Chemistry and Technology of Epoxy Resins PDF

Chemistry and Technology
of Epoxy Resins
Chemistry and Technology
of Epoxy Resins
Edited by
BRYAN ELLIS
Department of Engineering Materials
University of Sheffield
m
SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.
First edition 1993
© Springer Science+Business Media Dordrecht 1993
Originally published by Chapman & Hali in 1993
Softcover reprint ofthe hardcover lst edition 1993
Typeset in 1O/12pt Times by EJS Chemical Composition, Bath
ISBN 978-94-010-5302-0 ISBN 978-94-011-2932-9 (eBook)

DOI 10.1007/978-94-011-2932-9
Apart from any fair dealing for the purposes of research or private study,
or criticism or review, as permitted under the UK Copyright Designs and
Patents Act, 1988, this publication may not be reproduced, stored, or
transmitted, in any form or by any means, without the prior permission in
writing of the publishers, or in the case of reprographic reproduction only
in accordance with the terms of the licences issued by the Copyright
Licensing Agency in the UK, or in accordance with the terms of licences
issued by the appropriate Reproduction Rights Organization outside
the UK. Enquiries concerning reproduction outside the terms stated
here should be sent to the publishers at the Glasgow address printed on
this page.
The publisher makes no representation, express or implied, with regard
to the accuracy of the information contained in this book and cannot accept
any legal responsibility or liability for any errors or omissions that may
bemade.
A catalogue record for this book is available from the British Library.
Library of Congress Cataloging-in-Publication Data available.
Preface
Epoxy resins have been commercially available for about 45 years and now
have many major industrial applications, especially where technical
advantages warrant their somewhat higher costs. The chemistry of these
resins is fascinating and has attracted study by many very able scientists. The
technological applications of the epoxy resins are very demanding and there
are many new developments each year.
The aims of the present book are to present in a compact form both
theoretical and practical information that will assist in the study, research
and innovations in the field of epoxy resin science and technology. The
literature on epoxy resins is so vast that it is not possible to be encyclopaedic
and that is not the function of the present text. It is the editor's hope that the
selection of topics discussed will provide an up-to-date survey. There is some
overlap in the chapters but this is minimal and so each chapter is essentially
self contained.
As with all chemicals there are toxicological and other hazards. These are
not dealt with in this text since a little knowledge can be dangerous, but
material supplied can provide information regarding any safety precautions
that may be necessary. However, often these precautions are not onerous
and epoxy resins, or more specifically the hardeners, can be handled readily.
It is hoped that this text will provide an up-to-date outline of the science
and technology of epoxy resins and stimulate further research into unsolved
problems and assist further technological developments.
Bryan Ellis
Acknowledgements
As editor, I would like to thank all my co-authors for their contributions

without which there would be no textbook. I am grateful to the secretarial
staff of the Department of Engineering Materials, University of Sheffield
for their assistance. Finally, but not least, to the editorial staff of Blackie
who have transformed the manuscripts into a very well produced book.
B.E.
Note
For ease of reference a general index and a separate index of curving agents and hardeners
are provided.
Contents
1 Introduction to the chemistry, synthesis, manufacture and

characterization of epoxy resins 1
B. ELLIS
1.1 Epoxy resins 1
1.2 The chemistry of the epoxy group 7
1.3 The synthesis and manufacture of epoxy resins 14
1.3.1 Synthesis of epoxy compounds 14
1.3.2 Epoxy resins manufactured from epichlorohydrin 16
1.3.3 Oxidation of unsaturated compounds 26
1.4 Characterization of uncured epoxy resins 29
1.4.1 Chemical analysis 29
1.4.2 Quantitative analysis 30
1.4.3 Molecular structure 32
1.4.4 Physical properties 33
References 35
2 Curing agents for epoxy resins 37

W.R. ASHCROFf
2.1 Introduction 37
2.2 Nitrogen-containing curing agents 38
2.2.1 Aliphatic amines and derivatives 38
2.2.2 Cycloaliphatic polyamines and derivatives 51
2.2.3 Aromatic polyamines and derivatives 54
2.2.4 Catalysts and co-curing agents 56
2.2.5 Hydrazine and hydrazides 60
2.3 Oxygen-containing curing agents 60
2.3.1 Carboxylic acids and anhydrides 60
2.3.2 Phenol formaldehyde resins 65
2.3.3 Amino formaldehyde resins 66
2.4 Sulphur-containing curing agents 67
2.4.1 Polysulphides 68
2.4.2 Polymercaptans 68
2.5 Miscellaneous curing agent types 68
2.5.1 Amine-boron trihalide complexes 68
2.5.2 Quaternary phosphonium salts 69
2.5.3 Cationic salts 69
2.6 Summary 70
References 70
3 The kinetics of cure and network formation 72

B. ELLIS
3.1 Cure of epoxy resins 72
3.2 Gelation, network structure and glass transition temperature 74
viii CONTENTS
3.2.1 Branching theory 75

3.2.2 The glass transition 81
3.3 Techniques for monitoring cure 83
3.3.1 Monitoring cure 83
3.3.2 Direct assay of the concentration of reactive groups 84
3.3.3 Thermal analysis 86
3.3.4 Rheological changes during cure 87
3.4 Kinetics of cure 89
3.4.1 Introduction 89
3.4.2 Amine-curing agents 90
3.4.3 Carboxylic acid anhydrides 98
3.4.4 Diffusion control 99
3.5 Effect of cure on mechanical and related properties 102
3.5.2 Glassy moduli 104
3.5.3 Stress-strain curves and visco-elastic behaviour 106
3.5.4 Visco-elastic properties 109
3.5.5 Physical ageing 110
References 113
4 Additives and modifiers for epoxy resins 117

S.J. SHAW
4.1 Introduction 117
4.2 Diluents 117
4.2.1 Non-reactive diluents 118
4.2.2 Reactive diluents 118
4.3 Fillers 120
4.3.1 Physical/mechanical properties 121
4.3.2 Thermal characteristics 122
4.3.3 Shrinkage 124
4.3.4 Electrical conductivity 124
4.3.5 Viscosity 124
4.3.6 Toughness 125
4.4 Resinous modifiers 126
4.5 Flexibilisers/plasticising additives 128
4.5.1 Plasticisers 128
4.5.2 Reactive flexibilising additives 128
4.6 Elastomeric modification 131
4.6.1 Types of elastomeric modifiers 131
4.6.2 Compatibility and morphology 132
4.6.3 Toughening mechanisms 137
4.6.4 The hybrid modification approach 138
4.7 Thermoplastic modification 138
4.8 Miscellaneous additives 140
References 142
5 Fracture behaviour of epoxy resins 144

W.J. CANTWELL and H.H. KAUSCH
5.2 Linear elastic fracture mechanics (LEFM) 145
5.2.1 The G approach 146
5.2.2 The K approach 149
5.2.3 Crack opening displacement 150
5.3 Deformation mechanisms 150
CONTENTS ix
5.4 Modes of crack propagation 152
5.4.1 Stable brittle propagation 153
5.4.2 Unstable brittle propagation 154
5.4.3 Stable ductile propagation 156
5.5 Effect of test conditions 156
5.5.1 Temperature 156
5.5.2 Loading rate 158
5.6 Microstructural effects 159
5.7 Fractography of epoxy resins 161
5.8 Toughening strategies for epoxy resins 165
5.8.1 Mineral filler-modified epoxies 165
5.8.2 Thermoplastic-modified epoxies 166
5.8.3 Rubber-modified epoxies 167
5.8.4 Effect of particle size and volume fraction 169
5.8.5 Hybrid systems 171
5.9 Conclusions 172
References 172
6 Electrical properties of epoxy resins 175

G.P. JOHARI
6.2 Physical changes during the epoxy curing 176
6.3 Theoretical formalism for electrical properties 177
6.4 Dielectric effects of sol-gel-glass conversion 182
6.5 Ionic conductivity and sol-gel conversion 189
6.6 Time and temperature evolution of the dielectric properties 191
6.7 Chemical kinetics and dielectric behaviour 194
6.8 Curing and the high-frequency relaxation process 196
6.9 Ageing effects on electrical properties 200
6.10 Electrical applications of epoxy resins 203
References 204
7 Epoxy resin adhesives 206

S.J. SHAW
7.2 Theories of adhesion and wetting phenomena 207
7.2.1 Theories of adhesion 208
7.2.2 Wetting 210
7.3 Substrates and surface pretreatments 213
7.3.1 Solvent cleaning 214
7.3.2 Mechanical abrasion 214
7.3.3 Chemical pretreatment 215
7.3.4 Primers 217
7.4 Methods oftest 218
7.4.1 Conventional test techniques 219
7.4.2 Fracture mechanics approach 221
7.4.3 Environmental testing 222
7.5 Epoxy adhesive formulation 225
7.6 Properties of adhesive joints 228
7.6.1 Bulk properties of epoxy adhesive 229
7.6.2 Adhesive joint mechanical properties 233
7.7 Environmental effects 238
7.7.2 Moisture-related effects 239
x CONTENTS
7.7.3 Failure mechanisms 240

7.7.4 Approaches to improved durability 243
7.7.5 Other hostile environments 251
7.8 Applications 252
References 253
8 Composite materials 256

F.R. JONES
8.2 Fibre reinforcements 256
8.2.1 Manufacture of carbon fibres from polyacrylonitrile (PAN) precursors 259
8.2.2 Aramid fibres 263
8.2.3 Glass fibres 265
8.3 Fabrication of composites 267
8.3.1 The reinforcement form 267
8.3.2 Prepreg mouldings 267
8.3.3 Matrices for fibre composites 269
8.4 Mechanical properties of unidirectional laminates 276
8.4.1 Longitudinal modulus, E, 276
8.4.2 Longitudinal tensile strength, alll 277
8.4.3 Transverse modulus, E t 280
8.4.4 Transverse strength, a tll 281
8.4.5 Off-axis properties 283
8.5 Failure process in laminates 284
8.5.1 Crossplylaminates 284
8.5.2 Constraint cracking 287
8.5.3 Epoxy resin matrix failure strain 287
8.5.4 Thermal strains in crossply composites 287
8.5.5 Poisson-generated stresses and longitudinal splitting 289
8.5.6 Angle ply laminates 289
8.5.7 Discontinuous fibre composites 290
8.6 Effect of moisture on the performance of epoxy resins 290
8.6.1 Moisture absorption kinetics 291
8.6.2 Effect of resin structure 293
8.6.3 Effect of moisture on thermal residual strains 295
8.6.4 The combined effect of humidity and thermal excursions 296
8.6.5 Thermal spiking 297
8.7 Selection principles 298
8.8 Conclusions 298
8.9 Glossary of symbols 299
References 300
9 Coatings and other applications of epoxy resins 303

X.M. CHEN and B. ELLIS
9.2 Surface coatings 308
9.2.2 Surface preparation and primer 310
9.2.3 Solution coatings 311
9.2.4 Dip coats 313
9.2.5 Epoxy emulsions and other water-based coatings 314
9.2.6 Powder coatings 316
CONTENTS Xl
9.3 Industrial and related applications 318

9.3.1 Tooling 318
9.3.2 Civil engineering 319
9.3.3 Moulding compounds 321
9.3.4 Embedding 322
9.3.5 Miscellaneous 323
References 324
Index 327
Contributors
Dr W.R. Ashcroft Anchor Chemical (UK) Ltd, Clayton Lane,

Clayton, Manchester MIl 4SR, UK
Dr W.J. Cantwell Laboratoire de Polymeres, Ecole Polytechnique
Federale de Lausanne, CH-1014 Lausanne,
Switzerland
Dr X.M. Chen Department of Engineering Materials,
University of Sheffield, Sir Robert Hadfield
Building, Mappin Street, Sheffield S14DU, UK
Mr B. Ellis Department of Engineering Materials,
Professor G.P. Johari Department of Materials Science and
Engineering, McMaster University,
1280 Main Street West, Hamilton, Ontario,
L8S 4L7, Canada
Dr F.R. Jones Department of Engineering Materials,
Professor H.H. Kausch Laboratoire de Polymeres, Ecole Poly technique
Federale de Lausanne, CH-1015 Lausanne,
Switzerland
DrS.J.Shaw Materials and Structures Department,
DRA Aerospace Division, Farnborough,
HantsGU146TD, UK
1 Introduction to the chemistry, synthesis,
manufacture and characterization of
epoxy resins
B. ELLIS
1.1 Epoxy resins
The term 'epoxy resin' is applied to both the prepolymers and to the cured
/0"
resins; the former contain reactive epoxy groups, R-CH-CH2' hence their
name. In the cured resins all of the reactive groups may have reacted, so that
although they no longer contain epoxy groups the cured resins are still called
epoxy resins. The relative size of the market for epoxy resins is indicated in
Table 1.1 from which it can be seen that they are important industrial
polymers. Since they are more highly priced than other resins, they will only
find application when they have technical advantages. Many of the
applications involve high added value products.
Although the first products that would now be called epoxy resins were
synthesized as early as 1891 (see Dearborn et al., 1953; Lee and Neville,
1967a) it was not until the independent work of Pierre Castan in Switzerland
and Sylvan Greenlee in the United States that commercial epoxy resins were
marketed in the 1940s, although similar resins had been patented in the
1930s. The earliest epoxy resins marketed were the reaction products of
bisphenol A and epichlorohydrin and this is still the major route for the
manufacture of most of the resins marketed today, although there are many
other types ofresin available (section 1.3).
Pierre Castan investigated potential resins which could be readily
moulded at low pressures for the replacement of vulcanite as a denture base
material. The BPA-epoxy resins could be cured by reaction of epoxy groups
with phthalic anhydride without the evolution of low molecular species and
hence did not require high moulding pressures. An alternative acrylic resin
is now used for denture base and the patents were licensed to CIBA in 1942.
Epoxy adhesives and casting resins were marketed in the USA in 1946.
Greenlee working for Devoe and Raynolds produced resins which were
similar to those of Castan but with a somewhat higher molecular weight with
the objective of developing superior surface coatings. The epoxy coatings
developed by Greenlee offered improved adhesion, hardness, inertness and
thermal resistance compared with alkyd or phenolic resins. Following the
Table 1.1 Epoxy resin sales in the USA, 1990-1991
1990 1991
Resin 1000 tonnest Approximate' Approximate 1000 tonnest Approximate

sold relative cost relative value sold relative value
Epoxy 211 3.8 802 195 741

Phenolic 1286 1.3 1672.8 1163 1512
Unsaturated polyesters 558 2.9 1618 492 1427
Urea and melamine 761 667
Polystyrene 2297 1.15 2.64 2219 2552
Consumption in Japan in 1991 was 166 tonnes of epoxy resin and 386 tonnes of phenolic resin. t
• Relative to low density polyethylene = 1.0, Birley and Scott (1982).
t Modern Plastics International, 1992.
Me~ Cl
/ ° , I
HO-@-{\Q! OH + 2CH 2-CH-CH 2
Me Epichlorohydrin
Bisphenol A
(BPA) Alkali
MOH
/0, ---rc5'-¥~o /0,

CH2-CH-CH2-0~r~0-CH2-CH-CH2
Me
Diglycidyl ether of bisphenol A

(DGEBPA)
Catalyst
BPA
° ~M~_ OH M~__ °
CH'2~CH-CH2
1 0~{-&-0-CH2-tH-CH2 0--@-{----&-0-cH2-cH~CH2 t
Me Me
II
4 CHEMISTRY AND TECHNOLOGY OF EPOXY RESINS
Table 1.2 Number average molecular weight, Kin, of 'Ideal' epoxy

resins
n Mn No. of hydroxyl Epoxy equivalent

340 + 284n groups weight
0 340 0 170
1 624 1 312
2 908 2 454
10 3180 10 1590
first patent application (1948) Greenlee obtained about 40 patents for epoxy
resins. The major application of epoxy resins is still for surface coatings
which consumes about 50% of all epoxy resins produced. The relative use
pattern of epoxy resins is indicated in Table 1.3 and applications are
discussed in more detail in chapter 9.
Innovation in epoxy resin technology has involved the synthesis of epoxy
resins with specific characteristics and some of the more important are
discussed in section 1.3. Equally important has been the development of
hardeners which depend on their reactivity with epoxy groups. The more
important hardener systems are treated in detail in chapter 2. Much of the
chemistry of epoxy resins depends on the reactivity of the epoxy groups
which will be briefly outlined in the next section (1.2).
The cure of epoxy resins involves the formation of a rigid three-
dimensional network by reaction with hardeners which have more than two
·reactive functional groups, that is, functionality is f > 2. Often f ~ 4 for
common hardeners for BPA resins which often have an effective
functionality of two, but may be higher when the cure temperature is high
enough for the secondary hydroxy groups to react.
The cure of epoxy resins is complicated and it is useful to visualize the
process in several stages, which are illustrated in Figure 1.1, although except
Table 1.3 USA applications of epoxy resins (Modern Plastics International, 1992)
[1990] [1991]
1000tonnes % 1000tonnes %
Protective coatings 89 49 84 51
Electrical applications 25 14 22 13
Reinforced resins 14 7.5 13 8
Bonding and adhesives 13 7.5 12 7.25
Flooring 12 6.5 11 6.25
Tooling and casting 13 7.5 12 7.25
Other 15 8.3 12 7.25
Total 181 100 166 100
INTRODUCTION TO EPOXY RESINS 5
I
Viscous liquid
: rubbery .
Visco-elastic I visco-elastic
liquid solid glassy viscoelastic
Rheological
Newtonian fluid T = rj'Y solid
properties shear stress r
~~
rate of shear
power law fluid
r = myn
IJ i- const.
m and n ~ f (T" to)
complex shear modulus. G*

G* '" G' + iG"
-----------------------------T---------------------·
.... I
I
difunctional
~
Molecular epoxy ~ : formation
branched lof highly
structures pre polymer
crosslinked
.J-( molecules : incipient
network
hardener oligomers I 3D network
f. 4 I
I
average molecular I increasing
weight increases slowly

• highly branched : crosslinking
molecules I Ws < 1
ws= 1 I 0
W : 0 I wg >
Figure 1.1 The cure of epoxy resins.

Eo - E(tc)
The extent of reaction Xe = --"-----'..2-
Eo
where Eo is the initial concentration of epoxy groups and E(tc) is their concentration at cure time
tc' Tc is the cure temperature.
weight of soluble molecules
The sol fraction Ws = ---=---------
total weight of the sample
weight of cross-linked network, gel
The gel fraction Wgel = -------------=--
total weight of the sample
see chapter 3 for further discussion of cure processes.
for gelation, the process is continuous. Initially there is reaction between

epoxy and hardener reactive groups so that somewhat larger molecules are
formed. As cure proceeds, larger and larger molecules are formed but it
should be noted that the average molecular size is still small even when half
the reactive groups have reacted. When the molecular size increases as cure
progresses, some very highly branched molecules are formed and then more
and more highly branched structures develop. The critical point is gelation
when the branched structures extend throughout the whole sample. Prior to
gelation the sample is soluble in suitable solvents but after the gel point the
network will not dissolve but swells as it imbibes solvent. At the gel point
small and branched molecules are present which are soluble, hence the
curing sample contains sol as well as gel fractions. The gel initially formed is
weak and can be easily disrupted. To produce a structural material, cure has
to continue until most of the sample is connected into the three-dimensional
n~lwork so that the sol fraction becomes small and for many cured products
it has to be essentially zero. Cure and gelation are discussed in more detail
in ~hapter 3.
As cure proceeds there are major changes in the properties of the epoxy
resins. Initially the resin-hardener mixture is fluid and finally an elastic solid
is produced. The glass transition temperature of the curing resin increases as
cure proceeds and these changes can be represented in a time-temperature
trahsition (TTT) diagram introduced by Gillham (1986). Figure 1.2 is a
simplified version which illustrates the dominant effect of the onset of vitri-
fication as Tg increases to the cure temperature Te. For cure temperatures
well above Tg , the rate of reaction between the epoxy and hardener reactive
groups is chemically kinetically controlled. When ll.T = Te - Tg becomes
small the curing reactions become diffusion controlled, and will eventually
becpme very slow and finally stop. For products for which it is necessary to
ensure complete reaction of all epoxy groups it is a normal practice to post-
cure the resins at an elevated temperature. For successful application of
Cure
temperature
Te
Log te (cure time, tel
Figure 1.2 Simplified time-temperature-transition diagram (TIT). Tgo is the glass transition
temperature of the mixture of epoxy prepolymer-hardener-additive. For cure temperature
Tc < Tgo, the mixture is glass (1) and reaction of epoxy groups is inhibited. In glass (2) the
epoxy resin vitrifies before gelation. For glass (3) the glass transition temperature increases with
increased cross-linking of the network. Tg~ is the limiting glass transition temperature as the
concentration of epoxy group E -> O.
epoxy resins it is necessary to select a suitable hardener (see chapter 2) and

then cure the resin to attain a controlled network structure.
1.2 The chemistry of the epoxy group
The original discovery ofthe parent compound ethylene oxide, or oxirane, is

attributed to Wurtz who in 1859 published details of its synthesis from
ethylene chlorohydrin by reaction with aqueous alkali.
C('
I I
OH
I
CH 2-CH 2
Aq. alkali
------->
°
/"
CH 2-CH 2
This method is general for the synthesis of epoxy compounds but ethylene
oxide is now manufactured by direct oxidation of ethylene with air or oxygen
and a silver catalyst. Some of the early history of the synthesis and chemistry
of epoxy compounds has been discussed by Malinovskii (1965) with especial
reference to early Russian work. The synthesis of epoxy rings has been
discussed in detail by Gritter (1967) and Lewars (1984) and epoxy resins by
Tanaka (1988).
There are many methods for the synthesis of epoxy rings (Rosowsky,
1964), which are classified in Table 1.4. Although not the only ones, the
most important routes for the manufacture of epoxy resins are reaction of a
halohydrin with hydroxyl compounds, and the oxidation of unsaturated
compounds with a peracid. The first method is similar to the original
synthesis of ethylene oxide by Wurtz and may be illustrated by the reaction
of epichlorohydrin with hydroxyl compounds, such as phenols or aliphatic
alcohols.
/ "° CI
I
R-OH + CHz-CH-CH 2
Catalyst
------->
OHCl
I I
RO-CHz-CH-CH2
Alkali
/0"
R-0-CHz-CH-CH2 + MCI + H 20
MOH could be sodium or potassium hydroxide and has to be used in

stoichiometric concentration to neutralize the halogen acid, Hel in this case,
that is produced when the epoxy ring is formed. The application of these
reactions for the manufacture of epoxy resins is discussed in more detail in
section 1.3.
Unsaturated compounds can be oxidized to yield epoxy groups by the use
of peracids such as peracetic acid.
~ °
'c'-'c"- +
..-, °
CH 3-C?-
'OH
Table 1.4 Outline of the synthetic methods for epoxides
1. Oxidation of alkenes
a. Direct oxidation 02/catalyst, Ag/lOO-500°C

b. Inorganic oxidants e.g. chromic acid/permanganate
oII
c. Organic peroxides peroxy acids, R-C-OOH
d. Hydrogen peroxide H 20 2
2. From halohydrin
Hypohalous addition to alkene and then cyclodehydrohalogenation
X OH 0
HOX I I Alk I' .... / \ "
:C=C~ ~ :C-C~ M~~) "C-C.... + MX
3. From a-halocarbonyl compounds

a. Darzen's condensation
@-CHO +
Cl
I
0
II
CH 2-C-OEt ---
Na 0 0 0
II /\
@ - CH-CH-C-OEt
b. Reduction: lithium aluminium hydride
X-@-o Bf
III
C-CH2
c. Addition of alkoxide ion, followed by ring closure
o Sr 0- Sr
MeO.... 1 I
o
Me-~-tH2 MeO.... / \
MeO-)
"C-CH2 - "C-CH 2 + Sr-
Me Me
d. Addition of cyanide ion

Me o O-Me o
Me .... / \ "Me
I"H II CW I I"H
C-C ~ C-C "C-C....
Me .... Cl M e"ICN ....Cl CN H
e. Grignard reagent
OX
R3-~-t-Rl
R2
Compiled from: Gritter (1967); Lewars (1984); Malinovskii (1965) and Rosowsky (1964) who
give extensive lists of references to the very many routes available for the synthesis of epoxy
rings. Advances in Heterocyclic Chemistry should be consulted for more recent references.
This is Prileschaiev's reaction which was published in 1912 and has been
extensively discussed in several reviews (Rosowsky, 1964; Tanaka, 1988). A
possible mechanism for this reaction with epoxidation by perbenzoic acid is
I
CH
\I +
CH
I
I o
CH
I ........0
CH- + f-@
I H-O
Somewhat alternative mechanisms have been proposed which Gritter

(1967) discusses, pointing out that isotopically labelled peroxy oxygen is
incorporated into the ring. The mechanism proposed is
oII
Ar-C1 1r +'/C=L,
,-,/
o'0/c=o
where Ar is aryl group.

Many peracids (Swern, 1970) have been studied as well as reaction
conditions, since hydrolysis may occur with the formation of a glycol
I I
CH, Hydrolysis) CH-OH
I /0 I
CH CH-OH
I I
the formation of which has to be avoided to improve the yield of the epoxy
compound. The manufacture of epoxy resins from unsaturated compounds
by epoxidation with peracids is discussed in section 1.3.3.
The geometrical structure of the epoxy ring is planar, with bond angles
and lengths (Figure 1.3) determined from electron diffraction and
microwave spectroscopic measurements which have been discussed by
Lwowski (1984), Lewars (1984) and Peters (1967). The differences in bond
angle from dimethyl ether are considerable as can be seen from Figure 1.3,
and there must be considerable ring strain due to angular distortion from the
tetrahedral carbon angle of 109°. The dipole moment of simple ethers is 1.1
o o
(a) (b)
1.41 1\ 141pm
c c
c
""
""
H
e'
H
(c) (d)
F4
Figure 1.3 The structure of the epoxy ring. (a) Bond lengths and angles for ethylene oxide.
(b) C-O bond lengths and angles for dimethyl ether (the positions of the hydrogen atoms not
shown). (c) 'Bent' bonds in the strained epoxy ring. (d) Projection view of ethylene oxides.
to 1.3 D and that for ethylene oxide is 1.82 and 1.91 D in benzene solution
and the vapour phase respectively. The ionization potential of the oxygen
2p.n lone pair in ethylene oxide is 10.6 to 10.8 eV which is rather higher than
that of dimethyl (10.0 eV) and diethyl (9.5 eV) ethers which Peters (1967)
has compared with the ionization energies of other simple oxygen
compounds.
The electronic structure of three-membered rings poses difficult problems
o C
/" /" 3
since with a c e or C C bond angle of about 60° the 'normal' sp hybridiz-
ation with linear bonds between the ring atoms is impossible. The bonding in
cyclopropane has been discussed extensively and Halton (1991) in an
interesting review of ring strain in cyclic molecules considered the latest
evidence. The bonding in cyclopropane is abnormal with the interbond
angle compressed to about 60° which is required for ring formation with the
nuclei 'moving ahead' of the bonding electron density with the formation of
a 'bent' or so-called 'banana' bond as illustrated in Figure 1.3.
The geometry of the epoxy ring is similar to that of cyclopropane but
because of the electronegativity of the heteroatom the internal ring bond
angles and lengths are not equal (Figure 1.3). Parker and Isaacs (1959)
discussed the various structures that have been proposed for ethylene oxide.
It has been suggested that the carbon atoms are trigonally hybridized, that is
Sp2, and that one such orbital from each carbon atom overlaps with an
oxygen atomic orbital to form a molecular orbital which occupies the centre
of the ring (Figure 1.3). It is possible that the presence of the 'central' ring
orbital accounts for conjugation of the epoxy ring with other delocalized
electrons, which is shown by bathochromic shifts in UV (electronic) spectra
(Lewars, 1984) and NMR ring currents (Gritter, 1967). Of course such
conjugation does not prove that the electrons in the un substituted
compounds are de localized , and there has been dispute regarding the
possibility of ring currents in these compounds (Gritter, 1967). Although the
strain energies of cyclopropane and the epoxy ring are very similar, 27.43
and 27.28 kcaVmole respectively (Gritter, 1967), it may be that the bonding
is very different. For instance, from the NMR data compiled by Lwowski
(1984) the chemical shifts and coupling constants for the epoxy ring are
different from those for cyclopropane.
The many industrial applications of epoxy resins require the formation of
three-dimensional networks by reaction with suitable polyfunctiQnal
hardeners, which are discussed in detail in chapter 2. Many of these curing
reactions depend on the reactivity of the epoxy ring, which is very much
more reactive than the 'normal' non-cyclic ethers, R-OR', where Rand R'
are alkyl or aryl groups. In normal ethers the oxygen link is resistant to
attack by alkalis, ammonia or amines. Epoxy resins will react with some
aliphatic amines at room temperature; these amines may be used as curing
agents at ambient temperatures (see chapter 2). This increased reactivity of
cyclic ethers is due to the ring strain.
The chemistry of the epoxy ring has been reviewed comprehensively by
Parker and Isaacs (1959) and a more recent discussion is that of Lewars
(1984). The literature of heterocyclic chemistry (Katritsky and Weeds, 1966;
Katritsky and Jones, 1979; Belen'kii, 1988) including that of the epoxy ring
has been listed periodically; initially references to epoxy resins were listed
but not recently. However, these annotations are a useful source of
reference to information on the reactions of the epoxy group. The reactivity
of epoxy compounds is summarized in Table 1.5.
The reactions that are most important for the synthesis and cure of epoxy
resins involve either electrophilic attack on the oxygen atom or nucleophilic
attack on one of the ring carbon atoms. For the unsymmetrically substituted
/0"
epoxy compound R-CH-CHz, which occurs in most epoxy resins, several
factors determine ring opening reactions, such as, the nature of the reagent
or catalyst which may be either electrophilic or nucleophilic, the influence of
the substituent and the relative steric hindrance at the two carbon atoms.
With the general reagent HR', two possible products of ring opening may be
produced: 0 OH
/ " I
R-CH-CHz + HR' ~ R-CH-CHz-R'
Normal
OR t
I
~ R-CH-CHz-OH
Abnormal
Table 1.5 Some typical reactions of epoxy groups
1. Addition reactions by nucleophilic substitution

a. Hydroxylic nucleophiles
p" <j>H <j>RI

R-CH-CH2 + R10H -+ R-CH-CH 2
OH OH
0.2% H,SOJ100·C) I I
R-CH-CH 2
Rl = alkyl, alcohol; Rl = aryl, phenol

b. Acids
i. Mineral acids
o
/" HX
OH X
I I
X
I
R-CH-CH 2 -+ R-CH-CH 2 + R-CH-CH 2-OH
Product ratio depends on reactants and reaction conditions
X = F,CI,Br,I
ii. Carboxylic acids
C
Me-c~
ROO R
I II II I
I "
CI-CH-C-O-CH 2 -CH 2-C-O-CHCI
CI-CH-C-OH
iii. Ammonia and amines
R\ R2 =H: ammonia
R 1= H; R2 == alkyl, aryl: primary amine
R\ R 2 =alkyl, aryl: secondary amine
2. Electrophilic additions
a. Alkyl halides
~
-200"C
X==Br; R==Et
X=I; R=Me, Et, Pr
b. Isocyanates
c. Oxides of sulphur
S03
1,4 dioxane
3. Reduction
OH
I
/0, LIAIH
~~'
©
CH-CH, 100%
OH OH
I I
~-'"'
LiBH
~'" +
74% 26%
Products depend on the reducing agent
4. Oxidation
OH 0
HNO, ) I II
CI-CHz-CH-C-OH
l00'C
Compiled from: Gritter (1967); Lewars (1984); Malinovskii (1965) and Rosowsky (1964).
This is only a small selection of the very many reactions of epoxy groups that have been
reported. Base-catalysed rearrangements of epoxides are discussed by Yandovskii and Ershov
(1972), retention of configuration by Akhrem et al. (1968) and ring expansion by Grobov et al.
(1966).
a secondary alcohol or primary alcohol or a mixture. When HR' is an amine,

carboxylic acid or thiol, the 'normal' product, a secondary alcohol, is usually
formed.
o H OH
I
H
I
/ " I
R-CH-CH2 + H-N-R' --> R-CH-CH 2-N-R'
oII OH 0
I II
+ HO-C-R' -> R-CH-CH2-O-C-R'
OH
I
+ HSR' -> R-CH-CHz-SR'
In these reactions the attacking group donates a pair of un shared electrons

to the atom with the lowest electron density, that is, the methylene group
which is also less sterically hindered and hence the product is the secondary
alcohol. A mechanism for the base-catalyzed addition is regarded as
'borderline' SN2 (Streitweiser, 1956; Parker and Isaacs, 1959).
°
c-c
,\"
/".....
x-
~ [,I
C-C
0
0 .. -
" ]
,/:......
0-
~ -C-C-
I I
I
~ etc.
I
)(0- X
Acid catalysed addition involves proton attack on the ring oxygen atom,
, I~" \H+
"C C, ~ "C C, ~
'/~\+" x- [, ~
. ot
"c c,:
j H9
~ -C-C-
I
I
I
)(0- X
The kinetics of these reactions is discussed in detail by Frost and Pearson

(1961) and Parker and Isaacs (1959). The mode of addition may be reversed
when R is a strongly electron-attracting group or mixtures may be formed
depending on the importance of the various factors involved. Tanaka (1988)
discusses the configuration of the protonated epoxy group and Lewars
(1984) their basicity and calculations of proton affinity. The latter also
discusses other reactions which involve electrophilic attack on the ring
oxygen atom. These include Lewis acids, alkyl halides, halides, peroxy
acids, aldehydes and ketones. The reaction of epoxides with epoxides
initiated by electrophilic attack can lead to dimerization and also
polymerization (Price, 1967).
1.3 The synthesis and manufacture of epoxy resins
1.3.1 Synthesis of epoxy compounds

Ethylene oxide can be manufactured by the direct oxidation of ethylene
(Kilner and Samuel, 1960)
Oxidation /0"
0" catalyst
CHz-CH2
heavy metal hydride
90--!05'C
(60% conversion)
but unfortunately this process is not so efficient for higher olefines. The
methods that have been used for the synthesis of epoxy rings have been
discussed by Lwowski (1984) and Lewars (1984), the former in a general
review of the synthesis of small and large heterocyclic rings (Table 1.4). A
comprehensive review of the synthesis of epoxy compounds is that of
Tanaka (1988). The most important for the synthesis of epoxy resins are
(i) dehydrohalogenation of halohydrins and (ii) the epoxidation of alkenes

with peracids or their esters (section 1.2). Details of many actual syntheses
of epoxy resins including reaction conditions and yields are given by Sandler
and Karo (1977).
A very important intermediate for the production of epoxy resins is epi-
/0"
chlorohydrin, CI-CHr CH-CH2 , 2,3 epoxypropylchloride. This is because
the epoxy ring reacts readily with hydroxyl compounds such as phenols and
alcohols
/
°
"
R-OH + CHz-CH-CH2
aI I Catalyst
~ R-0-CHz-CH-CH2
OHO
I i I
MOH
Stoichiome~
/0"
R-0-CH 2-CH-CH2 + MCI
with the formation of a chlorohydrin. The epoxy group is formed by

dehydrochlorination with a stoichiometric amount of alkali, such as sodium
hydroxide.
The starting compound for the manufacture of epichlorohydrin IS
propylene which is chlorinated (Kilner and Samuel, 1960).
CH2 =CH ~ CI-CHz-CH-CI + CH2=CH-CHz-CI
I I
CH 3 CH 3
1,2 Dichloropropane Allyl chloride
The allyl chloride is converted to dichlorohydrin by reaction with

hypochlorous acid and is then de hydro chlorinated with lime to obtain
epichlorohydrin (Faith et al., 1965a).
OHICI
HOCI I I Ca(OH), )
CH2=CH-CHz-CI ~ CH 2-CH-CH 2CI
Industrially epichlorohydrin is either converted to glycerol by reaction with

sodium hydroxide or isolated by steam stripping and purified by distillation.
An alternative route is from acrolein, produced by oxidation of propylene
(Faith et al., 1965a).
0,
CH2=CH-CH 3 Stea~ CH2=CH-CHO + H20
catalyst Acrolein
CuO 350°C (85% yield)
The acrolein is chlorinated to yield 2,3 dichloropropionaldehyde which is

reduced to produce glycerol /3,y-dichlorohydrin, which is then
dehydrochlorinated.
Cl CI OH
I I I
CH2=CH-CHO CI-CHz-CH-CHO -- CI-CH2-CH-CH2
Epichlorohydrin may also be produced from allyl chloride by epoxidation

with a peracid.
1.3.2 Epoxy resins manufactured from epichlorohydrin

Epichlorohydrin is used for the production of a range of epoxy resins (Table
1.6) because the epoxy group reacts readily with hydroxylic compounds in
the presence of an alkali catalyst, MOH, and then a new epoxy ring can be
formed by dehydrochlorination.
o O-Na+
I \
CH,-CH-CH,CI I
R-OH + NaOH ---+ RO-Na+ + H 20 ) R-0-CHr CH-CH2CI + H 20
I
(catalytic
concentration)
OHP
/0" NaOH I I
R-0-CHr CH-CH2 + NaCI ( Stoichiometric RO-CH2-CH-CH2
concentrations
For the production of epoxy resins, the hydroxylic compounds are

multifunctional and many such phenols have been studied as possible
precursors; the more important are listed in Table 1.7. Also, some mono-
functional phenols have been used for the manufacture of resin 'modifiers'.
However, the most important phenol used for the manufacture of epoxy
resins is the difunctional bisphenol A which was originally studied by Cast an
(see section 1.1).
1.3.2.1 Resins manufactured from bisphenol A. Bisphenol A or

2,2'bis(p-hydroxyphenyl)propane is produced from acetone and phenol
with an acid catalyst such as 75% sulphuric acid or dry hydrogen chloride
Me
-@-~=-@-e
HO-@ + ~=O + @ - O H 50"
.
H2S0 4
• HO () C () OH
I
Me Me
acid gas (Faith et at., 1965b). The reaction conditions will depend on the
design of the production unit (Materials and Technology, 1972). The purity
of the product is high, >95% p,p'-isomer; the other isomers formed are o,p'
and 0,0'. For resin manufacture the p,p' isomer content should be at least
98%. The light yellow colour of some epoxy resins may be due to trace
impurities in the bisphenol A, such as iron, arsenic and highly coloured
organic compounds. Other names for bisphenol A are 4,4' -isopropylidene
diphenol and diphenylolpropane (DPP).
When a large excess of epichlorohydrin is reacted with bisphenol A with a
stoichiometric amount of sodium hydroxide at about 65°C the resin
produced contains about 50% diglycidyl ether of bisphenol A, DGEBA
Table 1.6 Epoxy resins derived from epichlorohydrin
1. Phenols
a. Many difunctional phenols have been investigated. BPA is the most important, others
are listed in Table 1.7
b. Monofunctional phenols: modifiers for epoxy resins
c. BP A resins: esterified with fatty acids
d. Halogen-substituted phenols
2. Alcohols
a. Multifunctional
/0.., /0"
l,4-butanediol~ CH2-CH-CH2-0-(CH2)4-0-CH2-CH-CH2
/0"
/0" yH2-O-CH,-CH-CH,
glycerol~ CH,-CH-CH,-O-yH /0"
CH,-O-CH,-CH-CH,
b. Monofunctional: Resin modifier
/0"
Butanol~ CH 3 -(CH,h-0-CH,-CH-CH 2
3. Phenolic and related resins

a. Phenol: formaldehyde novolac
b. Cresol: formaldehyde novolac
4. Carboxylic and fatty acids

a. Phthalic acid~ glycidyl esters
@t
o ° il
C-O-CH2-CH-CH~
rr-O-CH2-C~/.cH2
°
'\
° °
b. Long-chain acids ~ epoxy resin esters
c. Acrylic acid
BPA resins + CH,=<;"H ~ epoxy acrylates
C=O
I
OH
Acrylic acid
5. Nitrogen compounds °
a. Amines
aniline ~ @-N/ - CH,-CH~CH,
"\
CH2-CH-CH~
° -
'\
O-CH,-CH-CH, P"
p-amino phenol ~ ~ - -
/0, ~
CH2-CH-CH2-N-CH2-CH-CH~
/0,
Table 1.6 Continued
b. Cyanuric acid
c. Halogenated amines, e.g. tribromo aniline
6. Miscellaneous
Many other compounds have been reacted with epichlorohydrin. This table is a guide to
some of the more important. Other chlorohydrins and halohydrins may also be used,
MeO
\ / \
e.g. methylepichlorohydrin CI-CH 2-C-CH 2
(BADGE) and the reaction may be represented formally as
Me~
HO--@-{---&-OH
Me
With 'pure' epichlorohydrin it is necessary to add about 5 w/o water to

accelerate the reaction. Recovered and recycled epichlorohydrin usually
contains sufficient water. The excess of epichlorohydrin is required to limit
the production of higher molecular weight products. It is obvious that
DGEBA will react with bisphenol A and so on with the formation of higher
molecular weight resins, which have the general formula shown on page 20.
Table 1.7 Phenols used for the synthesis of epoxy resins
1. Bisphenol A (BPA)
Me
HO-@-{%OH
Me
2. Bisphenol F
3. Tetrakis phenylolethane
4. Resorcinol
HO%OH
5. Methylolated phenol
CH 0H
2 e dCH,OH
HO-@-{~~H Me
--> diglycidylether the -OR groups accelerate amine cure
6. Brominated and fluorinated phenols

Br Br
"oj§>-I~o" HO -@- C~
0 {~OH
CF1
Br Br
A range of species is present in any specific reaction product and commercial

resins are available with 0 :5 ii :5 14, and may be higher where ii is the
average number of structure units in the species. A major variable which
determines the molecular weight distribution and hence the average
molecular weight is the ratio of epichlorohydrin to bisphenol A, but other
reaction conditions will affect the product obtained, as will be discussed. For
o
I
:£
u
I
::r: ::r:
o-u
I
:£
u
I
o
.~.
~~
o
I
the production of higher molecular weight resins the chain extension process
is used (section 1.3.2.4). Batzer and Zahir (1975, 1977) have discussed in
detail the molecular weight distributions of epoxy resins prepared by
different processes (Ravindrath and Gandhi, 1979).
Although the commercially produced resins have a distribution of chain
lengths it is possible to obtain the DGEBA by molecular distillation and it
can also be crystallized. Pure DGEBA is a solid which melts at 43°C. Also in
liquid commercial resins that have been stored for prolonged periods some
crystallization of DGEBA will occur and the resin appears cloudy. Outlines
of the production process for the manufacture of a low molecular weight
resin are given by McAdams and Gannon (1986) and also Savla and Skeist
(1977c). Preparative methods are given by Sandler and Karo (1977).
Not all of the species present in commercial resins are diepoxides; side
reactions can occur when epichlorohydrin is reacted with bisphenol A.
These can become more important when higher molecular weight resins are
produced. The important side reactions are
i. Hydrolysis of epoxy groups
/0" OH\OH
NaOH 1 1
-CH 2-CH-CH2 ----> -CH2-CH-CH 2
H 20
a-Glycol formation
ii. Formation of bound chlorine

a. Reaction of epichlorohydrin with secondary alcohol
1
~0-CH2-CH-CH2
CI
+ CH
°
/ "
2-CH-CH2CI
1
~ ~0-CH2-CH-CH2
CI
1
OH
1
°
1
CH 2
1
/
CH=OH
CI CH 2 CI
1
~ 0-CH2-CH-CH 2
1
° 1
CH 2
1
CH...o
1 "
CH 2
b. Abnormal addition of phenolic hydroxyl
~OH+
lll. The presence of saponifiable or hydrolysable chloride is due to

incomplete dehydrochlorination
OHCl
I I
~CHrCH-CH2
iv. The formation of branched molecules due to reaction of epichloro-

hydrin with secondary hydroxyl groups.
The last traces of inorganic ions in the liquid resins may be difficult to
remove but their concentration can be minimized by suitable washing
procedures. The actual concentration of trace amounts of residual material
may affect the reactivity of the resins. Also, volatile low molecular weight
compounds have to be removed since only small amounts, ca. 1 w/o, can
significantly reduce the viscosity of the resin. Liquid resins find applications
which include coatings, castings, tooling and adhesives, and are also used as
starting materials for the manufacture of higher molecular weight and
modified resins.
1.3.2.2 Higher molecular weight bisphenol A resins. It is obvious that the

bisphenol A/epichlorohydrin ratio is important for control of the average
molecular weight of the resins produced with the repeat unit so that the
larger the value of n the smaller the epichlorohydrinibisphenol A ratio
required. The purity of the reactants is important and monofunctional
reactants are chain terminators and hence their concentration has to be
controlled. However, it is also necessary to optimize the reaction conditions
to achieve the degree of polymerization required. For the production of
oligomers with 1 :::; n :::; 4 the so-called 'Taffy' process may be used but for
much higher molecular weight polymers 3 :::; n :::; 20 the fusion or chain
extension process is used.
1.3.2.3 The 'Taffy' process. The charge is such that the epichlorohydrinl
bisphenol A ratio will yield a resin with the required value of the degree of
polymerization 1 :::; n :::; 4, so that the upper limit for the average molecular
weight produced is about 1500. A stoichiometric amount of caustic soda in
aqueous solution is added with stirring and the reaction temperature raised
to 45-50°C. As the molecular weight increases the reaction temperature is
raised to 90--95°C for about 80 min with maybe increased pressure and more
vigorous agitation. At the end of the reaction period the product is in a water
resin emulsion plus an alkaline brine. The epoxy resin is recovered by
separating the phases, washed with water to remove inorganic salts and the
water removed by drying at temperatures of up to 130°C and under vacuum.
For purification of the resin dissolution in an organic solvent may be
advantageous and removal of water may be assisted by the use of methyl
isobutyl ketone. However, it is then essential that the level of solvent
remaining in the resin is minimized. The recovery of the resin in this process
is a major disadvantage especially because of the large amounts of brine that
have to be removed. These problems are not encountered in the fusion or
advancement process (section 1.3.2.4).
In the 'Taffy' process integral values of n, the degree of polymerization,
are usually produced with n values of (0),1,2,3 whereas in the advancement
process n is even numbered. A typical product with a weight per epoxide
(wpe) ca. 500 and n = 2 has a softening point of about 70°C and the practical
limitation for this process is n = 3.7 and a softening point of 95-100°C.
These resins will not crystallize as will DGEBA as mentioned previously and
the determination of their softening point is discussed in section 1.4.
During the latter stages of this process the temperature is increased up to
about 95°C and hydrolysis of the epoxy groups occurs so that there is a
relative high a-glycol value of ca. 0.5 eq/kg. Although the formation of
a-glycol will reduce the molecular weight attained, their presence may be
advantageous since they catalyse the reaction between amine hardeners and
epoxy groups and hence accelerate cure.
1.3.2.4 Fusion process. This is also known as the chain extension process
for reasons which will become clear. The starting materials for the fusion
process are bisphenol A plus a liquid epoxy resin which is essentially
difunctional in epoxy groups, produced by the processes used for the
manufacture of DGEBA. Thus
/0,
CHo-CH-CHo-O
-@-~e-@-
C
/0"
O-CHo-CH-CHo
- - I --
Me
+
Me
Ho--@-F--@-OH
Me
The reaction is carried out at temperatures between 180 and 200°C for about
30 minutes with a nitrogen 'blanket' to minimize oxidative degradative
reactions. The reaction is very sensitive to type and concentration of catalyst
which must be very strictly controlled to ensure production reproducibility.
Thus there has been considerable development of specialized catalysts to
minimize chain branching and gelation by promoting reaction between
phenolic hydroxyl with epoxy groups and limiting the side reactions,
especially between epoxy-epoxy and epoxy-alcoholic hydroxyl groups
(McAdams and Gannon, 1986; Savla and Skeist, 1977c). It can be appreci-
ated that the chain extension process is a step-wise polymerization and the
relationship between average molecular weight and extent of reaction will
depend on the degree of stoichiometric equivalence in the reacting mixture,
i.e. the epoxy/-OH ratio and the extent of side reactions which effectively
remove either epoxy or -OH groups and hence alter the effective
epoxy/-OH ratio. The important parameters were investigated by Batzer
and Zahir (1975, 1977).
1.3.2.5 Phenoxy resins. From the previous discussion it would appear

that if side reactions are suppressed it should be possible to produce high
molecular weight resins from DGEBA and bisphenol A when there is exact
stoichiometric equivalence, i.e. epoxy-hydroxyl group ratio is unity, and
reaction conditions such that p ~ 1, where p is the probability that an
epoxy or hydroxyl group has reacted (Reinking et al., 1963). Thermoplastic
resins with average values of ii of about 100 and average molecular weights
of maybe 30 000 to 45 000 are available. These resins may not have terminal
epoxy groups, but each repeat unit has a secondary hydroxyl group which is
reactive, with, for example, isocyanates. Phenoxy resins are available in
solution for coating applications which harden when the solvent evaporates.
Granular resins may also be used for extrusions and injection mouldings
(McAdams and Gannon, 1986).
1.3.2.6 Resins from other phenolic compounds. Any multifunctional,

f ~ 2, phenolic compound is a potential starting material for the
manufacture of epoxy resins. Although many have been studied, only a few
have any commercial significance and these are 'formulated' to meet specific
requirements. Also, some monofunctional phenols have been reacted with
epichlorohydrin to produce monofunctional reactants for use as modifying
diluent agents. Some of the phenolic compounds that have been used as
starting materials are listed in Table 1.7 and only some of the more
important will be briefly mentioned.
The dihydric phenol which is produced by reaction of phenol with
formaldehyde is called bisphenol F.
OH
2 @ H
I
+ C=O
I
H
______. HO@-~-@OH
H
Mixed isomcrs
Bisphcllol F
Resins can be manufactured from bisphenol F by similar methods to those

used for bisphenol A and epichlorohydrin with a catalyst such as NaOH.
These resins have lower viscosities than the equivalent DGEBA.
Phenolic 'novolac' resins are manufactured by the reaction of phenol with
formaldehyde with P/F of :::::: 0.8 with the use of an acid catalyst. These
novo lac resins may be represented by the 'idealized' structure (Kopf, 1988),
ti LJ OH
~CH'l2JrI,--@J
n
with n depending on reaction conditions and P/F ratio. Not all the
substitution is in the ortho position but the reactivity of the ortho position is
higher than that of the para because of the 'activating' effect of the phenolic
group.
The reaction conditions used to manufacture epoxy-novo lacs is similar to
those used with BPA resins and the idealized structure of the product is
There are considerable 'variants on a theme' with this type of resin; use of
excess epichlorohydrin minimizes the reaction of phenolic hydroxyl with the
epoxy groups attached to the novolac resin and limits the amount of
branching that can occur. Also it is essential that all phenolic hydroxyl
groups have reacted because their presence would adversely affect the
storage life of the resin and also volatiles would be formed during cure. The
epoxy novo lacs have improved thermal and chemical resistance compared
to the BPA resins.
It is possible to partially esterify novolac resins before reaction with

epichlorohydrin or esterify afterwards for ester coating applications. Also
the novo lacs may be based on chlorinated phenols to obtain improved flame
resistance. Other halogenated phenolic compounds have also been used for
this purpose (Table 1. 7).
1.3.2.7 Other resins derived from epichlorohydrin. Many compounds

with reactive hydrogen atoms will react with epichlorohydrin and these have
been evaluated as potential resins. Some of the more important are listed in
Table 1.6. Formally many react similarly to phenols but there are important
differences. For instance, the chlorohydrin formed by reaction of epichloro-
hydrin with a secondary alcohol is much more sensitive to alkali than
aromatic ether-chlorohydrin and hence caustic alkali cannot be used as a
catalyst or for dehydrochlorination. A Lewis acid, such as BF3 or SnCI4 , is
used to catalyse the reaction between epichlorohydrin and the hydroxylic
compounds. Caustic alkali would also catalyse the epoxy-epoxy reaction
which leads to polymerization of the product. Aluminates affect dehydro-
chlorination without catalysing further side-reactions. Some epoxy resins
derived from aliphatic alcohols that are commercially available are listed in
Table 1.6. Resins have also been prepared from cycloaliphatic alcohols such
as hydrogenated bisphenol A, tetracyclohexylethane and hydrogenated
novolacs.
Glycidyl esters are manufactured from acids, such as phthalic and
hydrogenated phthalic acids. The viscosities of these esters are lower than
bisphenol A type resins but have similar reactivities. Also, bisphenol A type
epoxy resins can be reacted with fatty acids to produce vehicles for surface
coatings (Savla and Skeist, 1977a; Tess, 1988).
Another interesting class of compounds, some of which are available
commercially, are derived from amines, such as aniline, or p-phenylamine,
and these are also listed in Table 1.6.
Other chlorohydrins than epichlorohydrin could be used to produce
epoxy resins. Methylepichlorohydrin, derived from isobutylene, has been
used to manufacture resins by reaction with bisphenol A and also polybasic
acids and novolac phenolic resins (Savla and Skeist, 1977b).
1.3.3 Oxidation of unsaturated compounds

The oxidation of unsaturated compounds has been studied extensively with
a variety of reagents (Batzer, 1964) of which the most important for the
synthesis of epoxy resins are organic peracids and their esters. The
mechanism of such epoxidations has been discussed briefly (section 1.2) and
there is an extensive literature (Swern, 1970; Tanaka, 1988) on the methods
of preparing the peracids which affect the yields of both desired product and
impurities formed. Preparation of the peracid in situ often offers
advantages.
Numerous peracids, aliphatic, aromatic and cycloaliphatic have been
evaluated. Peracetic acid is widely used, either in aqueous, or non-aqueous
media in homogeneous or heterogeneous phase systems. It has high
epoxidation efficiency and stability at ambient temperatures but there are
handling hazards and reaction mixtures can be potentially explosive. A
major advantage of the peracid route for the synthesis of epoxy inter-
mediates is that since no species containing chlorine are involved in these
syntheses the resins do not contain hydrolysable chlorine, and they are also
low in ash and ionic content. Thus these resins have better weathering and
ageing properties than 'conventional' epoxy resins and find application
where 'good' electrical properties are required.
However, some impurities are formed during oxidation with peracetic
acid (Greenspan and Gall, 1953; Lee and Neville, 1967b).
0
..... / \ .,.-
"C-C, + H2 O
- HO OH
,I I"
"C-C, Glycol
"C-C,
0
..... / \ ,/
+ HOR
- OH
'C-C::'OR
" , Ether
"C-C,
0
...... / \
./
+ CH3-C'OH
O
'i
- OH
,I "
0
II
"C -C::·O-C-CH3 Ester
These reactions can be minimized by the use of peracetic acid in which the
sulphuric acid used in its synthesis is neutralized. With this reagent natural
oils can be epoxidized but it is not satisfactory for the production of
compounds with more reactive epoxy groups, even when reaction
temperatures are kept low and reaction times short. However, the hydroxyl
groups formed are reactive and can be employed in the cure of these resins.
The epoxidation of unsaturated compounds with a peracid is used in the
manufacture of cycloaliphatic epoxy resins. Not only are these resins free of
hydrolysable chloride and inorganic salts (ash) they do not contain aromatic
compounds and hence are more stable to UV exposure than the bisphenol A
derived epoxy resins. The presence of aromatic rings in BPA resins increases
the UV absorption of the resins and also degradative processes occur by the
formation of conjugated structures (Atherton et al., 1982; Bellenger et al.,
1986).
The starting materials for the production of cyclo-aliphatic resins may be
synthesized by Diels-Alder addition of unsaturated compounds. This can be
illustrated by the dimerization of butadiene to yield vinyl cyclohexene which
can then be epoxidized with a peracid.
"CH 2
CH Diels-Alder
I +
CH
''cH 2
/"
~xidized
~~~acid
with
The product, vinyl cyclohexene dioxide is available commercially (Union

Carbide, 1989, ERL-4206). The Diels-Alder addition of butadiene with
aldhydes or ketones may also be epoxidized. For example, with acrolein two
useful products can be synthesized:
CHO
I
CH
+ II •
CH 2
Acrylaldehyde
(prop-2-CI1-I-al) Tetrahydrobenzaldehyde
Butadiene
(buta-l.3-dielle)
Peracetic
acid
j Peracetic
acid
3,4-epoxycyclohexylmethyl - 3,4 epoxy 2-(3,4 epoxycyclohexyl- 5,5 spiro-3,4 epoxy)

cyclohexane 'carboxylate cyclohexane metadioxane
Union Carbide, 1989 ERL-422I Union Carbide, 1989 ERL 4234
Ciba-Geigy CY-179 Ciba-Geigy CY-175
Other epoxy products can be produced by peracid epoxidation, such as
epoxidized polyolefines, oils and fatty acid esters. The last are used as
plasticizers and stabilizers for polyvinyl chloride compositions.
1.4 Characterization of uncured epoxy resins
The generalized structure of bisphenol A type epoxy resins can be

represented by the structure formula
cH~'cH-CH,r o@-r@-o-cH,-b::-CH,t@-r@-O-CH'-cH~'cH'
T Me I Me
and in general there are different species present with different values of n,
the number of structural units. A first approximation in characterizing this
resin is to determine the concentration of epoxy groups present since this is
needed to calculate the amount of hardener required. A similar
consideration will apply to the other types of epoxy resin that are
commercially available.
However, the composition of epoxy resins is more complicated than that
represented by the general structure and characterization involves much
more than only determination of the epoxy content. Thus, the most
important parameters that manufacturers of uncured epoxy resins provide
as sales specifications include as well as epoxy content, hydrolysable
chlorine, specific gravity, colour and for liquid resins their viscosity at
ambient temperature. For solid resins the softening point and a solution
viscosity are often specified. Some resin manufacturers give detailed
instructions for some analytical procedures; for instance Ciba Geigy (1988)
give test methods for the determination of epoxy content and readily
hydrolysable and total chlorine contents. A detailed treatment of the
analysis of uncured resins is given by Urbanski (1977) with experimental
details of methods for the qualitative identification of epoxy resins and
procedures for their quantitative analysis. The characterization of uncured
resins involves determination of their detailed structure. Some of the
methods of analysis applied to uncured resins will now be considered.
1.4.1 Chemical analysis

Qualitative test for epoxy resins are available and procedures have been
discussed by several authors (Urbanski, 1977; Braun, 1988; Jahn and
Geotzky, 1988a). For uncompounded resins infrared methods offer a
convenient method of identification. It is useful to establish a set of
reference spectra determined with a repeatable set of conditions even
though there are compilations of spectra available. Spot tests although
simple to execute require cross-checking to ensure reliability. For resin-

hardener mixtures and compounded resins, a combination of infrared and
NMR spectroscopy (Hadad, 1988a) as well as chromatrographic methods
(Me stan and Morris, 1984) may be required.
1.4.2 Quantitative analysis
1.4.2.1 Epoxy content. This is the commonest analytical measurement

that is carried out on uncured epoxy resins. Methods for the assay of epoxy
content involve the halogen-acid cleavage of the epoxy ring to yield a
halohydrin:
OH\X
I I
--> -CH-CHz
This reaction has been discussed previously and is conceptually obvious as a

possible method for the determination of the number of epoxy groups per
unit mass present. However, study of the published methods which
essentially rely on the reaction being exclusive and quantitative, show that it
is by no means easy to ensure the accuracy of these methods. It is essential to
follow the detailed procedure implicitly, and to ensure that specified
conditions are strictly adhered to (Dobinson et al., 1969; Urbanski, 1977).
1.4.2.2 Determination of hydroxyl groups. For all resin species with n ~

there will be a secondary alcohol present unless it has reacted with bisphenol
A which of course may lead to the formation of a branched molecule. Hence
determination of the concentration of secondary alcoholic hydroxyl groups
is important for the characterization of the structure of epoxy prepolymers.
Hydroxyl groups may also be potentially reactive with hardeners and their
concentration is required to calculate resin/hardener stoichiometry. Also,
hydroxyl groups may catalyse the reaction between hardener and epoxy
groups. This is very important for the kinetics of cure with amine hardeners
which will be discussed in chapter 3.
Other types of hydroxyl group which may be present are a-glycol units
formed by hydrolysis of an epoxy group, and also phenolic hydroxyl due to
incomplete reaction of the phenol when the resin was manufactured,
especially when larger concentrations of bisphenol A are used.
Common methods for the assay of hydroxyl groups are inapplicable
because the epoxy groups present interfere (Urbanski, 1977). It is possible
to use lithium aluminium hydride which only reacts with active hydrogen
atoms.
4 ROH + LiAIH4 --> LiAI(OR)4 + 4 Hz
The hydrogen evolved is determined volumetrically or by gas liquid

chromatography. The procedure is difficult to carry out quantitatively and
corrections have to be applied to allow for the presence of water or other
active hydrogen compounds. A more convenient method is to use an
acetylation procedure in which both hydroxyl and epoxy groups react and
the total hydroxyl concentration is determined by difference since the epoxy
concentration can be determined separately (Jahn and Goetzky, 1988b;
Urbanski, 1977).
The a-glycol concentration may be determined by the periodic acid
method (Stenmark, 1956).
HOIOH
I I
R-C--CH2 + 10; -> 10:; + RCHO + HCHO + H20
a-Glycol
Since the a-glycol is usually only present in small amounts it is necessary to

take precautions to obtain accurate results. Phenolic hydroxyl is also usually
only present in low concentration and although several methods (Jahn and
Goetzky, 1988c) have been suggested their accuracy must be assessed to
ensure reliable results are obtained (Urbanski, 1977). The esterification
equivalent-weight defined as the gram weight of epoxy resin required to
esterify one gram-equivalent of carboxylic acid is determined for resins that
are converted to epoxy esters. It can be determined by acetylation with
acetic anhydride in pyridium chloride solution. The epoxy group will react
with two acid groups and the hydroxyl groups will also be converted to esters
(Jahn and Goetzky, 1988c).
Infrared spectroscopic methods have also been suggested for the assay of
hydroxyl groups but the method requires calibration because of the effects
of hydrogen bonding on the spectra (Dannenberg, 1963).
1.4.2.3 Chlorine content. Although 'pure' DGEBA would have zero

chlorine content, commercial resins may have up to about 1% chlorine. Self-
extinguishing resins may have an organically bound chlorine content of up to
30%, and bromine- and fluorine-containing resins are also available. The
resins with high chlorine content may have lower thermal stability
particularly when cured with amine hardeners. Flame retardancy is a
specialized topic and resin suppliers can formulate compositions which are
satisfactory for particular applications.
In bisphenol A type epoxy resins the presence of chlorine will adversely
affect the electrical properties as would any polar impurity. The colour and
reactivity of these resins may also be adversely affected by the presence of
chlorine. The active chlorine may block reaction of lower base catalysts,
such as tertiary amines, and hence its concentration may be critical.
However, when organically-bound chlorine is present the resin will have a
reduced functionality and hence the cured resin will have a less tight
network.
The total chlorine present can be determined by the classical method using
a Parr bomb where the material is oxidized and then the chloride ion present
is determined as silver chloride (Jahn and Goetzky, 1988d). An alternative
procedure has been proposed by Urbanski (1977) in which the chlorine is
determined hydrolytically by heating for 2 hours with 0.5 N potassium
hydroxide in an ethylene glycol-dioxane solution, followed by potentio-
metric titration with silver nitrate in a dioxane-acetic acid solution. The
organically bound chlorine is obtained by correcting for any ionic chloride
ions that may be present, which may be determined by a standard pro-
cedure. In commercial resins, the inorganic chloride content should be very
low, but of course resin manufacturers have to monitor its presence to
ensure all of the salt formed by the dehydrochlorination reaction has been
effectively washed from the resin.
The organically bound chlorine will be in the forms known as (i)
hydrolysable and (ii) inactive. The presence of the former is due to
incomplete dehydrochlorination and the latter to either reaction of
epichlorohydrin with the secondary alcohol or abnormal addition of the
phenolic hydroxyl group as discussed previously (section 1.3). The
hydrolysable chlorine, also called saponifiable chlorine, can be determined
by reaction with excess caustic soda and then back titration with standard
hydrochloric acid (Jahn and Goetzky, 1988d). The inactive chlorine can
be calculated by difference from the total organically bound and the
hydrolysable chlorine.
1.4.3 Molecular structure

Methods such as infrared and nuclear magnetic resonance (NMR) spectro-
scopy can be used to identify the type of epoxy resin, that is for determi-
nation of the structure of the repeat unit. These methods can also be used to
identify the hardeners and other additives present in epoxy resin-hardener
mixtures. There are many texts which deal with the theory and practice ofIR
and NMR spectroscopy and Rabek (1980a) discusses their applications in
polymer chemistry.
Hadad (1988b) gives a very good review of the application of various
chromatographic methods for the determination of the molecular weight
distribution (MWD) of epoxy resins. Batzer and Zahir (1975,1977) used gel
permeation chromatography to analyse the products obtained from the
synthesis of epoxy resins using a range of reaction conditions. They
compared the MWD actually produced with theoretical predictions and
suggested that deviations from the model are due to the higher reactivity of
epichlorohydrin with phenolic hydroxyl compared with the arylglycidyl
ether group. Chromatographic methods are discussed by Rabek (1980b) and
their application to the analysis of epoxy resins by Mestan and Morris
(1984).
1.4.4 Physical properties
1.4.4.1 Softening point. Although pure DGEBA will crystallize, even

when in a mixture with other higher molecular weight species, in general
epoxy resins will not crystallize. When cooled they form glassy, amorphous
solids with glass transition temperatures (Ives et al., 1971) which have been
determined using broad-line NMR spectroscopy (Ellis, 1972). The usual
method that is used is measurement of a softening point (Lever and Rhys,
1968) which is defined by the procedure used for its determination. Since the
resins soften over a range of temperatures, there will not be exact agreement
between methods. There are two standard specifications, the ball and ring
and Durran's method which yield somewhat different results. In the ball and
ring method the sample is held in a horizontal ring with a steel ball on top of
it. The temperature is raised at a prescribed rate and the softening point is
that temperature when the ball has moved downwards by a distance of
2.5 cm (1 inch) under its own weight. Durran's 'mercury method' involves
warming a known weight of resin in a standard test tube and then cooling it
so that it solidifies. A specified quantity of clean mercury is placed on top of
the resin, which is then warmed. The Duran's softening point is the
temperature at which the mercury sinks to the bottom of the tube,
(Stenmark and Weiss, 1956).
1.4.4.2 Viscosity. Control of the processing of epoxy resins requires

information on their rheological properties. A good general discussion of
rheological measurements is given by Schoff (1988a). For the liquid resins
which may have high viscosities it is very difficult to mix with hardeners so
that a uniform molecular dispersion is obtained. With some hardeners,
reactions occur even at room temperature and hence the viscosity increases
before mixing is complete and microgelation may occur and of course locally
the epoxy/hardener stoichiometry is out of balance. Even with hardeners
which require elevated temperatures it is often necessary to warm the resin
to reduce its viscosity so that a molecular mixture can be obtained. Some of
the morphological features observed in cured epoxy resins can be attributed
to inefficient mixing of the hardener with the epoxy prepolymer. For many
applications, such as casting, undiluted epoxy adhesives and laminating
resins it is essential that their viscosity during fabrication is low. Also, air
bubbles which rise to the surface very slowly, if at all, may be very
troublesome when the viscosity increases due to the onset of cure.
The viscosity of epoxy resins depends on their molecular structure,
molecular weight distribution, and is often a very sensitive function of
temperature. For 'compounded' resins which are liquid at room tempera-
ture, their viscosity is measured without dilution but for solid resins the
viscosity of a 40 w/o solution in diethylene glycol monobutyl ether is often
measured. Solution viscosities (Rabek, 1980c) can be measured by bubble
viscometers such as the Gardner-Holdt tubes (Schoff, 1988b), capillary

viscometers (lves et al., 1971) or instrumentally with a spindle viscometer
such as the Brookfield (Lee and Neville, 1967c). For these measurements
the concentration of the solution and its temperature must be kept to
specified values to obtain satisfactory measurements.
For measurement of the viscosity of the liquid resins, the Hoeppler
viscometer is often used. The principle of this viscometer is measurement of
the time for a ball to descend a tilted tube containing the resin. Both the size
of ball, tube diameter and angle have to conform to the specification. Even
with these precautions measured viscosities may differ by maybe 15% from
the values determined by other methods such as flow through tubes or
rotational viscometers. When only one set of conditions is used with a
Hoeppler viscometer the viscosity is determined at essentially a single
average rate shear. This would be satisfactory for Newtonian fluids but
higher molecular weight resins will have a power law relationship between
shear stress, r, and rate of shear, y
T = m yn
where m and n are power law parameters and both are temperature-
dependent. When the power law exponent n = 1, the fluid is Newtonian
with m = YJ, the viscosity. For polymer fluids n is less than unity and the fluid
is pseudoplastic or shear thinning. For epoxy resins such as Epon 828 the
value of n at ambient temperatures is approximately unity (Chen and Ellis,
1992), but with higher molecular weight resins it is necessary to confirm that
they are Newtonian. The most suitable instrument is a cone and plate
viscometer but use of a Hoeppler viscometer with timing of the rate of
descent of the ball for two different angles of tilt may give sufficient data to
determine m and n (Barnes, 1980).
The temperature dependence of viscosity of many simple liquids is
Arrhenius, YJ = YJe exp [EYJIRT]' so that a plot of In YJ vs liT is linear.
However, in some cases it may be satisfactory to represent the temperature
dependence on a log-linear plot, so that:
log 1] = A + BT
where A and B are empirical constants. Also, the Arrhenius temperature

dependence would not be expected to hold when the viscosity is measured at
temperatures insufficiently higher than the glass transition temperature of
the resin.
The structure of the epoxy resin affects its viscosity, for example the
aromatic, bisphenol A type epoxy resins have higher viscosities than the
equivalent cycloaliphatic which are themselves higher than linear alophatic.
Other measurements that may be made on uncured resins include, density
and refractive index which is very sensitive to the structure and purity of the
resin. Colour is also determined by reference to colour standards. The
vapour pressure, especially of solutions may be important as is the flash
point.
References
Akhrem, AA., Moiseenkov, AM. and Dobryin, V.N. (1968) Russ. Chem. Rev. (Eng. Trans.)
37,448--462.
Atherton, N.M., Banks, L.G. and Ellis, B. (1982) J. Appl. Polym. Sci. 27,2015-2023.
Barnes, H.A. (1980) In Rheometry: Industrial Applications, ed. Walters, K. Research Studies
Press, John Wiley and Sons, Chichester.
Batzer, H. (1964) Chem. and Ind. 5, 179.
Batzer, H. and Zahir, S.A. (1975) J. Appl. Polym. Sci. 19,585-617.
Batzer, H. and Zahir, S.A. (1977) J. Appl. Polym. Sci. 21, 1843-1857.
Belen'kii, L.I. (1988) Adv. in Heterocyclic Chemistry 44,269-396. Synthesis of epoxy groups
p. 304. Reactions of epoxy groups pp. 303-330.
Bellenger, V., Verdu, J., Francillette, J. and Hoarau, P. (1986) Polymer Commun. 27,
279-28l.
Birley, A.W. and Scott, M.J. (1982) Plastics Materials, Properties and Applications, Leonard
Hill, Glasgow.
Braun, D. (1986) Simple Methods for Identification of Plastics 2nd edn. C. Hanser Verlag,
Miichen, pp. 27, 84 and PIT chart.
Chen, X.M. and Ellis, B. (1992) Unpublished measurements.
Ciba-Geigy (1988) Araldite for Surface Coatings, Laboratory Manual, Ciba-Geigy Plastics,
Cambridge.
Dannenberg, H. (1963) Soc. Polym. Eng. Trans. (Jan. 1963) p. 78.
Dearborn, E.C., Fuoss, R.M., Mackenzie, A.K. and Shepherd, R.G. (1953) Ind. Eng. Chem.
45,2715.
Dobinson, B., Hofmann, W. and Stark, B.P. (1969) The Determination of Epoxide Groups,
Pergamon Press, Oxford.
Ellis, B. (1972) Amorphous Materials Ch. 38, eds. Douglas, R.W. and Ellis, B. Wiley-
Interscience, London. pp. 375-388.
Faith, W.L., Keyes, D.B. and Clark, R.L. (1965a) Industrial Chemistry 3rd edn. John Wiley,
New York. p. 404. (1965b) p. 152.
Frost, AA. and Pearson, R.G. (1961) Kinetics and Mechanism 2nd edn. Wiley, New York,
pp.288-307.
Greenspan, F.P. and Gall, R.J. (1953) Ind. Eng. Chem. 45,2722-2726.
Gillham, J.K. (1986) Encylopedia of Polymer Science and Engineering, 2nd edn. John Wiley,
New York, pp. 519-524.
Gritter, R.J. (1967) The Chemistry of the Ether Linkage, Ch. 9, section II, ed. Patai, S.
Interscience Publishers, pp. 381-390.
Grobov, L.N., Sineokov, A.P. and Etlis, V.S. (1966) Russ. Chem. Rev. (Eng. Trans.) 35, 67l.
Hadad, D.K. (1988a) In Epoxy Resins, Chemistry and Technology, 2nd edn, Ed. May, C.A.
Marcel Dekker Inc, New York, pp. 1111-1124. (1988b) pp. 1096-111l.
Halton, B. (1991) Adv. in Ring Strain in Organic Chemistry 1,1-17.
Ives, G.C., Mead, J.A. and Riley, M.M. (1971) Handbook of Plastics Test Methods, IIiffe
Books, London.
Jahn, H. & Goetzky, P. (1988a) In Epoxy Resins, Chemistry and Technology, 2nd edn, ed.
May, C.A. Marcel Dekker Inc, New York, pp. 1050-1056. (1988b) p. 1067. (1988c) p. 1069.
(1988d) pp. 1070-1072.
Katritsky, A.R. and Jones, P.M. (1979) Adv. in Heterocyclic Chemistry 25, 303-391, see
327-329.
Katritsky, AR. and Weeds, S.M. (1966) Adv. in Heterocyclic Chemistry 7, 225-299, see
232-233.
Kilner, E. and Samuel, D .M. (1960) Applied Organic Chemistry, Macdonald and Evans, p. 54.
Kopf, P.W. (1988) Enc. Polym. Sci. Eng. Vol. 11. Wiley-Interscience, New York, pp. 45-95.
Lee, H. and Neville, K. (1967 a) Handbook of Epoxy Resins, McGraw-Hili, New York, sections
1-2 to 1-4. (1967b) sections 3-6 to 3-8 (1967c) sections 4-13 to 4-15.
Lever, A.E. and Rhys, J.A. (1968) The Properties and Testing of Plastics Materials, Temple
Press Books, Feltham, U.K. pp. 166-169.
Lewars, E.G. (1984) Comprehensive Heterocyclic Chemistry, ed. Katritsky, A.R and Rees,
C.W. Vol. 7, section 5.0.5.4, ed. Lwowski, W. Pergamon Press, Oxford, pp. 114-118.
Lwowski, W. (1984) Comprehensive Heterocyclic Chemistry, eds Katritsky, A.R and Rees,
C.W. Vol. 7, Part 5, ed. Lwowski, W. Pergamon Press, Oxford, pp. 1-16.
McAdams, L.V. and Gannon, J.A. (1986) Ency. Polym. Sci. Eng. Vol. 6, Wiley-Interscience,
New York, pp. 222-382.
Malinovskii, M.F. (1965) Epoxides and their Derivatives, Section II. Israel Programme for
Scientific Translations, Jerusalem, pp. 38-101.
Materials and Technology (1972) Vol. IV. Petroleum and Organic Chemicals Longman,
London,pp.322-324.
Mestan, S.A. and Morris, C.E.M. (1984) Rev. Macromol. Chern. Phys. C24(l), 117-172.
Modern Plastics International (1982) Jan. p. 48.
Parker, R.E. and Isaacs, N.S. (1959) Chern. Rev. 59,737-799.
Peters, D. (1967) The Chemistry of the Ether Linkage, Ch. 1, ed. Patai, S. Wiley-Interscience,
New York (see especially p. 14).
Price, c.c. (1967) The Chemistry of the Ether Linkage, Ch. 11, ed. Patai, S. Wiley-
Interscience, New York, pp. 501-503.
Rabek, J.F. (1980a) Experimental Methods in Polymer Chemistry Wiley, Chichester, UK.
Ch. 15, pp. 221-254; Ch. 20, pp. 299-331. (1980b) Ch. 23,24,25. pp. 368-442. (1980c)
Ch. 9, pp. 123-141.
Ravindrath, K. and Gandhi, K.S. (1974) f. Appl. Polym. Sci. 24, 1115-1123.
Reinking, N.H., Barabeo, A.E. and Hale, W.F. (1963) f. Appl. Polym. Sci. 7,2135-2144.
Rosowsky, A. (1964) Heterocyclic Compounds with Three- and Four-membered Rings, Part I,
Ch. 1, ed. Weissberger, A. Wiley-Interscience, New York, Section II, pp. 31-181.
Sandler, S.R. and Karo, W. (1977) Polymer Synthesis, Vol. II, Ch. 3. Academic Press, New
York, pp. 75-113.
Savla, M. and Skeist, I. (1977a) In Polymerization Processes. ed. Schildkrecht, C.E. Wiley-
Interscience, New York, Ch. 16, p. 608 and Table 16.8. (1977b) p. 605. (1977c) pp. 589-591.
Schoff, c.K. (1988a) Ency. Polym. Sci. Eng. 2nd edn, Vol. 14. Wiley-Interscience, New York,
pp. 454-541. (1988b) pp. 64-71.
Stenmark, G.A. and Weiss, F.T. (1956) Anal. Chern. 28, 1784.
Streitweiser, A. (1956) Chern. Rev. 56,571-752, see pp. 682-690.
Swern, D. (1970) Organic Peroxides, Vol. 1, Wiley-Interscience, London, pp. 355-533.
Tanaka, Y. (1988) Epoxy Resins, Chemistry and Technology, 2nd edn, Ch. 2. ed. May, C.A.
Marcel Dekker, New York, pp. 9-283.
Tess, RW. (1988) In Epoxy Resins, Chemistry and Technology, 2nd edn, ed. May, C.A.
Union Carbide (1989) Cycloaliphatic Epoxide Systems Union Carbide Chemicals and Plastics
Technology Corporation.
Urbanski, J. (1977) Handbook of Analysis of Synthetic Polymers and Plastics Ch. 10. ed.
Urbanski, J. Ellis Horwood, pp. 295-318.
Yandorskii, V.N. and Ershov, B.A. (1972) Russian Chern. Rev. (Eng. Trans.) 41, 403-410.
2 Curing agents for epoxy resins*
W.R. ASHCROFT
2.1 Introduction
In order to convert epoxy resins to hard, infusible thermoset networks it is

necessary to use crosslinking agents. These crosslinkers , hardeners or curing
agents as they are widely known, promote cross-linking or curing of epoxy
resins. Curing can occur by either homopolymerisation initiated by a
catalytic curing agent or a polyaddition/copolymerisation reaction with a
multifunctional curing agent. Specific curing mechanisms will be detailed
within the ensuing sections.
Following publication of the first patents describing cure of epoxy resins
and the subsequent commercialisation, the choice of curing agents and
catalysts has multiplied. Originally only a relatively few types were known
and these included simple amines, acid anhydrides, phenol-formaldehyde
and amino-formaldehyde resins. Nowadays there is an extended selection
available and there are two reasons for the growth. Firstly, as alternative
types of epoxy resin were developed so was the potential for other organic
chemicals to function as curing agents recognised. This led to the
introduction of polyfunctional amines, polybasic carboxylic acids,
mercaptans and even inorganic chemicals: hence the diversity of product
types, as previously described by Mika and Bauer (1973), which have been
Table 2.1 Estimated world market for curing agents by type
Weight Relative use

(mmlb) (%)
Amines - aliphatics 45 10
- cycloaliphatics 30 7
- aromatics 15 3
- dicyandiamide 10 2
Polyamides 75 16
Polyamidoamines 30 7
Phenol- and amino-formaldehyde resins 75 16
Carboxylic acid functional polyesters 100 22
Anhydrides 55 12
Polysulphides and polymercaptans 15 3
Catalysts 10 2
Total 460 100
* Anchor Chemical (UK) Ltd-An Air Products Company. © Air Products and Chemicals,
Inc. 1992.
offered. Secondly, and consecutively with end application development for

epoxy systems there has been a need to modify available curing agents to
meet specific ultimate requirements - hence the comprehensiveness of the
speciality curing agent manufacturers' ranges. Table 2.1 describes the
estimated relative world market usage in 1989 for the commercially most
significant curing agent types.
2.2 Nitrogen-containing curing agents
2.2.1 Aliphatic amines and derivatives

Ethyleneamines, polyamides, polyamidoamines, imidazolines, adducts
with epoxy functional materials, Mannich bases, association salts, ketimines
and acrylonitrile adducts, propylenamines, higher alkylenamines, fatty
diamines, polyetheramines, araliphatic amines, alicyclic amines, polymeric
amines, dicyandiamide and biguanides.
2.2.1.1 Ethyleneamines and derivatives. Ethyleneamine epoxy curing

agents are derived from ethylene and ammonia and were amongst the first
curing agents reported by Greenlee (1952) for use with epoxy resins. Typical
examples are:
H 2N(CH2CH 2NH)xH
x = 1, EthyleneDiAmine (EDA[107-15-3J)*;
x = 2, DiEthyleneTriAmine (DETA[111-40-0J);
x = 3, TriEthyleneTetrAmine (TETA[112-24-3J);
x = 4, TetraEthylenePentAmine (TEPA[112-57-2J);
x = 5, PentaEthyleneHexAmine (PEHA[4067-16-7J)
* Chemical abstract services number (CAS No.).
The mechanism of cure with ethyleneamines typifies the reactiotfDetween

epoxy resins and all primary and secondary amines; their speed of reaction is
faster, however, than cycloaliphatic and aromatic polyamines particularly at
ambient temperatures. Scheme 2.1(a) illustrates the initial step which
involves the primary amine active hydrogen adding to the epoxy group. This
is followed by the resulting secondary amine adding to another epoxy group.
Generally, primary and secondary amines are used at mix ratios that provide
one amine active hydrogen for each epoxy group, i.e. the stoichiometric
amount. When the epoxy is a diepoxide and when the amine is a diamine the
resulting product is an infusible polymer (Scheme 2.1 b). The reaction is
accelerated by materials that stabilise the alkoxide ion intermediates
(Scheme 2.1c). Therefore, hydroxyl groups are accelerators including those
generated during cure. They can also be added as accelerators and the more
acidic they are, the more effective. Hydroxylic accelerators or hydrogen
CURING AGENTS 39
(a)
CHrCH(OH)-CHzO*
/
-> R1N
\
+ + ->
CHz-CH(OH)-CHzO*
(b)
CHrCH(OH)-CHzO* *OCHrCH(OH)-CH z
/ \
*OCHz-CH( OH)-CHz-N N-CHrCH( OH)-CHzO*
\ 1 /1
*OCHz-CH(OH)-CHrN
f ~
N-CH r CH(OH)-CH 20*
\ /
CHrCH(OH)-CHzO* *OCHrCH(OH)-CHz
(c)
*OCHrHC\-~Hz + HOR -> *OCHz-HC-CHz + R1NHz ->
o
\ I
ROH-O
*OCHrHC-CHrNHR 1 + HOR
I
OH
Scheme 2.1 Mechanism of cure of primary amines. -CH20* indicates that only the reactive
epoxy group is shown.
donors capable of catalysing the amine epoxy reaction in decreasing order of

efficiency are: -OH (Ar ~ PhCH2 > RCH2 > H); -C0 2H; -S03H;
-CONH2; -S02NH2'
Ethyleneamines are highly reactive due to their unhindered
polyfunctional nature and give tightly cross-linked networks due to the short
chain distances between the active sites. This results in cured resins with
excellent solvent resistance and mechanical strength, yet limits flexibility.
Their low molecular weights mean that they have low viscosities, however it
also means high vapour pressures with the outcome that they are corrosive
and irritant to the skin. They are also high in odour, are hygroscopic and
have poor compatibility with epoxy resins. This last gives them a tendency to
exude during cure giving rise to bloom and carbonation. This can result
when amines of poor compatibility are used with epoxy resins - they can
exude to the surface after mixing and before cure has trapped them into the
matrix. Blooming or blushing as it is also described, refers to the resulting
tacky surfaces. Carbonation or water spotting refers to the white patches of
amine bicarbonate salts which can appear and mar the surface appearance of
cured films. The more basic polyamines react with atmospheric CO 2 in
preference/competition with epoxy resin to form the unstable carbamic
acids and amine carbamates. These can revert to amine and CO2 which in the
presence of water (humidity or excess) generates carbonic acid at the curing
surface which can then react irreversibly with amines to form the
undesirable amine bicarbonates.
The low molecular weights of the ethyleneamines mean the usage rates
are low and therefore they have to be mixed in to the epoxy resin with a high
degree of accuracy - apart from the need to avoid excess additions which
would exudelbloom, additions under stoichiometric amounts lead to
embrittlement. Various modification techniques have been developed,
however, to overcome most of the disadvantages and it is the derived
products, rather than the unmodified ethyleneamines, which are now used
extensively in epoxy applications. Most involve increasing the molecular
weight of the amine to reduce the volatility and most lead to significant
changes in reactivity and cured mechanical and chemical resistance
properties.
2.2.1.2 Polyamides, polyamidoamines and imidazolines. Condensation

with carboxylic acids to form ami des is a simple technique for overcoming
the volatility, irritation potential and critical loading issues of
ethyleneamines. Reaction of the higher homologues of the ethyleneamines
series with dimerised or polymerised unsaturated fatty acids as recorded by
Cowan et al. (1948) leads to the popular general curing agent type the
polyamide.
¢C
I
R R'
+ (2) H2N (CH, CH 2NH),H -(2) H20.
C0 2 H
¢C
I
R R'
CONH(CH 2 CH 2NH),H
C0 2 H CONH(CH 2CH 2NH)xH
A range of polyamides are available commercially, based upon diverse

fatty acid/dimer acid/amine ratios which have been processed under various
conditions. They vary in molecular weight, physical form, amine content
and reactivity but all are low in volatility. Their mixing ratios with epoxy
resins are generally non-critical and cured properties can be deliberately
varied from tough and hard at below stoichiometry, to flexible and soft
around or above stoichiometry. They give good adhesion to a wide variety of
substrates, particularly epoxy-polyamide coatings. This is due to residual
secondary amine functionality on the polyamide backbone which is often
deliberately not fully converted. This ensures keying-in sites and promotes
intercoat adhesion. The fatty acid/dimer acid backbone provides excellent
corrosion resistance by its water repelling nature but also leads to a colour
stability problem due to its unsaturated nature.
Polyamides are, however, rather slow in curing particularly at lower
temperatures and tend to give greasy cures due to exudation, but addition
CURING AGENTS 41
of tertiary amines, phenolic amines or co-curing agents can help to reduce
both of the defects. Alternatively, condensation of ethyleneamines with
phenolic-containing carboxylic acids to give the so called phenalkamine type
curing agent, according to Kittridge and Michelli (1968), gives enhanced low
temperature reactivity curing agents, though colour stability is sacrificed.
An alternative approach to overcoming the exudation problem, which is
caused by poor compatibility with epoxy resin of both polyamide and
unreacted ethyleneamine which is frequently present from the polyamide
manufacturing process, is to pre-react it with an epoxy resin to improve
compatibility. In contrast to the polyamides, polyamide adducts do not need
to be given an induction period prior to use after mixing to ensure that the
system is homogeneous. Adducts also cure better at low temperatures under
conditions of high humidity and give faster tack-free cure times by virtue of
a physical drying phenomenon resulting from their increased molecular
weights. Physical drying does not involve a chemical reaction; it is a purely
physical effect whereby (pre)polymers, when deposited as a film, upon
solvent evaporation appear tack-free or hard. Further improvements in
drying rates have also been obtained, according to Wilson (1986), by
replacing part of the dimer acid by aromatic polybasic acids.
The lower viscosity, more reactive polyamides are used with liquid epoxy
resin in high solids coatings, adhesives, sealants, and cable jointing
applications. In contrast the higher molecular weight high viscosity
polyamides, adducts and aromatic modified poly ami des are used mainly
with solid epoxy resin in solvent-based corrosion-resistant coatings.
However, the high viscosity of polyamides and ad ducts means that they
cannot easily be used in reduced volatile organic content or high solids
coatings. This led Brytus (1986) to the development of new types of low
viscosity solvent-free polyamide ad ducts which give properties in cured
coatings comparable to those obtained with conventional polyamides and
adducts. They have the added advantage that when used with liquid epoxy
resins they give higher crosslink densities and hence improved mechanical
properties, chemical resistance and corrosion resistance. Richardson (1973)
prepared water dispersible polyamides by modification of standard
polyamides and found them to be useful low odour, non-flammable
alternatives to solvent-based polyamide solutions with all of their inherent
advantages, in concrete primers, new to old concrete adhesives and surface
coatings applications.
Reaction of ethyleneamines with aliphatic monocarboxylic acids leads to
polyaminoamides or polyamidoamines or amidoamines as they are also
frequently called, which can, on further heating, undergo ring closure,
forming the cyclic ami dines or polyaminoimidazolines.
R-CH=CH-CH=CH-C0 2H + H 2N(CH 2CH2 NH)xH ~
R-CH=CH-CH=CH-CONH(CH2CH2NH)xH
polyaminoamide
R-CH=CH-CH=CH-C~
N
J
/N(CH 2CH 2NH)x_IH
polyaminoimidazoline
Generally tall oil fatty acids are used as the mono-basic acid component
and the resulting mixtures of polyaminoamide and polyaminoimidazoline
vary in pot-life/reactivity depending upon the imidazoline content. High
imidazoline content mixtures are long working life, low reactivity curing
agents which in the presence of water, particularly substrate surface
humidity, ring open back to the more reactive polyaminoamides. In effect,
they function as semi-latent curing agents and are widely used on damp
concrete for their excellent wetting characteristics and adhesion in concrete
coating, bonding new to old concrete and other adhesives applications.
Their low viscosity and excellent corrosion resistance also makes them
suitable for electrical potting and filament winding applications.
Mono-basic acids other than fatty acids have also been used by Wilson
(1970) in the preparation of pure imidazolines for use in fibre-reinforced
epoxy lamination applications where low reactivity, high strength and good
chemical resistance are important. Additionally, imidazolines, prepared
from mono-basic acids, have been further reacted with polycarboxylic acids
to form solid salts, and used by Veba Chemie (1977) for their unique ability
to provide matt finishes in powder coatings applications.
2.2.1.3 Adducts with epoxy functional materials. One of the earliest

methods investigated for the reduction of the irritation potential of
ethyleneamines was partial reaction with epoxy functional materials to
produce low volatility adducts. Newey (1958) reported that mono-epoxides
such as ethylene oxide, propylene oxide and mono-glycidyl ethers react to
produce low viscosity mixtures of mono- and di-adducted amines with the
expected reduced volatility.
H2N(CH2CH2NH)xH + H 2C-CH-R -> H2N(CH2CH2NH)xCH2CH(OH)R
'd +
In addition to the reduced irritation potential and increased loading, this

type of curing agent also has a faster cure rate due to in-built hydroxy
substitution. Although this class of curing agent gives reduced cross-link
densities compared to the starting ethyleneamines they are used widely in
wet lay-up laminating, patch repair of boats, adhesives and tooling
applications. Unfortunately free, or unadducted ethylene amine is also
present as part of the statistical distribution of products and even though it
can be removed by distillation the resulting toxicologically safer products in
practice fail to give even the same levels of physical properties and chemical
CURING AGENTS 43
resistance as the 'crude' reaction products, let alone the starting

ethyleneamines. Adducts with di-glycidyl ethers on the other hand when
cured give the same level of physical and chemical properties as the amine
adducted as there is no overall loss of active hydrogens/reduction in
crosslink density.
H2N(CH2CH2NH)xH + H 2C-CH-CH20' 'OCH2-HC-CH 2 + H 2N(CH2CH 2NH)xH
'd 'd
-> H2N(CH2CH2NH)xCH2CH(OH)CH20' 'OCH2CH(OH)CH2HN(CH2CH2NH)xH
With liquid epoxy resin ad ducts prepared using an excess of amine, fairly
viscous materials of low volatility and higher mixing ratios result. Highet
mix ratios allow small errors to be made in weighing that would be critical in
the use of the parent amines. Once again due to the presence of in-built
hydroxyl groups, they are faster curing and thus ideal for laminating and
adhesives applications. Their ability to give excellent solvent resistance also
leads to usage with higher functionality resins for optimum solvent-resistant
coatings.
With solid epoxy resin, adducts can be prepared in two ways. In-situ
adducts are produced simply by addition of amine to a solution of the resin to
yield an adduct solution directly. This type is used widely for its ease of
manufacture, low cost, good film flexibility and excellent chemical
resistance, in solvent based coating applications. In-situ adducts do,
however, contain low levels of free amine that have a tendency to exude and
therefore mixtures with solid epoxy resin solutions often need to be aged
prior to application. Isolated adducts are prepared by addition of excess
amine to the solid epoxy resin solution followed by removal of unreacted
amine and reaction solvent by distillation. This yields solid isolated ad ducts
which need to be redissolved and used in the same way as the in-situ adduct
solutions. They do nevertheless exhibit a number of advantages associated
with the lack of any free amine induding low odour, and are better suited for
application at lower temperatures and higher humidity where bloom,
carbonation, loss of gloss and intercoat adhesion criteria are critical.
2.2.1.4 Mannich bases and association salts. An alternative approach

developed by Halewood (1962) for overcoming the disadvantages of
volatility, irritation and hygroscopicity of the ethyleneamines involves
reacting together the amine with phenol and formaldehyde to form low
molecular weight condensates or Mannich bases as they are commonly also
known.
HO-Ar + CH20 + H2N(CH2CH2NH)xH -> HO-Ar--CH2HN(CH2CH2NH)xH + H 20
These products cure rapidly at low temperatures and in the presence of
moisture, do not blush and are less susceptible to carbonation than other
aliphatic amine modifications. They also give excellent solvent and overall
chemical resistance and are used for a wide variety of civil engineering,
coating, adhesive and laminating applications. Products of this type
generally contain 10-20% of unreacted phenols which aid compatibility for
coal tar extended epoxy coatings applications. They are causing legislative
concern and when phenol rather than an alkyl phenol is used in the
manufacture, 'skull and cross-bones' labelling is required for the resulting
products, even though the free phenol is associated with the amine in salt
form.
HO-Ar + H2N(CH2CH2NH)xH --+ Ar-O- + H3N+(CH2CH2NH)xH
These simple salts or association products provide reduced dermatitic

hazard and are also used extensively for optimum reactivity under adverse
conditions. Their low viscosities make them ideal for use in the formulation
of concrete patch repair mortars and adhesives. When mixed with epoxy
resin, however, the salts break down liberating volatile ethyleneamines and
their related problems.
2.2.1.5 Ketimines and acrylonitrile adducts. Alternative approaches to

increasing the molecular weights other than condensation with carboxylic
acids or partial pre-reaction with epoxy functional materials involve
condensation reactions with ketones and adduction with acrylonitrile.
Ethyleneamines react reversibly with ketones to form ketimines as reported
by Shell (1964).
H2NCH2CH2NHCH2CH2NH2 + 2 RR'C=O --+ RR'C=NCH2CH2NHCH2CH2N=CRR'
+ PhOCH2-HC-CH2
(j
These polyamines are readily hydrolysed by moisture to regenerate the

amine and ketone, the latter of which volatilises out during the normal film
forming/cure process. When secondary amine-containing ethyleneamines
are used, so called reactive ketimines are formed which can be blocked by
adduction with a mono glycidyl ether. Both reactive and blocked ketimines
are characterised by their low viscosities, semi-latent nature, and rapid cure
when exposed in coatings to atmospheric moisture. Their main application is
in high solids coatings. There are, however, a number of limitations. To
ensure full through cure, sufficient atmospheric humidity is required and
applied film thicknesses need to be restricted to avoid solvent (ketone)
entrapment.
Farnham and Caldwell (1956) recorded ethyleneamines react readily to
form acrylonitrile adducts with variable degrees of substitution depending
upon the ratio of reactants and manufacture conditions.
H2NCH2CH2NHCH2CH2NH2 + CH2=CH-CN --+
H2NCH2CH2NHCH2CH2NHCH2CH2CN
CURING AGENTS 45
The cyanoethylated amines are less reactive than the parent amines, low
in viscosity and give physical properties generally inferior to those of the
parent amines. They are not lower in irritation potential though and care
must be taken in operation to avoid the reformation of acrylonitrile from the
retro-Michael reaction which takes place at around 120°C. Their ability to
wet minerals and glass fibres well and low reactivity makes them particularly
suitable for laminating and filament winding applications.
2.2.1.6 Propylenamines. Hydrogenation of adducts of acrylonitrile with

amines other than ethyleneamines, often secondary amines, leads to
another class of hardener - the propyleneamines.
RIRzNCHzCHzCHzNHz Rl,Rz = Me,Me: DiMethylAminoPropylAmine (DMAPA)
= Et,Et: DiEthylAminoPropylAmine (DEAPA)
Rl = H,Rz = C6H ll : CycIohexylAminoPropylamine
With the exception of the cyclohexylamino derivative [3312-60-5] all are

unsuitable as crosslinking agents in their own right due to insufficient
functionality: they give lower cured mechanical and chemical resistance
properties than the ethyleneamines. They are low in molecular weights and
thus are corrosive and irritant to the skin due to high vapour pressures. In
addition their strong basicity leads to a strong tendency to carbonate.
However, they do have a number of advantages: low viscosity; high
reactivity due to unhindered amine functionality, particularly D MAPA
[109-55-7] and DEAPA [104-78-9] which have inherent tertiary amine
functionality (see section 2.2.4). In addition they have the ability to act as
reactive flexibilisers/toughening agents by virtue of their low functionality
which reduces cross-link density and promotes chain extension. These
features render them particularly useful as co-curing agents with high
functionality (cycloaliphatics) amines in a variety of coatings and flooring
applications. They are also used alone for adhesive applications.
2.2.1.7 Higher alkylenediamines. Alternate aliphatic diamines with

longer chain lengths between the reactive sites than the ethyleneamines and
propyleneamines are:
H zN(CHz)6NHz HexaMethyleneDiamine (HMD)
HzN(CHzhCH(Me)CHzNHz MethylPentaMethyleneDiamine (MPMD)
HzNCHzCHzCH(Me )zCHzCH(Me )CHzNH z TriMethylhexaneDiamine (TMD)
HMD [124-09-4] which is more widely known for its use as an intermediate
for nylon-6,6 manufacture, is derived by tail to tail dimerisation of
acrylonitrile and subsequent catalytic reduction of the intermediate
adiponitrile. This synthetic route leads to a number of coproducts including
MPMD [15520-10-2] which is the hydrogenated head to tail dimerisation
product. TMD which was developed for use as a urethane intermediate is
actually an equal part mixture [25513-64-8] of the 2,2,4- and 2,4,4-trimethyl

hexamethylenediamine isomers.
All three materials have physical characteristics derived from their low
molecular weights: low viscosity, corrosivity, irritancy to the eyes and skin.
They are high in reactivity and have a strong tendency to carbonate due to
unhindered primary aliphatic amine functionality. They also have an ability
to give moderate film flexibilities due to the moderate chain distances
between the cross-link sites. Additionally, their high hydrocarbon contents
make them hydrophobic and, problematically, only moderately compatible
with epoxy resin. They thus tend to exude and lead to surface tack in
unmodified form. Adduction with epoxy resin, Mannich base formation or
salting with acidic accelerators all help to minimise the corrosivity/irritancy
hazard, reduce the carbonation tendency and improve their compatibilities
with epoxy resins. The principle commercial uses for the various
modifications are coatings and concrete repair applications.
2.2.1.8 Fatty amines. Amines derived from C16-C 18 saturated and

unsaturated fatty acids are also used as co-curing agents in epoxy
applications. They are used primarily to provide flexibility and enhance
water and corrosion resistance but are limited by their low reactivity.
2.2.1.9 Polyetheramines. Of the various polyglycol polyamines types

evaluated in epoxy applications only the polyoxypropylene derivatives have
achieved prominence. Most other polyglycol polyamines, notably ethylene
oxide derivatives, are too hygroscopic to be used extensively. Lee and
Winfrey (1966) found that by varying the functionality of the initiating
polyol and the molecular weight of the polyoxypropylene derivatives,
several different molecular weight amine functional materials could be
made for use with epoxy resin.
The 230 and 400 molecular weight diamines and the 403 molecular weight
triamine are all low in viscosity and exhibit low reactivity and reduced
Table 2.2 Polyoxypropylene derivatives of varying molecular weight
x Molecular
weight
2.6 230
5.6 400
33.1 2000
x+y+z
5.3 403
CURING AGENTS 47
carbonation tendency compared to other types of aliphatic amines. This is
due to the slightly hindered secondary aliphatic amine functionality. When
cured they do show good film flexibilities and thermal cycling properties due
to the long chain distances between cross-link sites. The flexible aliphatic
nature does, however, limit their solvent resistance and heat resistance. This
combined with an inconveniently slow cure rate at ambient temperature
means that they are commonly used in combination with other amines and/
or accelerators, principally for flooring, adhesive, electrical potting and
encapsulation applications.
2.2.1.10 Polymeric amines. Many reports have appeared in the patent

and academic literature describing amine functionalised polymers but few
have achieved commercial prominence. However, McFadden (1972)
developed reaction products of ethyleneimine with carboxylic acid terminal
prepolymers for water-based coatings applications where they were found to
combine the weathering and colour stability properties of conventional
acrylic systems with the corrosion resistance of epoxy coatings.
2.2.1.11 Araliphatic amines. The araliphatic amine MXDA [1477-55-0]

prepared by hydrogenation of isophthalonitrile is a di-primary aliphatic
"'NCH'l§jCH,NH,
MetaXyleneDiAmine (MXDA)
amine which behaves more like a di-primary cycloaliphatic amine than an

aliphatic diamine. This is most likely due to the rigid nature of the non-
conjugated aromatic ring backbone.
MXDA has a low viscosity, a light colour, and is insensitive to moisture.
The high basicity of the unhindered primary aliphatic amine functionality
imparts high reactivity and unfortunately a strong tendency to carbonate.
This combined with the corrosive and irritant nature means that it is
generally used in preadducted form or as a Mannich base. Resulting
products used for high solids and solvent-free coatings and flooring
applications, are characterised by their excellent sub ambient temperature
cure rates and high cured chemical resistance.
2.2.1.12 Alicyclic aliphatic amines. One of the very first aliphatic
N-AminoEthylPiperazine (AEP)
polyamines with a (hetero-)cyclic backbone bridging the active amino-

hydrogen sites to gain commercial prominence in epoxy applications was N-
AEP [140-31-8]. This higher homologue from the ethyleneamine stream of
products contains primary, secondary and tertiary amine groups, although
only the three primary and secondary active hydrogens participate in
reaction with epoxy resin. At room temperature N-AEP behaves like its
higher functionality acyclic homologue TETA (see section 2.2.1.1) in terms
of corrosivity, handling and in giving rapid gelation and high exotherm. Full
cure is not however possible with N-AEP due to the formation of the
so-called B-stage. This term B-staging, adopted from phenolic resin
technology, refers to an intermediate stage in the reaction of thermosetting
resins in which the partly cured material softens or reflows when heated,
allowing it to be moulded into another shape or form. This is desirable in
epoxy laminating applications yet undesirable in ambient cure structural
coatings where it would be ruinous. Blending with other amines and/or
acidic accelerators is required to achieve full cure at ambient temperatures
or alternately an elevated temperature post-cure is necessary. Fully cured
materials exhibit high rigidity and high impact resistance. Typical
applications include: ambient cured patch repair adhesives and flooring
applications; heat-cured adhesives and electrical casting applications.
Spiro-acetal diamines
n = 1--4
Spiro-acetal diamines developed by Ajinomoto Co. Inc. (1969) from an

intermediate in the manufacture of mono-sodium glutamate have gained a
market sector presence as curing agents with reduced skin irritancy, reduced
amine blush, improved toughness and impact strength and low colour for
epoxy applications. The basic spiro-acetal diamines are usually modified by
adduction and acceleration and applications include flexible coatings,
adhesives and castings.
1,3- [BisAminomethyl} TriCycioDecane diamine NorBornaneDiAmine

-Cyclohexane (l,3-BAC) (TCD diamine) (NBDA)
1,3-BAC [2579-20-6] prepared by reduction of MXDA (see section

2.2.1.11) or by exhaustive hydrogenation of isophthalonitrile is one of the
few commercially available products clearly recognisable as an alicyclic
CURING AGENTS 49
aliphatic polyamine. This low viscosity material provides superior colour
stability compared to its araliphatic precursor but unfortunately inferior low
temperature reactivity and sensitivity to atmosphere humidity. The elevated
relative cost also limits its utility to high performance coatings applications.
TCD diamine [26655-37-8] which is the hydrogenated hydrocyanation
product of dicyclopentadiene is a mixture of endo and exo, cis and trans
isomeric tricyclic primary diamines. The high viscosity and cost, however,
limits its usage to specialist high performance applications where high heat
resistance can be attained due to the rigidity of the fused ring backbone.
NBDA [56602-77-8] which is the related hydrogenated hydrocyanation
product of norbornadiene is a lower viscosity alternative with a potential
which has yet to be fully explored.
2.2.1.13 Dicyandiamide and biguanide derivatives. DICY or cyano-

guanidine [461-58-5] as it is less commonly known, the agrochemical
industry basic intermediate made from cyanamide (H2NCN), was one ofthe
first curing agents to be used with epoxy resins.
H 2 N-C-NH-CN
II
NH
Dicyandiamide (DICY)
It gives systems with pot lives/shelf-lives of six months with liquid/solid

resins by virtue of its insolubility in the resins at ambient temperatures and is
considered to be a highly latent epoxy resin curing agent. When heated in
the presence or absence of catalysts, rapid polymerisation commences to
yield highly cross-linked molecules exhibiting high strength, moderate heat
resistance, high solvent resistance and excellent electrical insulating
properties. High temperatures, however, are required to initiate reaction,
therefore DICY is normally used in conjunction with co-curing accelerators/
catalysts as the uncatalysed reaction with epoxy resins begins well above
1S0a C.
The mechanism of curing epoxy resins with DICY is complex and involves
initial reaction between all four active hydrogens with epoxy resin in the
presence of tertiary amine catalysts (Scheme 2.2a) and simultaneous epoxy
homopolymerisation catalysed by the tertiary amine (Scheme 2.2b). This is
followed by a final cure reaction between hydroxyl groups produced in the
partially cured epoxy resins and the cyano group in dicyandiamide (Scheme
2.2c). This latter step is also catalysed by tertiary amines.
DICY when mixed with liquid epoxy resin for one component adhesive
applications and solventless fibre reinforced laminate applications is used in
micronised form. This ensures rapid dissolution at reaction temperature and
optimum homogeneity in the cured product. Cure is generally initiated with
urea catalysts (see section 2.2.4.3). DICY is also extensively used in both
unground and micronised form with solid bis-A epoxy resins in hot melt
(a)
(*OCHz-CHOH-CH2)zN-C-N(CN)CHz-CHOH-CH2-0*
II
*OCHz-CH-CH2 + H2N-C-NH-CN ->
t;J
\ I II CH2-CH(0-)-CHz-0*
o NH
(b)
*OCHz-HC-CH2 + [R3N] -> *OCHz-HC-CH 2----
'cf I
(0-CHz-CH-CH20*)n
I
0----
(c)
R'z-N-C=N-R' R' z-N-C=N-R'
I I
R'-N-CN R'z-N-C=O
I
HO oI
I
*OCHz-CH-CH2- *OCHz-CH-CH 2-
R' = OCH2-CHOH-CHz-
Scheme 2.2 The mechanism of cure of dicyandiamide.
powder coatings applications (unground material being preferred in some

instances where the dicyandiamide can be ground down to application
particle size at the same time as pigments/resin). For powder coatings
applications, imidazoles (see section 2.2.4.2) are most frequently used as
catalysts. The largest usage of DICY however is in solution with high
molecular weight bis-A resins in electrical laminating printed circuit board
manufacture. Here the tertiary amines (see section 2.2.4.1) or imidazoles
are the preferred accelerators.
Dicyandiamide is also a useful intermediate for the formation of a number
of alternative types of curing agents which are likewise used in thermally-
cured epoxy resin applications.
Melamine (Fatty) guanamine Biguanide
The reactive nitrile group readily trimerises to form the cyclic guanamine
or melamine [108-78-1]. General Mills Inc. (1965) found that reaction also
occurs with other nitriles such as those derived from fatty acids as in the
production of fatty guanamines. Amines add to the nitrile group to form
biguanides and Gempeler and Zuppinger (1971) discovered that aromatic
CURING AGENTS 51
amine-derived biguanides provided a number of advantages over dicyan-

diamide for powder coatings, adhesives and moulding power applications.
These include improved reactivity due to improved solubility in molten
epoxy resin and improved flexibility due to reduced cross-link density.
Joyce (1971) also reported dicyandiamide as a co-curing amine/catalyst
for applications at temperatures down to O°e. Such systems have no latency,
and cure rapidly to give excellent adhesion and good chemical resistance;
end uses include speciality quick-setting adhesives and wet lay-up
laminating.
2.2.2 Cycloaliphatic polyamines and derivatives

This group of curing agents includes authentic cycloaliphatic di-primary
amines, as well as other di-primary amine types with cycloaliphatic
polyamine performance.
2.2.2.1 Authentic cycloaliphatic polyamines. Catalytic hydrogenation of

aromatic di-primary amines produces cycloaliphatic di-primary ammes.
3,3' -Dimethylmeth ylene-di( cyclohex y lamine)
Methylene-di(cyclohexylamine) or
H2 N - Q - C H 2D - N H 2 bi sParaAminoCyclohexylMethane (P ACM)
1,2-Cyc1ohexanediamine or
1,2-DiAminoCycloHexane (I ,2-DACH)
Of the many possibilities evaluated with epoxy resin only two have
achieved commercial success: 3,3'-dimethyJmethyJene-di(cyclohexyJamine)
[6864-37-5]; and the unhindered analogue PACM or methylene-di(cyclo-
hexylamine) [1761-71-3] as it is also known. Two grades of the latter are
available: PACM 48 which is a solid produced under conditions which
maximise the thermodynamically stable trans,trans [6693-29-3] isomer
content; PACM 20 which is a liquid produced with a low trans,trans and high
cis,trans [6693-30-7] cis,cis [6693-31-8] isomer content.
Both the 3,3'-dimethylmethylene-di(cyclohexylamine), which is also a
liquid, and PACM 20 are low in viscosity and light in colour. Like aliphatic
polyamines they are skin irritants but have lower vapour pressure which
reduces the hazard due to inhalation of vapour. When used to cure liquid
epoxy resins at temperatures in excess of 100°C they give excellent heat
resistance and high mechanical strengths and both are used in filament
winding, laminating, tooling and small-medium sized casting applications.
The 3,3' -dimethyl substituted material which is relatively hindered and thus
lower in reactivity being chosen for longer working life and the PACM 20
due to its higher conformational fluidity chosen for its superior fracture
toughness. PACM 20 has been found better suited to epoxy usage than
PACM 48 due to the ease of mixing with liquid resins and is also especially
suitable for modification as an adduct for ambient cure applications where
the flexibility and unhindered nature can be used to advantage.
1,2-DACH [694-83-7] is not produced by catalytic hydrogenation of an
aromatic precursor but is available commercially as the principal coproduct
from the catalytic reduction of adiponitrile (see section 2.2.1.7). This unique
1,2-diamine has a higher degree of reactivity even compared to the PACM
20 and this coupled with the high volatility and the limited flexibility due to
the short distance between cross-link sites makes 1,2-cyclohexanediamine
better suited for use in modified form in ambient cure epoxy applications.
2.2.2.2 Other cycloaliphatic poiyamines. 1,8-Menthane diamine [80-52-4]

was one of the first cycloaliphatic polyamines to be considered for heat cure
epoxy applications. This highly hindered di-primary amine prepared
by hydrocyanation of the terpenes pinene or limonene is low in viscosity,
provides long working-lives and gives high heat and chemical resistance.
3-Aminomethyl-3,5,5-trimethylcyclohexylamine commonly known as
IsoPhoroneDiAmine (IPDA) [2855-13-2] has however been most widely
adopted as the industry standard cycloaliphatic amine.
1,8-Menthane diamine
IsoPhoroneDiAmine
(IPDA)
IPDA which, like its stable-mate TMD (section 2.2.1.7), is also a product
of a multi-step synthesis starting from acetone, via isophorone. The low
viscosity, light colour, excellent heat resistance and cured mechanical
strength mean that in unmodified form IPDA is particularly useful for
small-medium sized casting, filament winding, electrical and general
laminating applications. The differential reactivity resulting from the
presence of primary aliphatic and cycloaliphatic amine functionality content
allows selective adduction with epoxy resin and selective formylation/
Mannich base formation. This results in low viscosity, activated derivatives
suitable for room temperature curing high performance coatings and civil
engineering applications.
CURING AGENTS 53
2.2.2.3 Accelerated cycloaliphatic polyamines, adducts and Mannich
bases. The lower reactivity of the cycloaliphatic primary amine function is
caused in part by steric hindrance from the ring( s) and though an undoubted
advantage for heat-cure applications it does mean that acceleration is
necessary to achieve reasonable room temperature cure rates. As cyclo-
aliphatic polyamines are corrosive, irritant to the skin and have a strong
tendency to carbonate/water spot due to their high basicities they are also
best modified for ease and safety of handling. Although only a limited
number of the modification techniques used with aliphatic polyamines are
applicable to cycloaliphatic polyamines a range of curing agents with widely
varying properties can none the less be produced. Usage rate, viscosity,
reactivity, cured mechanical properties and chemical resistance are all
dependent on the formulating technique used.
The simplest modification approach involves blending the cycloaliphatic
polyamine with an accelerator and a plasticising diluent. Accelerators which
can be used include aliphatic polyamines, simple tertiary amines and organic
acids. The most used plasticising diluent particularly in combination with the
most used cycloaliphatic polyamine IPDA is benzyl alcohol as this latter
reduces brittleness, provides a weak accelerating effect and helps the
hindered amine get through the B-stage. Applications for this low viscosity
curing agent type include highly filled screed flooring and coatings
applications.
Ciba Ltd (1970) found a further useful method of modifying cycloaliphatic
amines involving partial pre-reaction with epoxy resin coupled with addition
of plasticisers such as benzyl alcohol to reduce the brittleness and lower the
viscosity. Accelerators may be added but are not essential as the hydroxyl
group formed in the adduction process has an accelerating effect, as does
benzyl alcohol. IPDA, 1,2-DACH and PACM 20 are all used to produce this
type of curing agent; the IPDA-based products performing as industry
standard; the DACH-based materials are used for optimum chemical
resistance; the PACM 20 ad ducts providing the best overall carbonation and
water spotting resistance. Applications for these high gloss, blush-resistant,
colour-stable curing agents include decorative and chemical spill resistant
high solids coatings and flooring applications, as well as laminate/gel coat
applications.
Rutgerswerke (1969) reported cycloaliphatic polyamines can also be
reacted with formaldehyde and phenols to form Mannich bases and again
plasticising diluents such as benzyl alcohol are best added to lower viscosity
and reduce brittleness. The Mannich bases are darker in colour than the
adducts and blends and are not colour stable, but do however cure better in
the presence of humidity and at lower temperatures. IPDA is effectively the
only cycloaliphatic polyamine type used, as the other polyamine types give
products which are too viscous to be of any value. Applications for the
Mannich bases include flooring, old concrete priming, new to old concrete
bonding and coal tar modified epoxy coatings.
2.2.3 Aromatic polyamines and derivatives

Aromatic amines are less basic than aliphatic or cycloaliphatic polyamines
and react only slowly with epoxy resin. Typically they exhibit long pot lives
and require long periods at elevated temperature to attain optimum
properties.
2.2.3.1 Heat-cured aromatic amines. All heat cured aromatic polyamines

provide cured systems with excellent resistance to a variety of chemicals
including organic and inorganic acids. They are particularly notable for their
high retention of mechanical properties in operation at elevated
temperatures. Most of the aromatic amines are solids at room temperature
and are often liquified for ease of mixing with liquid epoxy resin by the
formation of eutectic mixtures or blending with cycloaliphatic polyamines
for heat cure applications.
Meta-PhenyleneDiamine (MPD)
H2N-©-CH2~NH2 4,4'-Diamino-DiphenyIMethane (DDM)

or MethyleneDiAniline (MDA)
RI.R4= NH,; R'.Rl= H 4,4'-Diamino-DiphenyISulphone

RI.R4= H; R'.RJ= NH, 3,3'-Diamino-DiphenyISulphone
(DDS)
Of all the aromatic amines, MPD [1477-55-0} provides the tightest cross-
link density and exhibits the best solvent resistance. Philipson (1959) found
MPD, when blended with other solid aromatic amines, formed low viscosity
eutectics useful in the manufacture of chemically-resistant filament wound
pipe. DDM or MDA as it is also known, prepared by condensation of aniline
with formaldehyde exists in a number of forms ranging from the pure 4,4'-
diamino-diphenylmethane solid [101-77-9} with a melting point of around
90°C to low softening-point mixtures of methylene bridged polyphenyl
polyamines which are easier to melt/mix with liquid epoxy resin, as reported
by Seeger and Fauser (1954). The low polarity of the pure DDM curing
agent which imparts excellent electrical insulation properties combined with
high mechanical property retention even under conditions of high humidity,
make this the most suitable candidate for electrical casting, laminating and
adhesives applications. Although the most widely used of the aromatic
amines, DDM is now classified as a potential human carcinogen and for
many applications is being replaced by other types of curing agents.
CURING AGENTS 55
4,4'.DDS [80-08-0] has the principle advantage over other aromatic
amines in providing the highest heat resistance. It has become the standard
curing agent, for use in combination with the BFrMEA catalyst (see section
2.5.1), with specialised epoxy resins such as the tetra-glycidyl ether of
DDM, in high temperature tooling and high performance military and
aerospace laminating applications. 3,3'·DDS [599-61-1]' although giving
reduced heat resistance compared to the 4,4'- analogue, has been adopted
for its enhanced honeycomb peel strength in aerospace laminating
applications.
~NH2
Et~Et
DiEthylTolueneDiAmines
(DETDA)
NH2
Teo"a·alkyl DDM
R R
H'N*c(CH'''@-C(C''')'~H' Tetra alkylated di(4·aminophenyl)

di·isopropyl benzene
~ ~
DETDA which is the 2,4-diamino-3,5-diethyltoluene/2,6-diamino-3,5-

diethyltoluene isomer blend [68479-98-1] is a low viscosity non-crystallising
liquid. Urech et al. (1982) reports that it is much less reactive than
(unhindered) DDM and needs to be used in conjunction with accelerators or
co-curing (cycloaliphatic) polyamines to provide acceptable cure rates even
at relatively high cure temperatures. Applications for the accelerated
variant include filament winding, casting and tooling, and for the cyclo-
aliphatic polyamine co-curing blends, syntactic foam and casting. Toxicity
however is now also a cause for concern with the DETDA. Alkylated
DDMs such as tetraethyl DDM [13680-35-8], and the tetra-alkylated
die 4-amino phenyl)-di-isopropyl benzenes such as the tetramethyl deriva-
tive [5432-99-5] developed by Bauer (1989) are 'safer' than DDM from a
toxicological point of view being Arne's test negative (a test providing an
idea of the potential for a mutagenic effect which is some indication of
carcinogenicity). They are used in a variety of laminating applications for
their long working life/ready compatibility with liquid epoxy resin and long
out-times/hot-wet resistance respectively.
2.2.3.2 Room temperature cured aromatic amines. Sinclair and Tilley

(1967) found aromatic polyamines, such as DDM, can be dissolved in
phenols to produce curing agents which will cure liquid epoxy resins at
temperatures below DoC. Dunn and Mebes (1963) revealed that addition
of organic acid accelerators to solutions of partially adducted DDM
in plasticisers such as di-butyl phthalate also produced effective room
temperature curing agents. By changing the plasticiser types/
concentrations, and the accelerator amounts, a range of curing agents with
pot-lives from a few minutes to several hours, and with differing mechanical
and chemical resistance properties are obtained. All exhibit lower heat
resistance compared to unmodified aromatic amines yet are still tolerant to
moisture, are resistant to water spotting and provide excellent resistance to
chemicals upon spillage and immersion. The high abrasion resistance,
organic acid resistance and non-tainting behaviour of certain plasticised
types has led to their use in the internal coating of beer and wine storage
tanks. These, and others, are also used for the internal coating of gas and oil
pipe lines and chemically-resistant flooring, marine and maintenance
coatings where dark colours are acceptable. Toxicity and colour stability are
current issues with room temperature curing aromatic amines however, and
for most applications they are being replaced by formulated cycloaliphatic
ammes.
2.2.4 Catalysts and co-curing agents

Tertiaryamines, imidazoles, ureas.
Catalytic epoxy resin curing agents can function in a variety of ways. They
can function as: catalytic curing agents which achieve cross-linking by
homopolymerisation; as accelerators and co-curing agents for polyamines,
poly ami des or anhydrides; or as activators.
2.2.4.1 Tertiaryamines. The reaction rates of most tertiary amines with

epoxy resins are low at room temperature. Tertiary amines are therefore
generally used catalytically as accelerators in elevated temperature cured
systems. They are rarely used as sole curing agents. The mechanism of cure
can follow two mechanisms. In the absence of hydrogen donors, epoxy resin
polymerisation proceeds directly via an anionic mechanism initiated by the
t-amine/epoxy zwitterionic adduction product (Scheme 2.3a). In the
presence of molar quantities of hydrogen donors, e.g. hydroxyl groups,
polymerisation also proceeds anionically, but through a two step initiation
process (Scheme 2.3b).
The most commonly used tertiary amines Dimethylaminomethyl phenol
[25338-55-0] and Tris(dimethylaminomethyl) phenol [90-72-2] contain
phenolic hydroxyl groups as an integral part of the structure of the amine. As
such they are particularly effective room temperature accelerators/co-curing
CURING AGENTS 57
(a)
R3N + H2C-CH-CHz-O* ---> R3N+-CHz-CH(O-)-CH 2-O*
'd
---> R3 +-CH2-CH-CH2-0*
I
oI
CH2-CH( 0 - )-CHz-O *
(b)
R3N+ -CHz-CH(O-)-CH 2-O* ---> R3N+ -CHz-CH-CHz-O*
I
+ HOR' OH + -OR'
R'O- + H2C-CH-CHz-O* ---> R'O-CHz-CH(O-)-CH2-0*
'd
Scheme 2.3 Mechanism of cure of tertiary amines.
agents for polyamine and polyamide-based coatings and adhesives

applications.
Along with BDMA [103-83-3] and DBU [6674-22-2], the phenolic amines
are also efficient activators for: polysulphides and polymercaptans (see
sections 2.4.1 and 2.4.2 respectively) in room temperature cure adhesives
and sealants applications; for dicyandiamide in elevated temperature cure
electrical, sports and industrial laminate manufacture; and for anhydrides in
filament winding and electrical casting applications. In all cases the amounts
of the different t -amines added to a system are determined empirically to
achieve an optimum balance of cure rate, working life and cured mechanical
properties. Too much catalyst in the formulations may help achieve faster
cure rates but usually at the expense of working-life, embrittlement and/or
shrinkage.
RI = OH; R2 = CH 2NMe 2; Rl,R4 = H Oimethylaminomethyl phenol
R I = OH; R2,Rl,R4 = CH2NMe2 Tris( dimethyl amino me thy I) phenol
R1= H; R2 = CH2NMe2; R-',R 4 = H BenzylDiMethylAmine CBOMA)
DiazaBicycloUndecene (OBU)
Salting of tertiary amines is a commonly used technique to prolong the

working lives particularly for electrical casting and lamination applications
where vacuum degassing needs to be carried out without any danger of
premature gelation. The tri-2-ethyl hexanoate salt of tris (dimethyl amino-
methyl) phenol [51365-70-9] which is a liquid and readily miscible with
liquid epoxy resin is actually used as a sole curing agent for electrical
casting applications. DBU, which is an extremely strong base (pKa
11.5), when reacted with organic and organophosphorus acids forms salts
with varying activation temperatures and pot-lives such as the long
latency 4-methylbenzenesulphonate salt [51376-18-2]. The mechanism of
cure of the tertiary ammonium salts involves initial esterification of the
blocking acid group with epoxy resin. The liberated amines are then able to
react with the epoxy group in the case of the tri-2-ethylhexanoate salt of
tris(dimethylaminomethyl) phenol, or activate anhydride co-curing agents
in the case of the DBU salts.
Quaternary ammonium salts such as Benzyl Trimethyl Ammonium
Chloride (BTAC) [56-93-9] also find use as latent accelerators for anhydride-
cured epoxy casting applications. They are unable to dissociate on heating to
the tertiary amine and little is known about their mechanism of action.
2.2.4.2 Imidazoles. Imidazoles are obtained by dehydrogenation of

imidazolines which are themselves generally prepared from reaction of
1,2-aliphatic diamines and nitriles. Many examples of 1-H-imidazoles or
1-(N)-unsubstituted imidazoles as they are also known, have been com-
mercialised and include: 2-Methyl-ImidaZole (2MII2MZ) [693-98-1];
2-Ethyl-4-Methyl-ImidaZole (2E4MI/2E4MZ) [931-36-2] and 2-Phenyl-
ImidaZole (2PII2PZ) [670-96-2]. They are efficient accelerators for
anhydrides and dicyandiamide and also act as catalytic epoxy curing agents
at moderate to high temperatures.
R"
5
c=c
~
c=c\
I
/
c=c\
I
I 3 \
HNI,2 ~ N HN "N R'N N
c~ 'c~ 'C~
I I
R' R"
I-H-Imidazole R'= Me; R"= H (2Ml/2MZ) R' = PhCH 2; R" = Me
R'= Et; R"= Me (2E4Ml/2E4MZ) 1-Benzy 1-2-methy limidazole
R'= Ph; R"= H (2PI/2PZ)
When used as a unique curing agent they react with epoxy resin at the 3-N
position to form a 1: 1 molar adduction product (Scheme 2.4a). They then
react at the 1-N position with a second molecule of epoxy resin (Scheme
2.4b) to form a 2 : 1 adduct which contains a highly reactive alkoxide ion
which initiates rapid anionic polymerisation of epoxy resin. For 1-(N)-
substituted imidazoles, 1 : 1 molar adducts with highly reactive alkoxide ions
CURING AGENTS 59
C=C
(a) C=C I \
I
\ + 'O-CH -HC-CH 'O-CH 2-HC-CH -N N
HN ,N 2 \ / 2 I 2, ~
'CR o HO CR
(b)
+ 'O-CH -HC-CH -
2 \ / 2
o
Scheme 2.4 Mechanism of cure of imidazoles_
are formed directly therefore these tend to be somewhat more reactive

catalysts and accelerators than the l-(N)-unsubstituted analogues.
1-Benzyl-2-methyl-imidazole [13750-62-4] which is one of the few examples
of a l-(N)-substituted imidazole to be commercially exploited is a highly
reactive catalyst for anhydrides and epoxy resins at elevated temperatures_
The 2-methyl-imidazole, 2-ethyl-4-methyl-imidazole and 2-phenyl-
imidazoles are used exclusively as accelerators for dicyandiamide-cured
high molecular weight epoxy resin in powder coating and electrical
laminating applications_ These 1-H-imidazoles, unlike the N-substituted
variants which are liquid, vary in physical form (low to high melting points)
depending upon the molecular weight. Although they are poorly soluble in
liquid epoxy resin at room temperature and have pot-lives of several hours,
for many applications including one-pack adhesives, solvent-free laminating
and filament winding, various analogues with considerably extended
latency/working lives have been developed. Essentially, improved latency
has been achieved by: Sawa and Gohda (1978) by general insolubilisation
using c-methylolation; Winslow and LaBelle (1979) by incorporation of
electron-withdrawing nitrile groups; and again Sawa et al. (1980) by salting
with sterically-demanding isocyanuric acid_
2.2.4.3 Ureas. Trisubstituted ureas, or urons as they are also known,

derived by blocking isocyanates with dimethyl amine have been used widely
as accelerators for the dicyandiamide cure of epoxy resin. The most common
types are:
R R = H: R' = Cl, R"= H 3-(4-chlorophenyl)-1,1-dimethy1urea [ 150-68-5]
M"NCOHN~R' R, R' = Cl, R"= H 3-(3,4-dichlorophenyl)-1,1-dimethylurea
R, R', R"= H 3-phenyl-1,1-dimethy1urea [l0l-42-8J

[330-54-1]
R"
R = Me, R'= H: R"= NHCONMe2: 2,4-to1uene di-isocyanate derivative
All exhibit outstanding latency at normal ambient temperatures and are

widely used in one-pack adhesive applications. Although the mechanism of
cure is not fully explained there is clearly a synergistic effect between
dicyandiamide and certain trisubstituted ureas. It is possible that

dicyandiamide activates deblocking in situ, causing release of dimethyl-
amine, which then functions as a conventional tertiary amine accelerator.
Toxicity issues with the traditional chlorinated aromatic substituent
components means that for most applications the 2,4-toluene di-isocyanate
derivative [17526-94-2], which is a more efficient accelerator, is presently
the preferred replacement.
2.2.5 Hydrazine and hydrazides

H2N-NHCO-R-CONH-NH2
R = -(CH2)C: AdipyJ
-(CH2h-: AzeIayJ
-(CH2)s-: SebaeyJ
Dihydrazides derived from flexible di-carboxylic acid esters and

hydrazine hydrate provide a class of curing agent which exhibits an unusual
combination of attributes according to 3M Co. (1960). These include long
shelf-lives, moderate pot-lives at advanced temperatures and the ability to
be heated and subsequently cooled without substantial curing. Furthermore
rapid cure at application temperatures, yielding heat-resistant, high
mechanical and adhesive strength products makes this class of curing agent
particularly suitable for coatings and laminating applications.
2.3 Oxygen-containing curing agents
2.3.1 Carboxylic acids and anhydrides

Carboxylic acid functional polyesters and anhydrides are the second most
important family of curing agents for epoxy resins. Although the acids are
only practical in heat-cured surface coatings, or stoving enamels, they are
nonetheless the most widely used epoxy curing agent type (Table 2.1).
Anhydrides are suited to most other heat cure applications.
2.3.1.1 Carboxylic acid functional polyester resins. The basic structure of

the carboxylic acid functional polyester resin is:
nHOOCR1COOH + (n-l)HOR 20H ~
HOOCRICO-(OR20-0CRICO)n_IOH + 2(n-l)H20
According to polyester resin technology, the epoxy resin component of a

hybrid epoxy/carboxyl functional polyester system is regarded as the cross-
linking or curing agent; di-glycidyl ethers of bisphenol A are furthermore
considered as modifiers. The reason for the remarkable success of carboxyl
functional polyesters according to Husbands (1987) was and continues to be
CURING AGENTS 61
(a)
*-OCR1COOH + H 2C-CH-CH20* ---> *-OCR1COO-CH 2-CH(OH)-CH20*
'd
(b)
*-OCR1COOH + HO-CH(CH 20*)2 ---> *-OCR1CO-OCH(CH 20*)2 + H 20
Scheme 2.5 Mechanism of cure of hybrid epoxy/polyesters.
the availability of a wide range of relatively cheap raw materials. These can
be combined together easily to give the necessary properties of correct
softening point and reactivity for powder coatings applications.
The mechanism of cure involves two stages: addition of the carboxyl
group to the epoxy functionality (Scheme 2.5a) and esterification with the
secondary hydroxyls on the epoxy backbone (Scheme 2.5b). The water
released in this latter condensation reaction then volatilises from the film
during the cure process.
The poor resistance to weathering of the bisphenol A, di-glycidyl ether
epoxies has precluded widespread use in external topcoat systems, but their
excellent corrosion resistance and adhesion make them ideal reaction
partners for carboxyl functional polyesters in a number of appliance and
metal-finishing powder coatings applications. Epoxy resins with high
aliphatic contents such as Tri-GlycidylIsoCyanurate (TGIC) on the other
hand provide improved weatherability and are used with carboxyl functional
polyesters for architectural powder coatings.
R = -CH 2-CH-CH 2
'd
TGIC
2.3.1.2 Carboxylic acid anhydrides. Dicarboxylic acid anhydrides were

amongst the first curing agents used with epoxy resins. They have achieved
widespread commercial importance due to a combination of their cure and
cured properties. Reaction with epoxy resin is characterised by a low
exotherm and although long periods at elevated temperatures are required
to achieve full cure the resulting low shrinkage, stress-free systems provide
excellent electrical insulation properties. These properties are retained even
under moderate to high continuous operating temperatures due to the good
thermal stability and have therefore helped in their adoption in the heavy
electrical industry.
The mechanism of cure of anhydrides is more complex than that of amines
due to several competing reactions which can occur, especially when

accelerators are added to enhance cure rates. In the absence of added
accelerators or catalysts (Scheme 2.6a) the anhydride ring is opened by a
hydroxyl group from the backbone of the epoxy resin forming a half-ester.
This is followed by the half-ester carboxylic acid group initiating reaction
with epoxy resin to form a di-ester-alcohol which can continue the
polymerisation process either by esterification with another anhydride
molecule, or etherification with an epoxy group. The latter etherification
reaction is the accepted route and in practice only 0.85 equivalents of
anhydride are required to provide optimum cross-link densities and cured
properties. Lewis bases such as tertiary amines and imidazoles are widely
used as anhydride accelerators. These are able to open anhydride rings to
form internal salts (betaines) which then act as initiators of cure (Scheme
2.6b). The resulting carboxylate ions react with an epoxide group to yield
alkoxide esters which undergo reaction with further anhydride molecules to
form carboxylate anion functional esters. These can then react with further
epoxide groups and continuation of this alternating sequence leads to the
formation of a polyester. In usage, optimum properties are produced when
stoichiometric equivalents of epoxy and anhydride are employed, in line
with this mechanism which does not involve etherification reactions.
(b)
(
CO
Co
P +NR3 -
(COWR3
coo- + H2C-CH-CH 20*

-
\/
o
CO COWR,
+ ( '0 (
cO COO-CH -CH-CH 20*
2 I
COO
(
COO-
Scheme 2.6 (a) Mechanism of cure of uncatalysed anhydrides. (b) Mechanism of cure of Lewis
base catalysed anhydrides.
CURING AGENTS 63
Lewis acids such as BFramine complexes and tetra-alkyl ammonium salts

are also catalysts for the epoxy-anhydride reaction although no fully satis-
factory mechanisms have been proposed. This form of catalysis does
however strongly favour the etherification reaction and as little as 0.55
equivalents of anhydride per epoxy equivalent are found to provide
optimum cured properties.
Apart from the effects of stoichiometry and accelerator or catalyst choice,
widely differing cured mechanical properties can obviously be obtained by
varying the anhydride curing agent type itself. The anhydride types vary
from dicarboxylic acid cyclic anhydrides to tetracarboxylic acid cyclic
di-anhydrides (and beyond) providing increasing cross-link density and thus
heat, chemical and solvent resistance respectively. Linear aliphatic
polyanhydrides are used as flexibilising modifiers.
2.3.1.3 Dicarboxylic acid cyclic anhydrides. With the exception of

Phthalic Anhydride (PA) [85-44-9] the most important dicarboxylic acid
anhydrides are cycloaliphatic. PA which is the least expensive and the most
difficult to handle due to a strong tendency to sublime during use, and the
hydrogenated derivative Hexa-HydroPhthalic Anhydride (HHPA) [85-42-7]
are used chiefly for casting and laminates for electrical applications. Tetra-
HydroPhthalic Anhydride (THPA) [26266-63-7] gives properties very
similar to HHP A though darker coloured products result. HHP A which is a
low-melting point solid does not sublime and is used with other anhydrides
to make low viscosity liquid eutectic mixtures.
o > (X
©reo I
eo> 0=>
co co co
PA THPA HHPA
Me
(XCD
.-
:
,
7
MTtIPA
\
/
co
0
(XCD'0
Me/.
MHHPA
/
CO ac
I
Me/.l'
MTHPA
P
CD
co
\
orNMA
MethylTetraHydroPhthalic Anhydride (MTHPA) [11070-44-3] prepared

by Diels-Alder reaction between isoprene and maleic anhydride is a liquid
eutectic mixture of regio- and stereo-isomers and is now one of the more
commonly used anhydride curing agents for filament wound pipe manu-
facture. Other applications include electrical casting, potting, encapsulation
and impregnation - here however the more exclusive hydrogenated deriva-
tive Methyl Hexa-HydroPhthalic Anhydride (MHHPA) [25550-51-0] is
preferred, and used with cycloaliphatic epoxy resins for optimum light
stability.
Methyl Endomethylene Tetra-HydroPhthalic Anhydride (METHPA or
Nadic Methyl Anhydride (NMA) [25134-21-8] as it is also known), prepared
by Diels-Alder reaction between methylcyclopentadiene and maleic
anhydride is used in components where electrical property retention at high
temperatures is required. NMA is the anhydride of choice here due to the
high rigidity of the fused cycloaliphatic ring backbone.
CI
C'W)
CI~CO
CI
HET DDSA TMA
Further examples of dicarboxylic acid anhydrides less frequently

encountered but none the less offering unique properties include; Hexa-
chloro Endomethylene Tetra-hydrophthalic anhydride (HET) [115-27-5], or
chlorendic anhydride as it is also known, like METHP AlNMA has a fused
ring structure and in addition substituent chlorine atoms which confer flame-
ret ardency for electrical laminate and casting applications; OoOecyl
Succinic Anhydride (OOSA) [51546-46-7] imparts flexibility to castings and
is generally used in blends with other anhydrides for improving thermal
shock resistance; Tri-Mellitic Anhydride (TMA) [552-30-7] is a very reactive
anhydride due to the free carboxyl group which tends to accelerate cures
with epoxy resins and is normally used with solid epoxy resins for high
performance laminate applications. Toxicity concerns have limited wider
usage of TMA.
2.3.1.4 Tetracarboxylic acid di-anhydrides. PyroMellitic Oi-Anhydride

(PMDA) [89-32-7] and Benzophenone Tetra-carboxylic acid Oi-Anhydride
(BTOA) [2421-28-5] are aromatic di-anhydrides which are used to achieve
high cross-link densities and high heat resistance.
/o ocIQr
0 co'0/
'oe CO
PMDA BTDA
PMDA which is the more compact cross-linking agent of the two gives the
better chemical resistance and the best resistance of all anhydride curing
agents. It is however a high melting solid and very reactive towards epoxy
CURING AGENTS 65
resin at high temperatures. It is used therefore in combination with other
anhydrides such as maleic anhydride to make it easier to dissolve and more
convenient to handle for a variety of electrical casting applications. Maleic
anhydride by itself is not a particularly useful curing agent as it produces
brittle cured epoxies. BTDA, also a high melting point solid, is somewhat
easier to handle due to its lower reactivity and is mostly used, alone or in
combination with other anhydrides, for high temperature-stability
electronic moulding powder and adhesive applications.
2.3.1.5 Linear poly-anhydrides. Black et al. (1964) have reported the use
of linear aliphatic poly-anhydrides derived by dehydration polymerisation
of adipic, azelaic and sebacic acids which contain terminal carboxylic acid
groups at either end of long chain aliphatic polymers. They are used
essentially as flexibilising modifiers for other anhydrides where
improvements in thermal shock resistance are required.
2.3.2 Phenol formaldehyde resins

Phenols can be reacted with formaldehyde to give resinous products,
Phenol-Formaldehyde (P-F) resins, with a wide variety of commercial
applications; Oldring (1987) records certain types being suitable as cross-
linking agents for epoxy resins.
2.3.2.1 Phenol novolac resins.
The novolac resins, which are the reaction products from formaldehyde and
excess phenol under acidic catalysis, when co-cured with high molecular
weight solid bis-A epoxy resins result in coatings with excellent adhesion,
film strength, flexibility and chemical resistance.
They are especially useful in powder coatings applications for corrosion
resistant pipeireinforcing bars (rebars) and with brominated epoxy resins for
FR3 electrical laminate production. The cure mechanism (Scheme 2.7)
involves poly-addition to epoxy resin and is activated by acids such as
p-toluenesulphonic acid.
Scheme 2.7 Mechanism of cure of novolac resins.

2.3.2.2 Resole resins.
The resole resins are the reaction products from excess formaldehyde and
phenol under basic catalysis. When co-cured with high molecular weight
solid bis-A epoxy resins in a poly-etherification reaction (Scheme 2.8)
because of the secondary hydroxyls on the epoxy backbone they yield even
higher cross-link densities and higher chemical resistance than novolac
resins. As such they are well-suited to drum and pail coating applications
where the high stoving temperatures easily drive off the water produced in
the condensation cure reaction.
Scheme 2.8 Mechanism of cure of resole resins.
2.3.3 Amino formaldehyde resins

Even more tightly cross-linked structures than attainable with phenol-
formaldehyde cured epoxy resins can be derived by heating Amino-
Formaldehyde (A/F) resins and epoxy resins at temperatures in excess of
150°C. Oldring (1987b) describes the most commonly used amino-resins
produced from: urea, the Urea-Formaldehyde (U-F) resins; and melamine,
the Melamine-Formaldehyde (M-F) resins. Amino-resins are invariably
also alkylated-etherified with alcohols under acidic conditions to restrict
possible methylol-methylol condensation/polymerisation reactions. The
degree of alkylation is an important factor in determining the performance
characteristics of the amino resins. As the degree of alkylation increases, the
viscosity and reactivity of the resin decreases, while compatibility with
epoxy resin increases.
CURING AGENTS 67
HOH 2C CH 20R 1
\ /
N N N
,NHCH 20R 1
Cp R10H 2C' Y Y 'cH 20R2
NHCH 20H NIN
NHCH 20H
U-F resin M-F resin
Rl = n-C4H9 , iso-C4H9 , C2Hs, CH3; R2 = another M-F resin
The mechanism of cure with epoxy resin involves: etherification with loss
of water or alcohol through the many secondary hydroxyls on the backbone
of the high molecular weight solid bisphenol A epoxy resins (Scheme 2.9a)
and, addition of N-methylol groups to the epoxy functionality (Scheme
2. 9b ). M-F amino-resins will also catalyse homopolymerisation of the epoxy
resin.
(a)
CH 20H + HO-CH(CH 20*h CH2-O-CH(CH 20'h + H20
***-N\ 1
---+ ***-N\ 1
CH20R + HO-CH(CH 20*)2 CH 2-O-CH(CH20*h + HOR
(b)
CH 20H + H2C--CH-CH 20* CHz-0-H 2C-CH(OH)CH20*
°
~
/ \ I /
'**-N ***-N
\ \
Scheme 2.9 Mechanism of cure of amino-resins.
The resulting cured films offer a combination of the best features of epoxy
and amino resins namely high adhesion, chemical resistance, gloss
retention, colour and colour retention and, with correct choice of amino-
resin, very good flexibility for metal decorating applications. M-F resins are
used for their superior film hardness and gloss properties where high
performance demands outweigh cost considerations, notably varnish and
moulding powder applications. U-F resins offer cost performance benefits
for fast bake enamel, stoving primer and can and drum top coating
applications.
2.4 Sulphur-containing curing agents
The thiol or mercaptan group (-SH) is able to react with epoxy resin in an
addition reaction. This requires catalysis at room temperature by amines
which promote production of reactive mercaptide ions (Scheme 2.10). With
*-SH + NR3 ~ *-S· H+NR3 + H2C-CH--CH20* --> *-S-H2C-CH(OH)CHzO* + NR3

\/
°
Scheme 2.10 Mechanism of cure of thiols/mercaptans.
mercaptan terminal polysulphides, whose functionality (typically 2) is too

low for efficient cross-linking of di-functional bisphenol A epoxy resin,
primary amines are used as initiators/co-curing agents. With polymer-
captans where the functionality (typically 3) is adequate for efficient cross-
linking, strongly basic tertiary-amines are used as initiators and rapid cure
times result.
2.4.1 Polysulphides
H(S-CH2CHz-OCH20-CH2CH2-S)X<SCH2CH(SH)CHzS)yH
Fettes and Gannon (1957) found low mercaptan content polysulphides

useful curing agents for epoxy resin. As the stoichiometric loading with
liquid epoxy resins is high, this leads to highly flexible products with very low
cross-link densities and poor mechanical properties. In practice however not
more than half the stoichiometric quantity is used with the proportion
varying according to the degree of flexibility required, and
co-curing polyamides, polyamidoamines, cycloaliphatic amines and
tertiary-amine catalysts are used to upgrade mechanical properties. Main
applications for epoxy-polysulphide systems are in building adhesives,
primarily because they give excellent adhesion to both concrete and glass
and also because they cure considerably faster than amine curing agents.
2.4.2 Polymercaptans
R-[O-{ CH 2CH2CH20-} nCH2-CH(OH)-CHz-SHh
n = 1 - 2; R = aliphatic hydrocarbon
The attractive reactivity characteristics of the mercaptan group with epoxy

resin have led to several attempts to commercialise other mercaptan-curing
agents. Diamond Chemicals (1968) designed certain tri-functional
polymercaptans to meet special application situations. Here the inherent
beta-hydroxy groups activate the mercaptan groups to give high reactivity at
low temperature and in thin films. When used at high loadings relative to
tertiary-amine co-catalyst, these polymercaptans also give the high lap-
shear adhesion required for Do It Yourself patch repair adhesive and
structural concrete bonding applications. The moisture resistances of the
cured systems are low however, due to the beta-hydroxy functionality, and
this with the disagreeable odour and skin sensitisation tendency has limited
their widespread adoption.
2.5 Miscellaneous curing agent types
2.5.1 Amine-boron trihalide complexes

R 1R 2R 3N+-B-X 3 Rl,Rz,R3 = alkyl or aryl; X = ForCl
CURING AGENTS 69
Greenlee (1955) reported that amines can react with boron trihalides to
form stable donor-acceptor complexes, which act as latent catalysts for
epoxy homopolymerisation. The reactivity of the complexes is controlled by
either the boron trihalide, the trichloride being a stronger acceptor than the
trifluoride, or the amine selection. The choice of amine also determines
physical properties such as solubility and miscibility with the epoxy resin, as
well as hygroscopicity: mono-ethyl amine-boron trifluoride (BF3-MEA) is
strongly sensitive to water; benzyl amine-boron trifluoride is not. Amine-
boron trihalide complexes cure epoxies to highly cross-linked, high heat and
chemical resistant materials which are somewhat brittle in nature;
plasticising carrier diluents are often used to improve compatibility with
liquid epoxy resin, and to reduce brittleness.
When activated by heating in the presence of epoxy resin, rapid reactions
occur accompanied by considerable evolution of heat therefore mass sizes
are limited. The complexes are used principally as catalysts with other curing
agents such as anhydrides for the curing of large masses for lamination,
casting, encapsulation and electrical insulating varnishes. Applications as
sole curing agents include adhesives, vacuum pressure impregnation,
encapsulation and electrical varnishes.
2.5.2 Quaternary phosphonium salts

Rl R2
,+/
,cP, -x Rl,R2,R3,R4 = alkyl or aryl; X = ClorI
R3 R4
Quaternary phosphonium salts such as the alkyl-triphenylphosphonium
halogenides are recorded by Smith (1978) as effective latent accelerators for
anhydride and phenolic resin cured epoxy resins. They have indefinite shelf-
lives, confer rapid cure rates at only moderately elevated temperatures, and
have no adverse effect on cured resin mechanical and electrical properties.
Although it is not clear exactly how they function their mechanism of cure
does not involve decomposition to phosphine. Typical applications with
anhydride/liquid epoxy resin include electrical casting, potting, encapsula-
tion and 'vacuum pressure' impregnation. Quaternary phosphonium salts
according to Martin (1976) are also effective catalysts for the advancement
of liquid epoxy resins with phenolic materials, in, for example, the prep-
aration of powder coating resin systems.
2.5.3 Cationic salts
Ar = Phenyl; -x = -BF4, -PF6 , - AsF6 , -SbF6

Cationic salts such as the tri-arylsulphonium salts have been found by
Crivello (1977) to be exceptionally efficient radiation cure photoinitiators
for cationic polymerisation of various resin types including epoxy. Only

cycloaliphatic epoxy resins however are highly sensitive to acid-catalysed
cationic polymerisation at ambient temperatures. Glycidyl ethers and
epoxidised oils require thermal treatment. The mechanism of cure with
these catalysts involves breakdown to a source of strong acid, HX where X is
BF 4, PF"6, AsF"6 SbF"6 and various radical fragments. Flexibility, hardness,
solvent and water-resistance properties depend upon the exact nature of
the resins and diluents selected. Cure rates depend upon the catalyst
concentration, coating thickness and irradiating light intensity.
Applications thus far developed for this type of curing system include can
coatings and a variety of specialist plastic coatings where adhesion is difficult
to obtain with other resin systems.
2.6 Summary
It can be readily appreciated that the precise choice of curing agent type can
markedly affect the properties of an epoxy resin when cured. As it is unusual
for a single catalyst or curing agent to provide optimum characteristics in any
given application, the identification and selection of a satisfactory curing
agent is often, therefore, in itself a complex procedure. Dunn (1984) has
however successfully rationalised the technique for surface coatings
applications and Ashcroft (1991) has attempted civil engineering/
construction applications.
References
Ajinomoto Co. Inc. (1969) Brit. Patent 1,155,973.

Ashcroft, W.R. (1991) Eur. Coatings J. 4, 229-241.
Bauer, R.S. (1989) US Patent 4,855,386.
Black, R.G. et al. (1964) SPE Antec, Jan.
Brytus, V. (1986) J. Coatings Technology 58(740),44-48.
CIBA Ltd (1970) British Patent 1,204,224.
Cowan, J.e. et al. (1948) US Patent 2,450,940.
Crivello, J.V. (1977) US Patent 4,058,401.
Diamond Chemicals (1968) Diamond Chemicals Technical Bulletin 3-800LC-61O.
Dunn, P.A. (1984) Polymers Paint Colour Journal Aug. 8/22.
Dunn, P.A. and Mebes, K.H. (1963) Farbe Und Lacke, 69, 346-353.
Farnham, A.G. and Caldwell, N.J. (1956) US Patent 2,753,323.
Fettes, E.M. and Gannon, J.A. (1957) US Patent 2,789,958.
Gempeler, H. and Zuppinger, P. (1971) US Patent 3,631,149.
General Mills Inc. (1965) British Patent 988,410.
Greenlee, S.O. (1952) US Patent 2,585,115.
Greenlee, S.O. (1955) US Patent 2,717,885.
Halewood, P. (1962) British Patent 886,767.
Husbands, M.J. (1987) In Resins for Surface Coatings, ed. Oldring, P. and Haywood, G. SrTA
Technology III, 63-167.
Joyce, J.D. (1971) British Patent 1,219,069.
CURING AGENTS 71
Kittridge, J.B. and Michelli, A.L. (1968) US Patent 3,390,124.
Lee, J.M. and Winfrey, J.e. (1966) US Patent 3,236,895.
McFadden, R.T. (1972) US Patent 3,634,372.
Martin, P.H. (1976) British Patent 1,429,076.
Mika, T.F. and Bauer, R.S. (1973) In Epoxy Resins: Chemistry & Technology, ed. May, C.A.
Marcel Dekker, New York, pp. 465-550 and reference cited therein.
Newey, H.A. (1958) US Patent 2,864,775.
Oldring, P. (1987) In Resins for Surface Coatings ed. Oldring, P. and Haywood, G. SITA
Technology II, 29-46. (1987b) idem, 46-71.
Philipson, J. (1959) US Patent 2,891,927.
Richardson, F. (1973) Pigment And Resin Technology May 41-43.
Rutgerswerke (1969) French Patent 1,568,011.
Sawa, N. and Gohda, K. (1978) US Patent 4,104,473.
Sawa, N. et al. (1980) US Patent 4,189,577.
Seeger, N.V. and Fauser, E.E. (1954) US Patent 2,683,730.
Shell International Research (1964) British Patent 960,236.
Sinclair, J.H. and Tilley, e.G. (1967) British Patent 1,083,657.
Smith, J.D.B. (1978) Org. Coatings & Plastics Chem. 39,42-46.
Urech, K. et al. (1982) US Patent 4,365,108.
Veba Chemie (1977) British Patent 1,460,505.
Wilson, A.F. (1970) British Patent 1,212,276.
Wilson, A.F. (1986) UK Patent Application GB 2,170,813A.
Winslow, J.M. and LaBelle, G.E. (1979) Deutsches Patentamt 2834371.
3M Co. (1960) British Patent 842,867.
3 The kinetics of cure and network formation
B. ELLIS
3.1 Cure of epoxy resins
Although the cure of an epoxy resin involves reaction between epoxy and
hardener reactive groups, as discussed in chapter 2, the full characterization
of the cure process involves many other factors. During cure, a liquid or fluid
resin-hardener mixture is converted to a solid (chapter 1, section 1.1 and
Figure 1.1). Critical features are gelation and the onset of vitrification. The
latter effect occurs as the glass transition temperature, Tg , approaches the
cure temperature, Te. The determination and alternative definitions of Tg
will be considered later (section 3.2). However, it should be noted that the
glass transition temperature is a function of the extent of reaction, X e ,
X = Eo - E(I,)
e E
o
with Eo and E(t,) being the concentrations of epoxy groups present initially
and at cure time te and has been used as a measure of the extent of reaction
(Wisanrakkit and Gillham, 1990). When the difference between the cure
temperature and the glass transition temperature becomes 'small', that is,
I::..T == Te - Tg(Xe) is small, curing reactions become diffusion controlled
because molecular mobility is rapidly reduced as I::..T -'> O. Cure may
continue very slowly even when I::.. T becomes negative. Then the glass
transition temperature exceeds the cure temperature.
Initially when Te > Tg (0) the rates of the curing reactions are chemically
kinetically controlled, and the rate expressions which have been proposed
for the early stages of cure will be discussed in section 3.4. With the
formation of larger more highly branched molecules, a critical point is
reached, the gel point, and then there is the formation of an infinite three-
dimensional network. If Te is too low, vitrification may occur before gelation
and then further reaction may be inhibited. An elevated temperature may
be used to reduce the viscosity so that the hardener can be mixed into the
resin and then the curing reactions can be quenched by rapidly cooling the
mixture to a storage temperature below Tg(O) , the glass transition
temperature of the 'compounded' resin. However, it is necessary to ensure
that mixing is complete before gelation, since the viscosity rises rapidly at
the gel point and the resin will not flow after formation of an incipient
network. Thus, information on gelation and vitrification is required to
KINETICS OF CURE 73
characterize an epoxy resin cure system and to specify an efficient cure

programme to ensure that the properties of the cured resin are optimized for
a specific application.
For the technological application of epoxy resins it is essential that mixing
of resin with the hardener and any other components should be complete
before the onset of gelation. The origin of some of the morphological
features that have been observed in cured resins may well be due to
incomplete mixing, so that there are regions rich in hardener and hence vice
versa. When the stoichiometry of the resin and hardener is considered, such
heterogeneity will limit the extent of chemical conversion and hence the
attainment of optimum physical and mechanical properties.
The 'pot-life' of the mixed resin and hardener is the time after mixing that
the resin can be processed for such applications as impregnation, casting or
other operations which involve 'flow'. Such operations have to be completed
before the viscosity of the resin starts to increase as the gel point is
approached. The viscosity could obviously be reduced by increasing the
temperature, but then curing reactions can occur and hence the pot life-time
will be reduced. For some applications diluents are added to reduce the
viscosity so that processing can be completed expeditiously. The pot-life
may be determined by use of so-called 'gel-timers'; commercial instruments
with a probe which 'prods' the resin periodically. When penetration is
prevented by gelation the time is recorded. Such measurements can be
sensitive to the sample size since the heat evolved in an exothermic reaction
causes the temperature to rise and hence the reaction proceeds more rapidly
and the curing resin may exhibit an 'exotherm'. For similar reasons the pot-
life will depend on the batch size, since the heat evolved will depend on the
volume which increases more rapidly than the surface area from which heat
can be lost.
In the cure schedule it is usual to specify a cure time te and a cure
temperature Te. However, the resin temperature may not ever be constant
at Te due to its low thermal conductivity and subsequently to an exothermic
heat of reaction (Amdouni et aI., 1990). With a resin sample of finite
thickness there will be a time-lag before the whole sample attains this
impressed temperature, Te. The time required for the centre of the resin to
attain Te will depend on the thermal diffusivity of the resin/hardener, the
product thickness and the heat transfer coefficient between for example a
metal mould and the initially liquid resin. This heat transfer coefficient may
change during cure because of shrinkage of the resin away from the metal
mould surfaces. For the cure of thicker sections the temperature of the
centre of the moulding may well exceed that of the mould surfaces, because
of the heat of reaction. To avoid excessive exotherms it is often necessary
to cure the resin at a relatively low cure temperature, e.g. 100a C for
BADGE + DDM, with a subsequent increase in the temperature for a
post-cure of up to maybe 180a C.
When the temperature of the curing system is near to Tg(tJ' i.e. ~ T =

(Tc - Tg(t,) becomes small, the rate of cure can become very slow and hence
post-cures at elevated temperatures are normally used to ensure that all of
the epoxy groups are consumed. Also, for many applications it is advan-
tageous to maximize the glass transition temperature; this is also achieved by
a suitable post-cure treatment.
Information regarding the chemical reactions occurring during cure are
difficult to monitor for the whole of the curing process. It is only possible to
carry out a 'wet' or similar chemical analysis on soluble resin hardener
mixtures. So alternative methods are required to monitor cure beyond the
gel point. The many methods that have been used will be discussed in section
3.3. Even with the extensive choice available it is not easy to monitor the
cure of epoxy resins in situ. Maybe measurement of the change in electrical
properties, either conductivity or dielectric properties, should be explored
more extensively.
3.2 Gelation, network structure and glass transition temperature
In this section the critical conditions necessary for gelation will be outlined.
After gelation the sol fraction, Ws decreases to essentially zero and the gel
fraction increases so that Wg ~ 1, where Ws and Wg are the weight fractions of
sol and gel respectively. With increasing extent of reaction the structure of
the network changes with more and more chains being 'tied' into the
network. Also the length of the network chains decreases as the degree of
cross-linking increases. Many of the physical and mechanical properties of
the cured resin depend on the details of the network structure. Because of its
crucial importance for the final stages of cure the glass transition
temperature will be considered, after discussion of gelation. There are still
many theoretical papers each year on the nature of the glass transition but
here only the essential features will be outlined which are relevant to the
cure of epoxy resins.
For an initial analysis of the change in macroscopic properties of epoxy
resins during cure it is usual to assume that the curing resin is homogeneous.
That is, the property changes are uniform throughout the sample. However,
there is some evidence for uneven curing (Ghaemy et al., 1982), and there
have been a number of studies of the morphology of cured resins. However,
with BADGE cured with DDM it was not possible to identify any structure
larger than a few bisphenol A repeat units but with other hardeners which
were less soluble in BADGE some features could be identified (Rashid,
1978). It would appear that at least some of the morphological features
observed in the cure of resins are due to poor mixing and dissolution of the
hardener in the resin. In other cases there may be either phase separation or
the formation of micro gel during the curing reaction. Ideally, before
interpreting macroscopic property changes the morphology of the cured
KINETICS OF CURE 75
resin should be examined. Some aspects of the inhomogeneity in cured

epoxy resins have been reviewed by Dusek (1986). However with these
cautionary remarks, further discussion in this chapter will be based on the
premise that the inhomogeneities that may be formed during cure are
unimportant.
3.2.1 Branching theory

From study of condensation reactions with monomers which have
functionalities greater than two it became clear that there was a critical
point, the gel point. With extents of reaction less than that required for
gelation the sample was soluble in suitable solvents and had measurable
viscosities. Beyond the gel point the sample was a visco-elastic rubbery solid
and would only swell with imbibed solvent. Carothers (1936) suggested that
the critical extent of reaction for gelation is Pg = 21fav where Pgis the extent
of reaction at the gel point and fay is the average functionality with
+ = 'f.NJ;
Jav 'f. N
I
where N; is the number of molecules with a functionality /; and the

summation is over all monomers in the system. For kinetic studies the extent
of reaction is defined as Xe = [Eo - E(dlEo but for branching theory it is
given defined by Pe = the probability that a reactive group has reacted.
Also, a is the branching coefficient. In this chapter usual terminology has
been used.
Unfortunately the Carother's equation is only approximate and predicts
higher extents of reaction at the gel point than are observed experimentally
(Pinner, 1956). An outline of gel point prediction is given in standard
textbooks on polymer chemistry (see for example Elias, 1976; Billmeyer,
1984; Labana, 1986).
The theory of gelation was developed systematically by Flory who gives an
account of these developments in his standard text (Flory, 1953). The
problem is illustrated in Figure 3.1 where the spatial configuration of highly
branched molecules is represented. It is necessary to predict the extent of
reaction required for the formation of an infinite molecule. A diepoxide
molecule f = 2 can be cured with anf-functional hardener and for simplicity
the reactivities of these hardener functional groups are regarded as equal.
Since the formation of an infinite molecule occurs at the gel point any repeat
unit on such a molecule must be connected to another unit via a branched
unit, in this case the hardener. The probability that this occurs can be
represented by a branching coefficient a which is defined by
a = probability that a functional group on a branch unit is
connected to another branch unit
\
\
\
\
\
\
\
\
\
\
\
\
\
(a)
(b)
Figure 3.1 The spatial configuration of highly-branched moledules. (a) Tetra functional
junction point. (b) Small portion of a three-dimensional network with tetra functional junction
points. --> indicates that the junction point is connected to other network chains - for a tetra
functional junction point there will be three such connections.
The formation of branched molecules can be represented as a branching tree

(Figure 3.2). A randomly chosen hardener branched unit with functionality
fis placed at the base. With a branching coefficient a there will be afreacted
hardener groups in Rank 1, that is the number of reacted branches in Rank
1 is B 1 = af. In Rank 2 one of the functional groups has been consumed
connecting it to Rank 1 so that the number of new reacted branches is
B z = a(f - 1). af = a Z f(f - 1). Generally:
Bn = an . f. (f - 1 r- 1
= af[a(f - )r -1 (3.1)
KINETICS OF CURE 77
rank
--------------~~~----------------o
Figure 3.2 Formation of branched molecules. (e), Reacted functional group; (0), reactive
functional group. For alternative representations and discussion of molecular trees and
networks see Gordon and Ross-Murphy (1975).
when [aU - 1)] < 1 the number of branches in the nth rank decreases as n
increases, Bn ~ 0 as n ~ 00. However, when [aU - 1)] > 1 then Bn > 0
for all n. Thus, there is finite probability that an infinitely large molecule will
be formed, i.e.
1
a =-- (3.2)
C f-l
where a c is the critical branching coefficient required for incipient network
formation.
It should be noted that this prediction of the critical condition for gelation
excludes the possibility of ring formation and hence the observed gel point
may occur at a higher extent of reaction.
The relationship between a and the extent of reaction for a simple case
may be obtained by the following argument. The calculation of a is
equivalent to the probability that a branched unit is part of the structure
~HE---EH~
where HE and EH are reacted epoxy and hardener groups. This structure
must be present when a chosen reacted hardener group is part of a link
joining two branched units. This chosen hardener group can only be part of
such a structure if it has reacted and is also attached to an epoxy group which
has also reacted. Thus
a = PH,PE (3.3)
where PH = probability that a hardener group has reacted
(3.4)
PE = probability that an epoxy group has reacted

_ [Elo - [Ell
(3.5)
PE - [Elo
where [Hlo and [Elo are initial concentrations of hardener and epoxy groups
and [HlI and [Ell are the concentrations at time t.
It should be noted that equal reactivity of hardener functional groups must
apply for this definition of PH to be exact. This is often not the case as will be
discussed in section 3.4 (Durand and Bruneau, 1983). When the initial
concentrations of hardener and epoxy groups are exactly equal and only
reactions between such groups occur PH = PE = P and a = l. The use of
equation 3.1 to calculate Bn , the number of branches in Rank n is only valid
when all reactions are intermolecular. Thus, it becomes more and more
inaccurate after gelation. However, it does illustrate the small degree of
branching for low extents of reaction. Also, the number-average molecular
weight only increases slowly with extent of reaction and is finite at the gel
point (Figure 3.3). The number of average molecular weight can be readily
calculated since it is the total mass divided by the number of molecules
present.
M = Mass (3.6)
n NI
where No is the number of molecules present initially and Nb is the number of
bonds formed, since for each bond formed the number of molecules present
decreases by one. The general case is treated by Macosko and Miller (1976)
who give relations for calculation of number, weight and Z average
molecular weights as a function of the extent of reaction. They have given a
general consistent treatment of non-linear polymerizations which is now
frequently cited in papers dealing with the cure of epoxy resins. Similarities
with the classical treatments of Flory and Stockmayer are pointed out so that
the relationship of their approach to that of others can be appreciated. In
Figure 3.3 the weight average molecular weight is also plotted and it can be
seen that it becomes infinite for P > Pc at the gel point. For P < Pc the
weight fraction of sol is unity, Ws = 1. For extents of reaction greater thanpc
the weight fraction of sol, w" decreases monotomically and the weight
fraction of gel, wg , increases correspondingly (ws + Wg = 1), as shown in
Figure 3.3. Network formation, in terms ofP(X3) and P(X4) is also shown in
Figure 3.3, the former has maximum for P - 0.85, and P(X4) approaches
unity only when P ~ 1. Thus, for the network to be fully 'cross-linked' cure
has to be essentially complete.
A general treatment of gelation was given by Case (1957) and re-
examined by Tanaka and Kakiuchi (1965) who assessed its applicability to
the cure of epoxy resins with acid anhydrides. Other work on the application
KINETICS OF CURE 79
5-------------------------
ws
~--.,1.0
Average I Weight
molecular 4 I 0,8 fraction
weight / and
(arbitrary I probability
units) / P(X n )
3 I
I 0,6
I
I
I
P(Xj__ "
~ 'f', 0.4
/" I',
/ / \
I I \
I I \
"/ IIp (X.) \ 0,2
/, / \
/' '')c./ \
, ,.~ ,.
;' ........
.... -
°0~----OL,'----0~,2----0~,-3---0~.4----~0,-5--JWO,~6~~~-0~,;~--OL,8---~-~0~,-9--~1.8
Extent of reaction - p
Figure 3.3 Effect of the extent of reaction on molecular properties for a stoichiometric mixture
of a diepoxide with a tetra function hardener, e,g.

o
/ \
0
/ \ +
N-R-N
H,
.... H
When the
H"" 'H
amine hydrogens have equal reactivity (but see section 3.4) and epoxy groups only react with
amine hydrogens. (---), weight fraction of sol (ws); ( - - - ) weight fraction of gel (wg ); Ws +
Wg = 1. Mn = number-average molecular weight = (MA + 2ME)/(1 + 2 - 4p), Mw = weight-
average molecular weight. For BADGE + DDM initially Mn (p = 0) = 261 and at the gel
point Pc = Pgel = 0.58 (ideally) Mn(Pc) = Mn(O)/(l - 'l3(Pc)) = 4.4 Mn(O) = 1150 (see Macosko
and Miller, 1976).
of branching theory to the cure of epoxy resins has been discussed by Tanaka
and Mika (1973) who annotate research published by Japanese workers_ A
more general review is that of Dusek (1986) who shows that it is necessary
to consider the kinetics of the epoxy reactions to model the curing process
when there are substitution effects. The statistical theory yields only
approximate results due to neglect of stochastic correlations when a
substitution effect is important. The importance of the gelation and network
formation has led to the development of several different mathematical
models. Gordon has published a large number of papers using graph theory
to model the structure of networks, and a good starting point is the paper
by Gordon and Ross-Murphy (1975). The sol-gel transition is a critical
phenomenon and its relationship to other critical processes is outlined by
Stanley (1986). Ziman (1979) considers macromolecular disorder and deals
with gelation as part of a general discussion of models of homogeneously
disordered systems. Scaling concepts have been used to predict the sol-gel
transition by de Gennes (1979) who introduces the percolation model. The

origin of percolation theory (Grimmett, 1989) is attributed to Flory (1941).
His theory of gelation which was referred to earlier is well worth reading for
an analysis of the basic concepts that any model must satisfy, although it is
essential that a refined theory should allow for the formation of loops due to
intramolecular reactions, especially for accurate prediction of the gel point.
Staufer (1985) has written an introduction to percolation theory and also
reviewed its application to gelation (Staufer et al., 1982). The bond
percolation model is applied to sites or monomers which are linked together
on a d-dimensionallattice by random bonding with a probability p. At a
critical bond probability Pc an infinite cluster appears. Thus P = Pc is the
critical condition for gelation. For P > Pc the largest cluster became
attached to the infinite molecules to form a network, whilst the sol fraction
Ws and the average size of the sol molecules decreases as P ~ 1. Near the
critical condition as P ~ PC' the properties of the system diverge as powers
of f = Ip - Pel. For instance the viscosity is related to Pc by 17 ~ f -k as
p ~ Pc and for p > Pc the elastic shear modulus increases, G ~ fZ. The
experimental determination of a gel point is not as facile as this discussion of
a mathematical discontinuity might imply. Although for p < Pc (or a < a c),
Wg = 0, the extent of reaction is continuous through the gel point.
Measurement of a sol/gel ratio is inaccurate for p only slightly greater than
Pc and the increase in viscosity is not satisfactory, even when it is due to an
increase in molecular complexity and not due to !1T = (Tc - Tg) becoming
small. A method that does yield a good estimate of the gel point is the change
in slope of a curve of log Gil vs tc where tc is the cure time and Gil is the
imaginary part of the complex shear modulus (G* = G' + iG"). This
method was proposed by Harran and Laudourd (1986) and has been
confirmed by Chen and Ellis (1992a) for log E" vs tc. Recently, percolation
theory has been applied to account for the changes in the visco-elastic
properties of an epoxy resin during cure. Harran and co-workers (Serrano
et al., 1989, 1990) determined a relationship between the viscosity of the
curing resin and the extent of reaction X,
(3.7)
where Xc is extent of reaction at the percolation threshold, that is the gel
point, A is a constant and v is the critical percolation exponent. Xc is treated
as an empirical parameter which is determined by a curve fitting procedure.
For cure of tetraglycidyldiaminodiphenylmethane with diaminodiphenyl
sulphone (TGDDM-DDS) they found that Xc and v were temperature-
dependent and concluded that this was due to a change in reaction
mechanism with cure temperature. This could be due to a temperature
dependence of the relative rates of reaction of primary and secondary
amines with epoxy groups (see section 3.4). They exclude the possibility that
in their studies the increase in viscosity was due to the onset of vitrification,
KINETICS OF CURE 81
that is /).T = (Tc - Tg) becoming small. However, it is essential that it is
established that gelation occurs before vitrification for their interpretation
to be generally applicable.
A more detailed examination of the change in visco-elastic properties of
epoxy resins during cure is that of Adolf and coworkers (Martin, 1988; Adolf
and Martin, 1990; Adolf et al., 1990) who also used a percolation model and
determined the critical viscosity exponent.
3.2.2 The glass transition

During cure as well as gelation the other critical phenomenon is vitrification,
as illustrated in the TIT diagram (chapter 1, Figure 1.2). The glass transition
in polymers has been discussed extensively and is treated in textbooks (e.g.
Sperling, 1986; Ward, 1983). The review of Shen and Eisenberg (1966) is
still of value. There is a universality of glass transition phenomena for all
glass-forming systems, such as low molecular weight organic molecules,
organic polymers or inorganic glasses, such as soda-lime-silica and many
others (Scherer, 1986; Rekhson, 1986; Elliott, 1990).
The definition of a glass transition temperature Tg is related to its method
of measurement and will also depend on the thermal history of the sample.
For instance in determination of the volumetric glass transition temperature
a sample is usually cooled from above Tg whilst the differential scanning
calorimetric method (DSC) involves increasing the temperature of the
sample at a prescribed rate from T < Tg , although other thermal histories
may be prescribed. The previous thermal history of the sample and rate of
cooling or heating affects the measured glass transition temperature. For a
curing thermosetting resin, the decrease in molecular mobility with
increasing extent of reaction causes the glass transition temperature of the
sample to increase. Thus, it is with progress of the chemical reactions that
the sample vitrifies and this process will be affected by the cure temperature,
To and the rate of cure which depends on the reactivity of the hardener with
the resin.
Differential scanning calorimetry is a common method for the
measurement of the glass transition temperature of curing epoxy resins,
since such measurements can be combined with studies of the rate of cure
(see section 3.3). However, many different methods may be used to measure
glass transition temperatures (Sperling, 1986; Rabek, 1980a). Also,
different methods of defining Tg have been used (Figure 3.4), such as
1. The temperature of the maximum in the mechanical tan <5 peak.
2. The point of inflexion of the log E' vs T curve or the 'mid' point of the
transition where E' is the storage modulus.
3. An intercept method.
With these alternative definitions there can be large differences between the
(a)
tan I)
Tg Temperature
(b)
9~'----- --~--
I
I
I
----'* -----
I
midpoint
log E' 8 I I
I I
I I
I I
I
7 I
I
Tg m Temperature
(c)
\ .
\ Intercept
9 --11----
log E' 8
Temperature
Figure 3.4 Definitions of the glass transition temperature, Tg • Measurement of dynamic

mechanical properties using DMTA with standard rate of temperature rise. The differences
between these estimates of Tg depend on the 'sharpness' of the transition. For example,
BADGE type resin + DDM: cured short time (V2 hr) at lOO°C ~ - Tt = 40°C; post cured
for longer times at 180°C ~ - T~nt = lOoC (Bell etal., 1992). The 'sharpness' of the transition
increases with extent of cure. ~ is the glass transition determined from the temperature of the
(tan b)max (method a). T~nt is the glass transition determined from the temperature of the
intercept (method c).
estimates of glass transition temperatures. These differences are greater the

lower the ext~nt of cure. Also, it should be appreciated that any method that
involves increasing the temperature of the sample may allow cure reactions
to proceed, thus the glass transition temperature increases whilst it is being
measured. With care it is possible to obtain self-consistent sets of data, but
comparisons between different sets must be treated with caution because of
the range of operational parameters involved.
KINETICS OF CURE 83
Prior to gelation the viscosity of the curing resin increases due to both
an increase in molecular weight and also an increase in its glass transition
temperature. The viscosity of glass-forming systems has a temperature
dependence which is not Arrhenius and for polymeric materials it is often
possible to 'fit' a set of viscosity measurements over a range of temperatures
to the Williams, Landel and Ferry equation (see for example, Sperling, 1986
or Ward, 1983).
- C (T- T)
log (YJ/YJ Tg) = 1 g (3.8)
C2 + (T- Tg)
where the 'Universal' constants are C1 = 17.44 and C2 = 51.6 or
alternatively C1 = 101.6 when the reference temperature Ts = Tg + 50°C is
used in place of Tg. With an increase in the extent of cure Tg increases, I:1T
(= Tc - Tg) decreases and the viscosity of the sample at Tc also increases.
Similar non-Arrhenius viscosity temperature dependence has also been
known for a long time for inorganic glasses. The equivalent temperature
dependence is represented by the Tammann-Vogel-Fulcher equation
(TVF), (Scherer, 1986; Jha and Parker, 1989)
B
In YJ = A + (3.9)
T- To
where To < Tg may be regarded as an 'ideal' glass transition temperature
but is often treated as an arbitrary constant.
Further discussion of the complexities of the glass transition would be out
of place here. However, Stutz and coworkers (1990) have presented a
generalized theory for cross-linked and uncrossed-linked polymers.
3.3 Techniques for monitoring cure
3.3.1 Monitoring cure

During the early stages of cure it is in principle possible to determine the
concentration of epoxy or hardener reactive groups by 'wet' analysis or
alternatively chromatographic methods, similar to those discussed for the
characterization of epoxy pre polymers (chapter 1, section 1.4). However,
there may be difficulties because of interference due to the presence of
different types of reactive groups. When cure has reached the gel stage it is
no longer possible to use methods which involve dissolution of the whole
sample. With finely powdered resin, reactants diffuse only slowly into
particles with a diameter of c. 5.um (Bell, 1970a) and assay of finely
comminuted resin is not a common method for monitoring cure. However,
since cure involves the formation of a network it is not sufficient just to
determine the extent of chemical reaction but changes in the physical and
mechanical properties of the resin with cure treatment, time and
temperature (te, Te) are also important. The methods that are used to
monitor cure as functions of Te and te may be classified as follows.
1. Direct assay of the concentration of reaction groups present, usually
the epoxy concentration.
2. Indirect estimation of the extent of chemical reaction.
3. Measurement of the changes in physical and mechanical properties.
For a full interpretation of the changes occurring during cure it is necessary
to monitor both the concentration of reactive groups and other properties
(Verchere et al., 1990; Adolf et al., 1990).
To obtain data amenable to a kinetic analysis of the rates of the competing
consecutive reactions that occur during cure, the concentration of the
reactive groups must be assayed as a function of cure time, to at a specific
isothermal cure temperature Te. In some cases non-isothermal data have
been analyzed but they are not so reliable as isothermal data. However, it
should be noted that collection of isothermal rate data may involve con-
siderable experimental difficulty. For instance it is not possible for samples
to reach the cure temperature instantaneously, heat has to be supplied to the
sample to raise its temperature, i.e. Tambient ~ Te. Also, the resin is curing so
that there may be temperature gradient across the sample and subsequently
the heat of reaction may cause its centre to become much higher than the
nominal cure temperature Te. This may cause sampling problems since the
extent of reaction may be inhomogeneous.
3.3.2 Direct assay of the concentration of reactive groups

Some of the methods that can be used for the assay of the concentrations of
epoxy and other groups may be divided into two groups:
1. Wet analysis or chromatographic methods which involve dissolution
of the partially cured resin.
2. Assay without dissolution where the whole sample is assayed using
methods such as infrared spectroscopy.
As noted the former methods are only applicable prior to gelation, whilst
the others can be used to the limit of their sensitivity. However, it should be
noted that the sensitivity of both groups decreases as cure time increases. It
is usually difficult to assay the extent of reaction for the 'final' few percent of
the reactive groups. It is easier to detect the presence of residual epoxy
groups but not possible to assay their concentration accurately. Thus, these
methods are insensitive for the 'final' stages of cure when there are often
very significant changes in the properties of the cured resins. Measurement
of the glass transition temperature which has been correlated with an
KINETICS OF CURE 85
estimate of the extent of reaction has been proposed for monitoring cure
since it is sensitive to the processes occurring in the final stages (Wisanrakkit
and Gillham, 1990).
The 'wet' methods of analysis discussed for the characterization of
uncured resins (chapter 1, section 1.4) have been adopted to the assay of
reactive groups still present in partially-cured resins. Problems due to
interference must be eliminated to obtain reliable results (Steinmann, 1989)
and often more than one method of assay is necessary.
Chromatographic methods have been discussed in detail by Meston and
Morris (1984) and both gel permeation chromatography and nuclear
magnetic (NMR) spectroscopy have been used to determine the degree of
branching in a partially-cured epoxy resin (Rogers, 1972). For partially
soluble resins it is possible to determine their high resolution NMR spectra,
usually proton chemical shifts, but the l3C spectra can also be measured. For
example, Sojka and Moniz (1976) have tabulated l3C chemical shifts for
DGEBA and the products of its reaction with piperidine. A general outline
of the theory and experimental methods of NMR spectroscopy with its
application to polymeric materials is given by Rabek (1980b), with also a
description of 'magic angle' spinning, a technique which enables the
chemical shifts in solid polymers to be measured. The application of this
technique to the study of epoxy resins has been discussed in detail by Mertzel
and Koenig (1986).
The integrated intensity of an infrared absorption band is directly related
to the concentration of the absorbing species and two spectral regions have
been used for the assay of epoxy groups. In the 4000-7000 cm -1 (2.5-15 Jlm)
region there is an absorption band at 915 cm- 1 (10.9 Jlm) assigned to the
epoxy ring and in the overtone region 16000-4000 (0.6 to 2.5 Jlm) the band
at 4535 cm -1 (2.205 Jlm) has been used for the assay of epoxy group
concentration. Listings of the infrared absorption spectra of uncured epoxy
resins are given by Lee and Neville (1967) and Hadad (1988a) outlines the
application of infrared spectroscopy to the analysis of epoxy resins. Rabek
(1980c) gives a review of the theory of infrared spectroscopy and also the
experimental techniques available for its application to the study of
polymeric materials.
Quantitative analysis with a dispersive infrared spectrometer for the
4000-700 cm -1 region is now considered to be insufficiently accurate for the
assay of epoxy groups due to the presence of adjacent absorption bands to
the epoxy band and also the insensitivity of the method especially when the
concentration of epoxy groups is low. Some of these problems for the
measurement of 915 cm -1 epoxy band have been resolved by the application
of Fourier-Transform InfraRed spectroscopy (Ff-IR) which has been
reviewed by Mertzel and Koenig (1986).
Measurement of the near infrared absorption (16 000-4000 cm -1) offers
advantages since only strong bands have significant absorption in the
overtone region and hence the spectra are less complicated so that problems
due to the presence of overlapping bands are avoided. Also, the sample path
length is longer, 1-10 mm, and hence sampling is more representative and it
is possible to use a glass cell to confine a liquid sample. Goddu and Delker
(1958) studied the overtone region and Dannenberg (1963) extended their
work to establish calibrations which enable accurate assay of epoxy and
amine group concentrations during cure. Recently (Ellis et al., 1992) the
method has been extended by the design of transparent soda-lime-silica
glass cells so that the extent of cure could be studied from initiation until the
epoxy absorption becomes too low to measure.
3.3.3 Thermal analysis

The basis for the application of differential scanning calorimetric (DSC)
methods depends on the assumption that the measured thermograms can be
directly related to the extent of chemical reaction, with usually a simple
linear relationship between thermal response, e.g. exothermic heat evolved,
and the consumption of epoxy groups. The technique is facile and requires
only small samples and allows a determination of the glass transition
temperature. An outline of DSC methods together with a brief discussion of
sampling techniques with an extensive bibliography is given by Rabek
(1980d). Prime (1981) discusses the detailed experimental procedures that
are required for the accurate application of DSC methods for the monitoring
of the cure of thermosetting resins, including epoxy resins which are also
dealt with by Barton (1985) and Hadad (1988b). Fava (1968) has discussed
methods of analysing DSC thermograms.
Recently, in a study of the cure of DGEBA (Dow-DER 332) with a
stoichiometric concentration of trimethylene glycol di-p-aminobenzoate
TMAB (Polaroid Corp. Polarcure 740M) Wisanrakkit and Gillham (1990)
estimated the fractional conversion from
x = 1- f1Ht
f1HT
where f1Ht and f1H T are the amounts of exothermic heat evolved in the time
internal t = 0 to tc and the total heat evolved. The method involved
integrating the areas under exothermic peaks which requires the speci-
fication of a procedure for 'drawing' a baseline, which has been discussed by
Barton (1985) and Prime (1981) cites a case where a dynamic DSC scan does
not return to the baseline. However, Wisanrakkit and Gillham (1990)
established a good correlation between the extent of reaction and the glass
transition temperature of the curing resin samples.
Barton (1985) discussed, with reference to most of the early work, non-
isothermal reaction kinetics when the temperature of cure is increased at a
heating rate, f3 = dT/dt. Many mathematical approximations (Flynn and
KINETICS OF CURE 87
Wall, 1966) have been applied to analyse the data, and some have been
re-examined by Keenan (1987) who evaluated the errors involved in
estimating activation energies for the curing of two epoxy resin adhesives.
A recent suggestion for the transformation of dynamic DSC measurements
into isothermal parameters for the cure of thermosetting resins is that of
Khabenko and Dolmatov (1990).
3.3.4 Rheological changes during cure

During cure, an epoxy resin changes from a viscous liquid to a rigid glassy
solid, as illustrated in chapter 1, Figure 1.1. Concomitant with the changes in
structure, there are very large changes in its rheological properties, which
are shown schematically in Figure 3.5. It can be appreciated that at tc = 0 the
real part of the complex shear modulus is zero, i.e. G' = 0, but will increase
to about 3 GPa as cure progresses and essentially all of the epoxy groups are
consumed. Ideally G* and its components G' and G" (or E*, E' and E")
should be monitored throughout the cure process. However, it is difficult to
design instruments which are capable of responding to the large changes in
G* = f(Tc> tc). Analogous measurements of the complex dielectric constant
E* and its components as functions of both frequency and tc over a range of
cure temperatures, Tc> are relatively facile and are discussed in detail by
Johari in chapter 7. Also, it is possible to monitor cure in situ using
microdielectrometry sensors (Bidstrup and Sentura, 1989), but interpret-
ation will involve an analysis such as that given in chapter 7 (see also Senturia
and Sheppard, 1986).
Although initially the rate of reaction may be high, the size of the
molecules increases slowly, as was discussed previously. Thus, the early
stages of cure are often monitored by measurement of a shear viscosity. One
of the most convenient instruments is a cone and plate viscometer
(Whorlow, 1980), since the sample size is small and temperature control is
good. Evaporation of a volatile hardener at high cure temperature may alter
the epoxy/hardener stoichiometry and thus cure kinetics will be inaccurate.
However, an advantage of this instrument is that the rate of shear is essential
constant for the whole sample, and it is possible to test for the onset of non-
Newtonian behaviour (Chen and Ellis, 1992a). Macosko (1985) gives an
outline of the application of rheological measurements to monitor the cure
of thermosetting resins including epoxies and the very large increase in
viscosity at the gel point is discussed in detail by Malkin and Kulichikhin
(1991). A fuller understanding of cure processes is attained by measuring
other property changes as well as the shear viscosity. For instance Verchere
et al. (1990) measured the shear viscosity with a concentric cylinder
viscometer and also the extent of reaction by FT-IR spectra, size exclusion
chromatography and DSC. Mijovic and Lee (1989) compared
chemorheological models by combining viscosity and DSC measurements.
8.0
log E*
7.5
7.5
log E"
~------=" 6.5
Log E' Log E"

(arbitrary (arbitrary
units) 7.0 6 units)
5.5
6.5
4.5
30 90 120 150 180
(a) Cure time, tJmins
log E'
0.8
~,
7.5
I \
I \
! \
I
I \
0.6 I \
tan B I \
(arbitrary I \
I \ 7.0
units) I
\
Log E'
tan S / \
\ (arbitrary
0.4
/iJ' \ tan 8 units)
I I \
I I \
I I \
! I
\
\
I \ 6.5
I \
,
0.2 I \
I \
I \
J
---
\\\
I
J
30
---- ~/
60 90 120 150 180

(b) Cure time, tJmins
Figure 3.5 Dynamic mechanical properties as a function of cure time. (a) Log E* and log E"
vs te' The gel time, t e .ge !> can be determined from the change in slope of log E" vs te' (b) Log E'
and tan c5 vs te' The shoulder at lower cure times on the tan c5 vs tc curve is due to gelation.
The shear viscosity of a resin during cure may increase due to either
gelation or the onset of vitrification, that is, I:!T = Tc - Tg(tel becomes
small. Thus, the whole curve process cannot be monitored by measurement
of a shear viscosity, it is necessary to measure a complex modulus and its
KINETICS OF CURE 89
components (e.g. G*, G' and G"). Such measurements are difficult for a
sample which is liquid initially and eventually solid (Figure 3.5). An early
solution to this problem was the torsional braid technique introduced and
developed by Gillham and coworkers (see Aronhime and Gillham (1986) for
a review). An instrumented Torsional Brain Analyser is commercially
available and there are several other instruments available for the
measurement of the dynamic mechanical properties of polymers. The
technique is known as dynamic mechanical thermal analysis, DMT A, or
DMA (Flynn, 1989). The instruments are equipped with a computer and
suitable programmes which enable measurement of the changes in the
complex modulus and its components by factors of 103 .
Adolf and Martin (1990) have measured the changes in G* and its
components as functions frequency and extent of reaction and interpreted
their results using a percolation model. Harran and Landourd (1986) found
that the slope of a plot of log G" vs tc decreases at the gel point, and Chen and
Ellis (1992b) have confirmed that d (log E")/dtc decreases at the gel point.
More recently Harran and coworkers (Serrano et aI., 1990) have applied
a percolation model to analyse their complex shear modulus data, as
mentioned previously.
3.4 Kinetics of cure
3.4.1 Introduction
The determination of the kinetics of cure of epoxy resins involves more than
just measurement of the rates of reaction of epoxy groups with the hardener.
This is because it is necessary to locate the gelation time, tc(gel), and also the
rates of reaction become diffusion-controlled with the 'onset' of
vitrification, i.e. when I1T = Tc - Tg(t,) becomes small. It is possible to
inhibit reaction between hardener and epoxy groups by reducing the
temperature so that the resin/hardener vitrifies. However, for elucidation of
the mechanisms of the chemical reactions occurring during cure it is
necessary to study the initial rates of reaction and also assay the extent of
reaction between epoxy and reactive hardener groups. Usually the epoxy
group conversion is determined but assay of the concentration of hardener
reactive groups is also desirable, especially when their reactivity is unequal,
as for example primary and secondary amine hydrogen atoms. Two
approaches have been used:
1. Generalized empirical rate equations.
2. Rate equations derived from proposed chemical mechanisms.
Both of these approaches will be considered and also the effects of
vitrification when the rates of the curing reactions become diffusion-
controlled.
3.4.2 Amine-curing agents
3.4.2.1 Introduction. The importance of amine-curing agents is clear

from chapter 2, and there have been many studies of the rates of cure of
epoxy resins with amine hardeners. Barton (1985) tabulates 26 resin amine
hardener systems, together with associated rate equations that have been
determined using DSC methods.
The possible reactions that may be of importance in the cure of epoxy
resins by polyamines are reaction of epoxy groups with primary amines,
secondary amines, hydroxyl groups or other epoxy groups. Reaction of
epoxy groups with primary amines
OH
/0" k~ . I
~CH-CH2 + H-N-H ~ ~CH-CHz-N-H
~ ~
where k~, is the rate constant for the uncatalysed reaction, but the reaction is
often catalysed by hydroxyl or other groups and then the rate is faster, i.e.
kJ > k~
OH
/0" k, I
~CH-CH2 + H-N-H + [Catlo ~ ~CH-CH2-N-H + [Catlo
$ ~ .
where [Cat]o is the effective concentration of the adventitious catalyst

present initially in either or both the resin and hardener, and its concen-
tration is regarded as essentially constant. It will be noted that for each
epoxy group that reacts a hydroxyl group is formed which can also catalyse
the reaction between amine and epoxy groups. Thus the reaction will be
autocatalytic and the rate equation will contain a rate constant kll and
concentration term, [P-OH], for the concentration of active hydroxyl groups
formed, and its concentration will increase during cure. A similar set of
equations can be written for the reactions of secondary amines with rate
constants k~, k2 and k 22 . Other reactions may also occur, such as that
between epoxy and hydroxy groups:
-t- OH
~ I
H-C-0-CH 2-CH ~
$
and also epoxy groups may react with each other when a suitable catalyst is
present and the cure temperature is high enough.
3.4.2.2 Approximate kinetic analysis. A full kinetic scheme for the

possible reactions that may occur during the cure of an epoxy resin with a
primary amine would require the formulation and solution of a set of
coupled differential equations. There would be a rate equation for each
KINETICS OF CURE 91
species that is involved in the curing process, with a term for every reaction
of that species. Thus, the rate equation for the consumption of epoxy groups
by reaction with primary amine hydrogen could be
dE 0
- - = k j EA j + k j Co EA j + k11 EAj[P-OR]
dt
+ k~ EA2 + k2 Co EA2 + k22 EA2[P-OR]
+ kOH E[P-OR] + kE E2[Cat] (3.10)
where E, Aj, A 2, etc. represent concentrations and the ks are rate constants.
The first set of terms is for the reaction of the primary amine, with
concentration Aj, the second for the reaction of the secondary amine with
concentration A 2, which of course will depend on the rate of consumption of
primary amine. There are also terms for the reaction of epoxy groups with
the hydroxyl group which is a product of the epoxy reactions and also for
reaction with itself. In view of this complexity, various simplifications have
been used and are often the basis for the empirical equations that have been
proposed.
The kinetic treatment of Rorie and co-workers (1970) has been the basis
for many subsequent treatments of the kinetics of amine cure. They
considered that only the catalysed reactions of epoxy groups were
important, that is k~ = k~ = kOH = kE = O. It is useful to define the relative
rate, K of the catalysed reaction of secondary to primary amine hydrogens,
that is K = k2/kj, and when K = 0.5 the rate equation 3.10 can, using
Xe = Eo - E and some algebraic manipulation be expressed as
(3.11)
When K is approximately a half then with use of K = 0.5 + ,'!,x and also
using the stoichiometric relation for reacted epoxy and active hydrogen
atoms Aj + A2/2 = Ao - xe/2, when A 20 = 0, equation 3.11 can be re-
arranged to
1 . dxe = (k j Co + k 11 x e) [1 + 2A2 .I"!.,x ] (3.12)
(Eo - xe) (Ao - xe/2) dt 2Aj + A2
A plot of the left hand side of equation 3.12, regarded as a reduced reaction
rate, versus xe will be initially linear when the conditions used to derive
equation 3.12 apply. This is because the term [2A2 .I"!.,K/(2Aj + A 2)] is
initially zero and small compared with unity up to maybe 50% consumption
of epoxy groups. Rorie et al. (1970) regarded their linear plots of reduced
reaction rate versus conversion to be satisfactory for the initial stages of
cure. Subsequently these plots have a maximum because the rate of con-
version decreases as the curing reactions become more and more diffusion-
controlled as I"!.,T = (Tc - Tg) becomes small.
Often the extent of reaction is expressed in terms of a normalised con-

version Xe = (Eo - E)/Eo = xe/EO and hence
dXe _ 1
dt - Eo
(dx e)
dt .
When the initial stoichiometry is exact and the concentration of active
amine hydrogen atoms and epoxy groups are exactly equal, Ao = EoI2.
Then using these substitutions equation 3.12 becomes:
1 2. dXe = [K1 + K2X e] He (3.13)

(1- Xe) dt
where K1 = V2(k1 Co Eo); K2 = V2(k11 E6) and He = [1 + (2A2 • /)..1(/

(2A1 + A 2)]. He is the Horie connection factor for the deviation of K from
0.5. Thus, when a stoichiometric concentration of hardener is present a plot
of [1/(1 - Xe)2( dXe/dt)] vs Xe should be linear until He becomes significantly
larger than unity and/or the rates of the curing reactions decrease because
they become diffusion-controlled.
Another approach which has often been applied by others is derived from
that of Sourour and Kamal (1976) who accept the postulates of Horie et al.
(1980) except that they regard the relative reactivity of primary and
secondary amines as equal, i.e. K = 1 and the rate equation they propose is
(3.14)
where 1<1 = k1 Co Eo and 1<2 = 1(11 E6 Xe are constants and R = HolED the
ratio of the initial concentrations of active hydrogen atoms to epoxy groups.
Equation 3.14 can be derived for a catalysed reaction between active
hydrogen atoms and epoxy groups and also a hydroxyl group autocatalysed
reaction:
-Ill
E + H + Co --> A
E + H + [P-OH] ~ A
The rate equation is

dE
- - = k1 Co EH + k11 EH[P-OH] (3.15)
dt
which can be rearranged using E = Eo (1 - X e ), H = Ho - (Eo - E) =

Ho - EoXe = Eo(R - Xe) with R = HolED and noting that [P-OH] =
Eo - E = EoXe. For a stoichiometric composition, R = 1, equation 3.14
can be put into the form
(3.16)
KINETICS OF CURE 93
which has the same form as equation 3.13 but with He = 1 and the constants
are different. Thus, the linearity of plots of the left hand sides of equations
3.13 or 3.16 vs Xe will not be sensitive to the value of I( if it is either 0.5 or 1.0.
Thus, the value of the relative rates of reaction of primary and secondary
amine hydrogen can only be found when the values of kl' k2 or 1(1 and 1(2 can
be determined separately with sufficient accuracy.
A generalisation of equation 3.16 which has been used and regarded as
'quasi-theoretically' justified is:
dXe(K +K Xffi)(l-X)ffi (3.17)

dt 1 2 e e
However, it is difficult to relate this equation to the reaction steps that occur
during an amine cure of epoxy resins. A simplified form of equation 3.17
with K] = 0 has also been used by Kamal and Sourour (1972,1973):
dXe = KXffi(l _ X )n (3.18)

dt e e
Macosko (1985) and Camargo et al. (1983) suggest that a general rate
expression which is useful for fitting conversion data is
dXe = k(l- xet ~ k;X~ (3.19)

dt ;=0
where in some cases the second term can be regarded as an autocatalytic

parameter. If the reactions are not autocatalytic m = 0 and hence equation
3.18 becomes
d!e = k(l - xet (3.20)
which is the rate equation for an nth order reaction (Wisanrakkit et al. ,
1990).
Barton (1985) gives a list of the rate equations that have been fitted by
about 20 different kinetic studies with the extent of cure measured by DSC
methods. However, from the present discussion it can be appreciated that
the kinetics of cure may be complicated, even before the rates of reaction
become diffusion-controlled due to vitrification. The empirical equations
yield only limited information on the mechanisms of the chemical reactions
involved. However, they have utility by providing a rate function for the
analysis of kinetic data, such as that obtained from DSC measurements.
Thus they may be sufficient for specification of cure conditions for specific
resin-hardener combinations. Also, they may be useful for quality control
and the in situ monitoring of the cure process.
Mijovic and coworkers (Mijovic, 1990; Mijovic and Wang, 1989; Mijovic
et al., 1984) have been studying the cure of epoxy resins over a number of
years and recently (Mijovic and Lee, 1989) used equation 3.17 with the
'overall' reaction order, m + n = 2, and solved it using a fourth order
Rungge-Kutta numerical integration to obtain curves for the extent of
reaction versus time of cure for several isothermal cure temperatures. Their
system was DGEBA-cured with a mixture of 4,4 methylene-dianiline
(MDA) and m-phenylene diamine (m-PDA) and they applied branching
theory to model the increase in viscosity prior to the gel point.
Sheppard and Senturia (1986) as part of their programme of studies of the
effects of cure on the dielectric properties of epoxy resins have also
measured the extent of cure using DSC methods. They found that plots of
1 dXe
- vs X
(1 - Xe)2' dt e
(equation 3.16) were linear for the cure of DGEBA with DDS with cure
temperatures in the range 410-460 K. They also compared the time
required for 50% primary amine conversion estimated from the kinetic
model and obtained satisfactory agreement with experimental determina-
tions (see also chapter 7).
An even simpler rate equation is obtained when m = 0 in equation 3.18,
an expression used by Acitelli and coworkers many years ago (1971) and
recently Wisanrakkit and coworkers (1990) obtained a first order fit with
n = 1. For cure of DGEBA (Epon 828) with PACM-20 (see chapter 2) plots
of dXe/dt vs (1 - Xe) for cure at 61°C were linear. Keenan (1987) discusses
in detail a method for determining the parameters 1(2, m and n with Kl = 0
when a simpler form of equation 3.17 is applicable. His method involved
measurement of an exothermic peak and the procedure was satisfactory for
representation of the cure of FM123-5 (American Cyanamid Co. ), a nitrile
modified-dicyandiamide-cured epoxy adhesive supported on a nylon
carrier.
An interesting alternative approach is that of Pin sheng and Chune (1989)
who apply the non-equilibrium thermodynamic fluctuation theory proposed
by Hsich (1982). In this theory the change in a mechanical property is
measured and the rate equation derived after some simplification is:
G -G(t) = exp [- (tIT) fJ]
00
Goo - Go
where Goo, G(t) and Go are the final, time-dependent and initial values of
the physical property, f3 is a constant which depends on the width of the
relaxation spectrum and r is the relaxation time. Pingsheng and Chune
studied the cure of a commercial DGEBA resin E51 (Shanghai Resin
Factory) with imidazole as the hardener and silica filler. They were able to
model the change in the shear modulus as a function of both cure
temperature and time.
KINETICS OF CURE 95
3.4.2.3 Analysis of competitive consecutive reactions. The rate equation
for the consumption of epoxy groups, equation 3.10, contains terms for the
concentration of secondary amine and hence to specify the kinetics of cure,
additional rate equations are required. When k~ = k~ = 0 the additional
rate equations are
- dAI = ki Al ECo + kl1 Al E[P-OH] (3.21)

dt
dA 2 = ki Al ECo + kl1 Al E[P-OH] - k2 A2 ECo - k22 A2 E[P-OH]

dt (3.22)
dA3
- = k 2A 2ECo+ k22A2E [P-OH ] (3.23)
dt
and solution of these coupled equations is difficult, even when the
autocatalytic terms are unimportant, which would be equivalent to the case
discussed in detail by Frost and Pearson (1961) (see also Benson, 1960).
However, it has been accepted, especially after the work of Smith (1961),
that the reaction of amines with epoxy groups in epoxy resins is
autocatalytic. An interesting approach to the solution of these coupled
differential equations is that of Dobas et al. (1975). They regarded k~ =
k~ = 0 = ki = k2 = kOH = kE so that equations 3.10 and 3.21-23 are
reduced to
- -dE = k l1 EA I [P-OH] + k 22 A 2E[P-OH] (3.24)

dt
- dAI = kl1 EAI[P-OH] (3.25)

dt
dA 2 = kl1 EAI[P-OH] - k22 A2 E[P-OH] (3.26)

dt
- dA3 = k22 A2 E[P-OH] (3.27)

dt
Thus, in this reaction scheme, only autocatalytic reactions are con-
sidered. Dobas et al. (1975) proceed by defining Y = EAI[P-OH] and
Z = EA2[P-OH] so that from equation 3.25
kil = - dA I . ~ (3.28)
a dt Y
and from equation 3.27
k22 = _ dA 3 . ~ (3.29)
a dt Z
Other estimates of k12 can be obtained from three equations of the form
k == [dA2 _ dE] _1_ (3.30)

llb dt d t ' 2Y
and a similar set of three equations for k22
k 22b == _ [dE + dA 2] . _1_ (3.31)

dt dt 2Z
Thus there are a total of eight equations which allow estimation of kll and
k22 from measured concentrations, E, Aj and A2 and also dE/dt, dAj/dt
and dA 2/dt. The concentrations were measured using near infrared
spectroscopic assay and complete sets of all constants were determined
for the reactions of seven diamines with p-tolylglycidyl ether at 80°C and
2,2-bis(4-(2,3-epoxypropyl) phenyl propane at several cure temperatures
and related to the pKa . j values of the diamines.
It is possible to 'remove' times as a variable (Dusek et al., 1975) by
dividing equation 3.26 by 3.25 to obtain:
dA2 == k22. A2 _ 1 (3.32)
dA j kll Aj
which can be integrated using the initial condition that Aj == AlO when
A2 == 0, that is only primary amines are present initially. Charlesworth
(1980) used this procedure to determine the value of 1(, the ratio of the rate
of reaction of primary and secondary amine hydrogens. Re studied the
reaction of phenyl glycidyl ether with amine at 100 and 150°C, I( == 0.44 at
100°C and I( == 0.71 at 150°C.
Williams and coworkers (Riccardi et al., 1984; Riccardi and Williams,
1986a, 1986b) have published a number of studies of the reaction of amines
with epoxy resins and in a recent paper (Verchere etal., 1990) report a study
of the reaction of a diamine with a monoepoxide, 4,4'diamino-3,3' dimethyl
dicyclohexylmethane (3DCM) and phenyl glycidyl ether. They distinguish
the rates of reaction of a primary amine hydrogen on alternate amine groups
in the diamine (see Figure 3.6). It should be noted that this hardener is
'hindered' (see chapter 2).
Thus they have a set of six coupled differential equations for the reaction
of the four amine hydrogen atoms. These rate equations are formally
autocatalytic so that for the reaction to start, they postulate that the
concentration of hydroxyl groups is:
[OR] == Co + Xe Eo
where Co is the concentration of catalytic impurity. This is equivalent to
regarding the. rate constants for the two catalysed reactions of amine
hydrogens with epoxide groups as essentially equal, that is, k j == k12
and is equivalent to equation 3.25 but with [P-OR] replaced by [OR] ==
KINETICS OF CURE 97
~
k',
1':
k",
KA,A,
A,-A, A,-A,
l~
t-<
A,-A,
Figure 3.6 Reactions of diamines. A]-A] unreacted primary diamine; A]-A2 unreacted
primary amine and a secondary amine; A3 reacted secondary amine, i.e. a tertiary amine. The
epoxy prepolymer unit is omitted except for A]-A2 . Verchere et al. (1990) measured the
concentrations of ArA2 plus A]-A3; A2-A3 and A3-A3 by size exclusion chromatography, and
used a kinetic scheme with k] = k'{ = k; = k] and k'2 = k z = k 2 .
Co + XeEO. The set of differential equations was solved by eliminating the

dependence on time by division, as discussed previously, and then solved.
The concentrations ofPGE, [A2 - A 2] plus [A3 - Ad and [A3 - A 2] plus
[A3 - A 3] were measured as a function of reaction time at 50°C by size
exclusion chromatography. They obtained agreement between experi-
mental and calculated concentrations as a function of Xe with a value
of 1< = 0.4. Verchere and coworkers (1990) regarded their study of the
reaction of 3DCM with PGE as a model for the cure of DGEBA with that
hardener. They also treated this curing reaction as purely autocatalytic in
the pre-gel stage, and measured the decrease in concentration of BADGE
by SEC and were able to calculate Mn and Mw as a function of X e, and
compare experiment with branching theory. They found that vitrification
does not affect the statistics of network formation.
In a study of the cure of BADGE with five aromatic amines, Grillet and
coworkers (1989) regarded the predominant reaction mechanism to be non-
catalytic for higher cure temperatures, that is Tc > lOO°e. They also
eliminated the time variable by dividing the set of differential equations by
the one which contained only one concentration-dependent term, a similar
procedure to that proposed by Dusek eta!' (1975) and also used by Verch ere
(1990) referred to previously. From viscosity measurements and SEC they
determined the reactivities of the diamines and found that they increased in
the order:
4,4' DDS < 3,3' DDS < BAPP < BAPS < DDM
They also discuss methods for estimating the relative reactivities of

secondary to primary amines, K = kzlk 1 and found that K < 1, with K - 0.8
to 0.9 for DDS diamines.
An interesting study that shows that the reaction of secondary amine
hydrogens with an epoxy group is very much faster than either epoxy-
hydroxyl or tertiary amine catalysed epoxy-epoxy reaction is that of Adolf
and Martin (1990). They studied the reaction of Epon 828 with
diethanolanine,
....CH 2-CH 2-OH
H-N
'CH 2-CH 2-OH
and the secondary amine reacts completely within 5 min at 90°C with about
115 of the epoxy groups to produce a secondary hydroxyl group. The
reaction then proceeds by tertiary amine catalysed reaction of both the
primary and secondary hydroxyl groups with equal probability. The gel
point occurs after about 3 h at 90°C. Fully cured resins have a glass transition
temperature of c. 70°C. This study confirms that in equation 3.10
kOH = kE = 0 compared with the rates of reaction of primary and secondary
amines with epoxy groups.
For the cure of BADGE with either TET A or diaminodiphenyl methane
Chen and Ellis (1992a) have extended the approach given by Frost and
Pearson (1961) so that both catalysis by impurities and also the product
hydroxyl groups are included in the rate equations. For TET A it was found
that the relative rates of reaction of secondary to primary amines increased
with cure temperature, at 22°C K = 0.25 and increased to K = 0.75 at 50°C.
Thus, the activation energy for the reaction of secondary amines was larger
than that for primary.
3.4.3 Carboxylic acid anhydrides

Although acid anhydrides are used extensively as hardeners for epoxy resins
(see chapter 2), studies of the kinetics of their curing reactions are far fewer
than those for amine cures. Also, the kinetics appears to be more
complicated and at the present time there is not a generally accepted rate
equation for cure with these hardeners. The orders of the reaction that have
been reported range from 0 to 4 and the statement of Mertzel and Koenig
(1986) that 'a generally accepted reaction mechanism is not available' still
applies.
Barton (1985) has reviewed the kinetic DSe studies of acid anhydride
cures of epoxy resins, listing the kinetic parameters derived using an
KINETICS OF CURE 99
empirical rate equation
dX
dC I = k{1-Xt (3.33)
dt dt Co
where Co is the initial concentration and k is an 'apparent' rate constant

defined by k = ko C3 -1, and has an Arrhenius temperature-dependence.
The values of n in equation 3.33 ranged from 0.2 to 2 and the apparent
activation energy from 56 to 159 kJ/mole, depending on the resin/hardener/
catalyst combination studied. Although about half of the reactions are first
order, n = 1, most of the others are fractional and it is not possible to relate
them to a simple reaction mechanism.
Peyser and Bascom (1977) discuss the analysis of DSC data in some detail
and found that for the cure of DGEBA with hexahydrophthalic anhydride
with benzyldimethylamine as a catalyst, the cure was first order with an
activation energy of 25 kcallmole. Stevens and Richardson (1983) defined
the extent of reaction, P(t,) , in terms of the time-dependence of the glass
transition temperature.
P _ Tg(t,) - Tg(o)
(tel - T T
(3.34)
g(oo) - g(O)
and of course the use of such a parameter depends on the definition and
experimental method used to determine Tg(O) , Tg(Tel and Tg(oo). Also, it is
essential that the limited value Tg(oo) is determined unambiguously. They
found that for cure of DGEBA type prepolymers, Ciba-Geigy CT200 and
CY207, with phthalic anhydride and also a mixture of phthalic and tetra-
hydrophthalic anhydrides that the overall kinetics is neither first nor half-
orders, but first order kinetics was satisfactory for 85-90% of the
converSIOn.
Steinmann (1987, 1989) has considered many of the previous reports on
anhydride cures and reports on her own work in which she studied the
reactions of the several epoxy resins with hexahydrophthalic anhydride
(HHPA) and tertiary amine catalysts.
3.4.4 Diffusion control

It is obvious that for any chemical reaction to occur the reactive centres must
be in close proximity to each other and also be orientated so that a product is
formed. Usually in liquids these processes may be regarded as relatively fast
and their rate determines the observed reaction rate constant (North, 1964;
Entelis and Tiger, 1976). For the cure of epoxy resins prior to vitrification or
gelation the reactions occur in the 'liquid' state and the rate of cure is
controlled by 'chemical kinetics'. However, when the viscosity increases the
rate of cure then depends on diffusion of reactant to a reactive centre, that is
the reactions are 'diffusion-controlled' (Eyring et al., 1980). This is the

factor which limits the extent of reaction at lower cure temperatures. The
reaction between epoxy resin and hardener may be quenched by storing well
below the glass transition temperature. Cure may be 'activated' by raising
the temperature.
The onset of diffusion-controlled polymerizations and cure of
thermosetting resins has been reviewed in considerable detail by Mita and
Horie (1987). They discuss some epoxy resin systems and show that the
effect of diffusion control of their rates of cure can be represented on
Gillham's TTT diagram (Enns and Gillham, 1983; Aronhime and Gillham,
1986), see chapter 1, Figure 1.2. It is importantto note that it may be difficult
to determine the rate constants for the reaction of primary and secondary
amines with epoxy groups without the onset of complicating processes, even
when the product is not a covalently bonded cross-linked network. For
instance, Rozenberg (1986) discusses the autoinhibition of the reaction
between phenyl glycidyl ether with aniline. Even at longer reaction times the
conversion is limited to about 0.8 due to effects of hydrogen bonding which
reduces the rate of reorientation of the reaction product. This is a factor that
may affect the ratio, 1( = k21kb of the rate constants for the reaction of
primary and secondary amines with epoxy groups in this and other systems.
Not only is it necessary for the mutual reorientation of potentially reactive
groups to occur but a prior requirement is that they should diffuse together
so that they are within a 'reaction zone' (Chiu et al., 1983; Sharma and
Soane, 1988). It is the diffusion coefficient ofthe smallest free molecule with
a reactive group which will have the greatest effect on the overall rate of
reaction. Most of the discussion of the factor affecting the onset of diffusion
control of the rate of cure of epoxy resins has only considered the increase in
the shear viscosity of the reacting mixture. However, it is a 'local molecular
viscosity' which will determine the rate of diffusion of the smallest reactive
molecule to a reactive site on an oligomeric molecule. The local viscosity has
been determined for small molecules in some polymeric matrices. Guillet
(1985) estimates that the microviscosity of polymeric systems is many orders
of magnitude lower than their 'macroscopic' viscosity, that is their shear
viscosity. He also observes 'that this fact has gone unrecognized in many
studies of polymer reactions'.
At more advanced stages of cure, intramolecular reaction between epoxy
and other reactive sites on the same molecule or in the gel become
dominant. Thus, the rate of segmental reorientation dominates the cure
process and has been discussed by Mita and Horie (1987) and also
Rozenberg (1986) who outlines a model based on reptation theory (Larson,
1988). Rozenberg also discusses the 'topological reaction limit', since it is
possible for potentially reactive groups to be isolated within the three-
dimensional network. However, it appears that for a 'fully' cured resin, the
concentrations of such unreacted groups is small, but because of the
KINETICS OF CURE 101
insensitivity of analytical methods, it is not possible to give a precise estimate
of their concentration.
Dusek (1986) discusses diffusion control of the cure of epoxy resins in
terms of a WLF temperature-dependent rate and concludes that branching
theory can be applied to reactions occurring in the melt, rubbery and glassy
states. In the vicinity of T g , molecular mobility is a function of free volume
(Dusek, 1985) so that log k ex r, the segmental mobility (Lunak et aI., 1978)
and epoxy curing reactions are quenched when !1T ~ - 30 K (Dusek,
1985).
Others (Huguenin and Klein, 1985; Wisanrakkit and Gillham, 1990;
Sandford and McCullough, 1990) have discussed diffusion-controlled curing
reactions using free volume models coupled with the Rabinowitch theory
(1937), a procedure which was also suggested by Havlicek and Dusek
(1987). The overall rate constant, ka, has both a temperature and extent of
reaction dependence which can be formally represented by
In the treatment of Wisanrakkit and Gillham (1990) kT is the scaled rate

constant with an Arrhenius temperature-dependence and kd is the
'diffusion' rate constant which has a WLF type temperature-dependence.
They give a discussion of a method for determining these two rate constants.
Matsuoka et al. (1989) have proposed a model based on a Kohlrausch-
Williams-Watts (Williams and Watts, 1970) distribution of relaxation times,
which is a function of the extent of cure. They applied their theory to the
cross-linking of a commercial epoxy-novolak resin which is used for the
encapsulation of integrated circuit chips. With a cure temperature of 140°C,
the limiting glass transition temperature attained was 173°C, with very little
increase in extent of cure for 1 < tc < 10 h. Thus, the cure reactions were
quenched at about 90% conversion when !1 T ~ - 30°C as suggested by
Dusek (1985).
Yu and van Meerwall (1990) report a very interesting application ofNMR
pulsed gradient spin echo methods for the measurement of the diffusivity
of molecules during cure. They studied two epoxy resins, DGEBA and
diglycidyl ether of 1,4 butanediol (DGEB), both cured with DDS at 120 and
140°C. The self-diffusion of both polymer and probe-molecules including
free hardener were measured. There was no effect of gelation on the
diffusivity but Dpolymer decreased by three orders of magnitude during cure.
The change in Dprobe was much smaller which indicates that it is the 'local'
viscosity which is the important parameter determining the rate of cure as
!1 T = Tc - Tg becomes smaller and smaller.
This discussion of diffusion control of the extent of cure when !1 Tis small
provides an explanation for the normal industrial practice of elevated
temperature post-curing to attain optimum glass transition temperatures.
The relationship between extent of cure and the glass transition temperature
is discussed in detail by Stutz et al. (1990). However, it may be difficult to
determine the extent of cure as the concentrations of reactive groups
approach zero. Even a DSC zero residual heat does not necessarily indicate
'full' cure (Stutz et al., 1990; Stutz, 1992). Also, Stutz (1992) recommends
control of the final stages of cure by measurement of the amount of solubles.
Although this parameter may require experimental experience to obtain
accurate measurements it can be used to calculate a degree of cure using
branching theory (Macosko and Miller, 1976; see section 3.2).
3.5 Effect of cure on mechanical and related properties
3.5.1 Introduction
Epoxy resins find applications in areas such as coatings, adhesives or matrix
materials for composites, as discussed in chapters 7, 8 and 9. One of the
properties required for these applications is that the material should be
essentially a rigid solid at the use temperature, Tuse, although in some cases
a degree of flexibility is required (see chapter 4). The use temperature in all
these applications is below the glass transition temperature of the cured
resin, Tuse < Tg• For higher temperature applications there is continued
development of epoxy-hardener systems with higher and higher glass
transition temperatures and hence higher Tuse. Of course it is important to
ensure that chemical degradation, due to oxidative, photochemical or other
processes, do not impair the properties of the cured resin. These
relationships are represented schematically in the TTT diagram, Figure 1.2.
In many applications of epoxy resins it is their glassy state elastic moduli
that are of paramount importance. Modulus maps for amorphous polymers
have been proposed by Gilbert et al. (1986) who confined their discussion to
thermoplastics, so-called 'linear' polymers, and considered poly(methyl
methacrylate) and poly(styrene) in detail. There is a need to represent the
structure-property relationships for thermosetting resins. The elastic
properties will be a function of the structure of the network and the
measurement temperature relative to the glass transition temperature of
the cured resin. Thus, an elastic property P = f(T,C) where T is the
measurement temperature and C is a cure parameter. Choice of measurable
quantities to define C presents several problems. Of course, C is related to
extent of conversion, X e , which is difficult to measure accurately when the
concentration of reactive groups becomes very low, as discussed previously
(see section 3.3.2). Also it is important to note that the relationship between
Tg and Xe is non-linear (Wisanrakkit and Gillham, 1990; Wang and Gillham,
.... Tg CXl
sol fluid region

. ~.:'fJ(I _ glass transition
I
\l1..{~\C! -
_4--
I
TgO ....
Gel
o
Cure Parameter, C
Figure 3.7 The CfP diagram: a modified version ofa TgIP diagram (Wang and Gillham, 1992)
C = Tg - Tgo
Tg", - Tgo
Tgo is the initial glass transition temperature of the uncured resin; Tg", is the glass transition
temperature of the fully-cured resin; Tg is the glass transition temperature of the partially-cured
resin.
1992), and would complicate any representation of the relationships

between properties and extent of conversion. Wang and Gillham (1992)
suggest that the glass transition temperature can be used as a measure of
cure, C, and have presented a Tg - TP diagram. Their proposal has been
slightly modified by replacing Tg by C in Figure 3.7, in which a normalized
definition for the state of cure is used,
Tg - Tgo
C = ----::~~ (3.35)
Tgoo - Tgo
where Tgo is the initial and Tgoo is the 'fully cured' glass transition
temperature. Such a definition requires that Tg is a measure of the structure
of the cured resin regardless of the precise cure treatment, that is, cure
temperature and cure time. With this definition C increases from zero to
unity as cure progresses.
An earlier review of the effects of cross-linking on the properties of
polymers is that of Nielsen (1969), which is still useful as a starting point for
the study of the relationships between properties and cure of thermosetting
systems. It should be noted that there has been more recent discussion of the
relationship between the glass transition temperature and the structure of

the cured resin (see for example Banks and Ellis (1982) and Stutz et al.
(1990)). Kaelble (1973 and with co-authors, 1988) and Oleinik (1986) review
epoxy resins. The latter discusses much of the original work published in
Russian.
The change from a liquid to eventually a visco-elastic solid during cure of
an epoxy resin was discussed in chapter 1 and illustrated in Figure 1.1. The
importance of the TgTP (or CTP) diagram (Figure 3.7) introduced by Wang
and Gillham (1992) is that it provides a generalized framework for
correlating the physical properties of thermosetting materials.
3.5.2 Glassy moduli

When visco-elastic relaxation effects are absent the elastic moduli will
have a low temperature and/or high frequency limit, such as Ego, for the
Young's modulus. The moduli that are measured experimentally at higher
temperatures and lower rates of strain will be lower than Ego. When either
the measurement temperature decreases or the rate of strain is increased the
moduli increase. A major cause of this increase in modulus with decrease in
temperature is due to a decrease in the volume of the resin. As the density
increases the separation between adjacent chain segments decreases so that
the van der Waals forces between intermolecular atoms increase, which of
course leads to an increase in the elastic moduli. For thermoplastics, Gilbert
et al. (1986) suggest a linear relationship
(3.36)
where Eo is the Young's modulus at 0 K, am is the temperature coefficient of
modulus, which is regarded as constant in the range T to Tg • It should be
noted that equation 3.36 is only applicable for the glassy modulus, that is
T < Tg • Bondi (1968) has suggested an equation 3.37, which has a similar
form to equation 3.36, for T < 0.9 Tg •
E*(T) = Eo [1 - Eob .:£]

2
Tg
(3.37)
where E* = EVw/D..Hs is a dimensionless reduced modulus, and Hs is a

lattice contribution to the enthalpy and Vw is the van der Waal's volume of a
repeat unit. Bondi gives data for the calculation of molecular structure
increments and a value for b, a dimensionless constant, of about 30. Morel
et al. (1989) have measured Young's modulus for a series of epoxy resins and
find that Bondi's method overestimates the elastic modulus.
The major factor that determines the elastic moduli of glassy polymers is
the cohesive forces between molecules, or molecular segments. An estimate
of these attractive forces is provided by the cohesive energy density of the
material and Tobolsky (1960) suggested that the bulk modulus, B, of a
polymeric glass can be estimated from
B = 8.040 2 (3.38)
where 0 is the solubility parameter (Barton, 1983), which is the square root
of the cohesive energy density. Kaelble (1973) and Kaelble et al. (1988),
have used Tobolsky's method to estimate the elastic moduli of epoxy resins.
Since they are interrelated by
E = 3B(1 - 2v) = 2G(1 + v) (3.39)
where v is Poisson's ratio and G is the shear modulus it is possible to
calculate E from B when v is known. Kaelble sets v = 0.33 to obtain
E = 2.67G = 8.04 02 (3.40)
Kamon and Furakawa (1986) used the same approach but considered the
relationship unproven. Further, they noted that the elastic modulus of
annealed samples was higher than for those that had been quenched, an
effect that will be discussed later (see section 3.5.5). However, on annealing
there is an increase in the ambient temperature density of an epoxy resin.
Thus there will be an increase in the intermolecular forces and so also an
increase in the elastic modulus.
A systematic examination of the relationship between the structure and
elastic moduli of cured epoxy resins has been reported by Morel and
coworkers (1989). They compared experimental determinations of elastic
moduli for a range of resin-hardener systems with several methods of
estimating the limiting glassy moduli, such as those discussed by van
Krevelen and Hoftyzer (1976). Ultrasonic pulse measurements of both the
longitudinal transverse wave velocities enable elastic constants to be
calculated, from the relations
VL = ~. 3(1 - v) (3.41)
Q (1 + v)
(3.42)
where VL and VT are the velocities ofthe longitudinal and transverse waves,
and Q is the density. Thus, the shear modulus can be calculated from
equation 3.42 and also using equation 3.39 with equation 3.41 Young's
modulus can be calculated.
Poisson's ratio may be estimated from its relationship with the
longitudinal wave velocity by
VL = [ V ] 6 [3(1 -v)] 112

(3.43)
V (1- v)
where V is the sum of molar group contributions, Vi' i.e. V = LVi and V is
the molar volume. The ViS are known as Rao additivity increments and are
listed by Morel et al. (1989) for their calculations for epoxy resins. A more
complete listing is given by van Krevelen and Hoftyzer (1976). Use of this set
of molar contributions Ui underestimates the longitudinal wave velocity and
hence would lead to an underestimate of the elastic moduli.
Morel et al. (1989) also evaluated the applicability of equation 3.38 by
plotting the bulk modulus B vs 0 2 , the cohesive energy density. A straight
line plot was obtained but with a slope of 11.5 compared with the predicted
value of 8.04. They considered that their estimates of the solubility
parameters neglected the hydrogen bonding that is well known to occur in
epoxy resins. To estimate the degree of hydrogen bonding they measured
the equilibrium water absorption, Wm , at 100°C and 95% relative humidity.
There was a good linear fit between the experimental data and the water
absorption and elastic modulus, equation 3.44.
M = Mo + kWm (3.44)
where M is the modulus and the slope, k, depending on the contribution due
to hydrogen bonding, and Mo depending on the aromatic group
concentration. Good linear correlations were obtained for all three moduli,
B,EandG.
The moduli that are usually measured in the temperature range 50 to
150°C below the glass transition temperature of the resin are often con-
siderably lower than the limiting low temperature high frequency moduli
which have been discussed in this section. The stress-strain behaviour of
epoxy resins will be considered next.
3.5.3 Stress-strain curves and visco-elastic behaviour

Stress-strain curves for epoxy resins are often non-linear, as illustrated
schematically in Figure 3.8 for extension (Pink and Campbell, 1974). It is
usual to determine Young's modulus from the limiting slope at low applied
stresses and hence strains. This limiting slope increases slightly with
increased rate of extension (Yamini and Young, 1980) and decreases with
increase in temperature. Similar behaviour was also observed with
conventional compressive tests. Yamini and Young (1980) also studied the
effect of changes in the concentration of hardener and found that the tensile
modulus decreased as the concentration of hardener increased (Shell 828
with TETA, 7.4 to 14.7 phr). The cure time and temperature obviously also
affected the modulus. The change in relative modulus with Tg , as a measure
of the state of cure, is illustrated in Figure 3.9, which shows that the modulus
increases rapidly in the region if vitrification reaches a maximum and then
decreases (Wang and Gillham, 1992). This maximum is not related to
gelation, which has been suggested by some authors. Gelation was discussed
previously (see section 3.3.4) and it was noted that there is a change in the
slope of log E" or log Gil vs cure time, te , at the gel point. However, it is
8 306
367
''~"" 5
(jj
4 381
x
3
2 I--- 4 °/o---t
i = 3.3 x 10- 4 sec'
Strain (%)
Figure 3.8 Stress-strain curves for an epoxy resin at temperatures in the range 77 to 381 K.
Stress, kp/mm2; strains, displaced for clarity, see gauge length (from Pink and Campbell, 1974).
pertinent to note that when the structure of a cured epoxy resin is altered by
changing the concentration of hardener the Young's modulus has a
minimum when there is an exact stoichiometric mixture of epoxy groups and
reactive amine hydrogen atoms (Bell, 1970b; Selby and Miller, 1975; Kim
et al., 1978; Gupta et aI., 1985). It may be that the modulus is related to the
density of the resin at the measurement temperature, which also passes
through a maximum with cure time and hence extent of reaction (Bell et al. ,
1992). The maximum in the relative modulus is found at all cure tempera-
tures studied by Wang and Gillham (1992) with the relative modulus
increasing with cure temperature and the maximum occurring at lower
extents of reaction as the cure temperature increased (Figure 3.9).
From Figure 3.8 it can be seen that the curvature of the extensional stress-
strain curves is more pronounced at higher test temperatures. At room
temperature it has been found that for stresses lower that the yield stress the
stress-strain curve fits a power law relationship a = MEn with a value of
n = 0.93 for a BADGE type resin cured with DDM (Bell et al. 1992).
Williams (1979) suggests the use of the power law equation for the stress-
strain relationship for polymeric materials which do not obey Hooke's law
exactly with n = 0.9. When strains are low, M is approximately equal to
Young's for values of n not too different from unity.
theoretical gelation
end of glass
transnion
20
o 20 40 60 80 100 120 140 160 180 Tg

o 0.5 1 C*
Extent of Cure (as measured by Tg, 0C)
Figure 3.9 Relative isothermal elastic modulus versus Tg (extent of cure). Resin: DGEBA type
(Dow, DER332); hardener: Trimethylene glycol di-p-aminobenzoate (TMAB, Polaroid
Corp.) in a stoichiometric concentration. Measurement temperature (0) 30°C; (.) 50°C; (D)
70°C; (e) 90°C; (.6.) 110°C (from Wang and Gillham, 1992). * C is the cure parameter as
defined in Figure 3.7.
Although yield is observed in extension (Figure 3.8) especially at higher

measurement temperatures, it is often more precisely measured in uniaxial
compression (Moehlenpah et al., 1969). Yamini and Young (1980) have
studied the yield of a TETA-cured epoxy resin measuring yield stress as
functions of resin/hardener concentration, post-cure and rate of straining.
The most marked effects were due to systematic changes in the resin-
hardener stoichiometry and the test temperature. The yield stress decreased
as expected as the test temperature increased up to the glass transition
temperature of the resin, which increased with increased hardener
concentration for resin given the same cure and post-cure. They also found
that both the yield theories of Argon (1973) or Bowden (1973) were
satisfactory, but that Bowden's applied over a wider temperature range up
to the glass transition temperature of the resin.
Above their glass transition, epoxy networks behave as elastomers with
elastic moduli, En very much lower than the glass moduli. Banks and Ellis
(1982), compared f1E = E/E g for several cross-linked thermosetting resins
and found that for epoxy resin, f1E was in the range 7 X 10- 3 to 1.3 X 10- 2,
whereas for a very highly cross-linked phenol-formaldehyde resin f1E was
only 1.45 x 10- 1. Bell et al. (1992) have found that this ratio, f1E, changes
monotonically with epoxy conversion from 1.05 x 10- 2 to 1.16 X 10- 2 for
a BADGE type resin cured with DDM, and is not much higher even after an
extensive post cure (6 h) at 180°C.
The theory of rubber-like elasticity has been treated in considerable detail
(Treloar, 1975), which it would be inappropriate to discuss here. The
rubber-like behaviour of epoxy networks above their glass transition
temperature have been discussed by Oleinik (1986) and Ilavsky et al. (1984).
For an ideal rubbery material the moduli are related to the network
structure by
E = 3G = 3QRT (3.45)
r r M
e
where Q is the density, R is the gas constant, T is the absolute temperature

and Me is the average molecular weight of the network chains. From
consideration of the ratio of the chain contour length to the root mean
square distance between junction points Kaelble et al. (1988) discuss the
calculation of the maximum extension ratio, A = 111o, for epoxy resins and
other network polymers. For epoxy resins it is 1.4 to 1.6 compared with 6.85
for a silicone rubber. Oleinik (1986) also concludes that the largest possible
elongation of a cross-linked epoxy resin is about 50% (i.e. E = 0.5 = A - 1,
and A = 1.5). The highest measured elongations at break is somewhat
lower, 35% (range 12-35%). Also, the stress-strain curve can be repre-
sented by
(3.46)
for extension ratios of up to 1.2, where a is the applied stress, equal to the
applied force divided by the un strained area of the sample. Thus, from this
experimental value of Gr it is possible to calculate the equivalent molecular
weight of the network chains. This is a direct method of determining a
parameter which is a characteristic of the network structure. Other methods
involve branching theory and either assumptions or direct determination of
the relative reactivities of the reactive groups, such as the reactivity of
primary and secondary amine groups.
3.5.4 Visco-elastic properties

The rate and temperature effects on the stress-strain curves, Figure 3.8, is a
manifestation of the visco-elastic properties of epoxy resins. The change in
the dynamic mechanical properties with cure were briefly discussed in

section 3.3.4 (see Figure 3.5). As well as the glass transition temperature,
known as the a process, there are other relaxations, (3, y, etc. (Pogany,
1969, 1970; Williams, 1979; Ochi et al., 1985). These processes may be
observed by measurement of dynamical mechanical properties as a function
of temperature, and also creep or stress relaxation rates. These methods
have been reviewed by Neilson (1969) and Kaelble (1973; Kaelble et al.,
1988) discuss the visco-elastic properties of epoxy resins and Kong (1986)
extends the discussion to the properties of composites, as does Wolfe and
Tod (1989).
Arridge and Speake (1972) report on the effects of cure on the low
temperature y relaxation. Crowson and Arridge (1979) constructed a time-
temperature master curve from creep data, and determined the shift factors
aT. They found that the temperature dependence of the log aT could not be
represented by the WLF equation. A similar temperature dependence of the
shift factors, log aT, was found for master curves constructed from DMTA
measurements (Bell et al., 1992). Thus, further analysis of the temperature
dependence of visco-elastic data is required.
Gupta and coworkers (1984-1985) have studied an epoxy resin and found
as well as the a and (3 relaxation a third broad but not very intense relaxation
situated between the a and (3 peaks which they call an a' transition. Dillman
and Sefferis (1989) report their study in which they analysed both the cure
processes and the dynamic mechanical behaviour, and pointed out its
analogy with dielectric relaxation. Wang and Gillham (1992) have
monitored the changes in rigidity and log (decrement) with cure for an epoxy
resin. They located Tg and Tp from the maxima in the log (decrement) versus
temperature plots. These data were used for the construction of their Tg-TP
(CTP) diagram Figure 3.7. An interesting feature of their results was the
'jump' in the Tp vs Tg plot, Figure 3.10, which occurs prior to gelation. After
gelation Tp increases linearly with Tg •
3.5.5 Physical ageing

When a glass forming material is quenched from above its glass transition
temperature there is an 'instantaneous' contraction followed by a time-
dependent reduction in volume. The rate of the time-dependent densi-
fication depends on the temperature to which the glass has been quenched.
For temperatures only moderately below the glass transition temperature
the rate of volume relaxation is relatively rapid but becomes slower and
slower as the quench temperature is reduced. Such volume relaxations were
studied many years ago and Kovacs (1958, 1977) carried out systematic
studies on polymeric materials. Such volume relaxation processes can be
treated in a similar way to a visco-elastic analysis of creep behaviour.
With a decrease in volume the free volume will also be reduced and hence
Or-----------------------------------,
-40
- 80
-120
-160
theoretical gelation
-200~~~--~~r_--r_r_r---r---~--~--4
o 20 40 60 80 100 120 140 160 180 Tg
o 0.5 1 C*
Figure 3.10 The fJ relaxation temperature Tfi versus the glass transition temperature T g • The
resin-hardener system is the same as in Figure 3.9 (from Wang and Gillham, 1992). * C is the
cure parameter as defined in Figure 3.7.
chain mobility which depends on molecular packing and is at least partially

dependent on free volume will also be reduced. Thus, as the volume
decreases so will molecular mobility and hence the rate of volume
relaxation. Unfortunately it is still not possible to define free volume
uniquely or unambiguously and hence these effects can only be explained
qualitatively in terms of a change in free volume. However, the concept is
very convenient as an aid to an understanding of the processes involved.
The properties of a polymeric glass will depend on its packing density
which affects the distances between molecular sub-groups and hence the van
der Waals forces between molecules. Generally, when a polymeric material
is cooled through its glass transition temperature it will become stiffer and
more brittle. Also, its visco-elastic properties change, the creep and stress
relaxation rates decrease, as does the damping or tan O. These properties
also change with densification of the glass and these effects have been
attributed to 'physical ageing', to distinguish the process from 'chemical
ageing' such as photo-chemical or thermal degradation which are due to the
breaking or formation of covalent chemical bonds. These chemical
processes lead to permanent changes in the structure of the material. It is
generally regarded that physical ageing effects are reversible and by
increasing the temperature of the sample above its glass transition
temperature its past history can be expunged.
Studies of physical ageing have been the subject of a monograph by Struik
(1978) who observed that partially aged samples with identical thermal
histories subjected to large deformations relaxed more rapidly than when

subject to only small deformations. With the application of a tensile stress
there will be an increase in volume of the sample. The magnitude of this
increase in volume will depend on the value of Poisson's ratio for the
material at the ageing temperature. Thus, with this increase in volume the
rates of molecular relaxation processes will be more rapid. The term that has
been used for this effect is 'erasure of ageing' but more recently
'rejuvenation' appears to be accepted.
Most of the initial studies of physical ageing were on thermoplastics, so
called 'linear' macromolecules and it is only more recently that more
attention has been directed to such processes in network polymers. A major
problem with the study of cured resins is control of the network structure,
since it is possible for either or both continued cure or degradative processes
to obscure the physical ageing effects. A useful review and report of his own
work on epoxy resins is given by Kong (1986). He reports that the lower the
cooling rates from T> Tg to TA < Tg, where TA is the ageing temperature,
and/or with increased ageing time, the sample increases in density and this
leads to lower impact strength, fracture energy, fracture strain, mechanical
damping and lower creep and stress relaxation rates. With physical ageing
there is a loss of ductility which is important in epoxy resin composites which
are also discussed by Kim (Kim et at., 1978).
McKenna and co-workers (see Lee and McKenna, 1988, 1990) have
published a lot of detailed research work on the physical ageing of epoxy
resins. In two recent papers (Santore et al. , 1991; McKenna et al., 1991) they
question the possibility of rejuvenation due to the applications of
mechanical stresses. Specifically they find that the underlying volume
recovery kinetics, which are much slower than the mechanical relaxation,
remain unaltered in spite of the imposition of mechanical deformations.
Thus, they report that their results support the argument that mechanical
stimuli neither alter the underlying (non-equilibrium) thermodynamic state
of the glass nor erase physical ageing.
Study of both curing and physical ageing processes have been carried out
by Plazek and co-workers (Choy and Plazek, 1986; Plazek and Choy, 1989;
Plazek and Frund, 1990). They have measured the change in volume ofthe
resin during cure using a specially constructed pressurized volume
dilatometer, and a frictionless magnetic-bearing torsion creep apparatus to
measure creep and recovery in order to calculate viscosities and recoverable
steady-state compliances of the cured resins. In their later study (Bero and
Plazek, 1991) of volume-dependent rate processes in epoxy resins they
report on four types of interrelated experiments, (i) specific volume with
constant rates of cooling, (ii) time-temperature volume measurements, (iii)
elongational creep compliances and (iv) shear creep compliances. A very
interesting result they report is that the volume-temperature cooling cures
are asymptotic to their glass lines irrespective of the rate of cooling. Thus,
the lower temperature limit of the glass transition region is independent of

the cooling rate. Hence, the molecular motions that occur in the glass
transition region are inhibited at a specific temperature irrespective of the
actual volume of the specimen. This observation is of considerable
importance theoretically and practically and requires further study, as
planned by Bero and Plazek (1991).
References
Acitelli, M.A., Prime, R.B. and Sacher, E. (1971) Polymer 12, 335-343.
Adolf, D. and Martin, J.E. (1990) Macromolecules 23,3700-3704.
Adolf, D., Martin, J.E. and Wilcoxon, J.P. (1990) Macromolecules 23,527-531.
Amdouni, N., Sautereau, H., Gerard, J.-F. and Pascault, J.-P. (1990) Polymer 31,1245-1253.
See Fig. 2 and Table 2, p. 1348.
Argon, A.S. (1973) Phil. Mag. 28,839.
Aronhime, A.T. and Gillham, J.K. (1986) Adv. in Polym. Sci. 78,83-113.
Arridge, R.G.C. and Speake, J.H. (1972) Polymer 13, 443-454.
Banks, L. and Ellis, B. (1982) Polymer 23,1466-1472.
Barton, A.L.M. (1983) Handbook of Solubility Parameters and other Cohesion Parameters.
CRC Press, Boca Raton, USA.
Barton, J.M. (1985) Adv. in Polym. Sci. 72, 111-154.
Bell, J.P. (1970a) I. Polym. Sci. A2. 6,417-436.
Bell, J.P. (1970b) I. Appl. Polym. Sci. 32, 1901-1906.
Bell, J., Ellis, B. and Found, M.S. (1992) Unpublished experimental measurements.
Benson, S.W. (1960) The Foundations of Chemical Kinetics. McGraw-Hili, New York,
pp.42-46.
Bero, c.A. and Plazek, D.T. (1991) I. Polym. Sci., B. Polym. Phys. 29, 39-47.
Bidstrup, W.W. and Senturia, S.D. (1989) Polym. Eng. Sci. 29, 290-294.
Billmeyer, F.W. (1984) Textbook of Polymer Science 3rd edn, Wiley, New York, pp. 40-47.
Bondi, A. (1968) Physical Properties of Molecular Crystals, Liquids and Glasses, Section
13.3.1, Wiley, New York, pp. 397-400.
Bowden, P.B. (1973) In The Physics of Glassy Polymers Ch. 5, ed. Haward, R.N. Applied
Science Publishers, London, pp. 279-339.
Camargo, R.E., Gonzalez, V.M. and Macosko, C.W. (1983) Rubber Chemistry and
Technology 56, 774--778.
Carothers, W.H. (1936) Trans. Far. Soc. 32,39-53.
Case, L.C. (1957) I. Polym. Sci. 26,333-350.
Chen, X.M. and Ellis, B. (1992a) Paper submitted to I. Appl. Polym. Sci.
Chen, X.M. and Ellis, B. (1992b) Unpublished data, but see Chen, X.M. Ph.D thesis Sheffield
1988).
Charlesworth, J. (1980) I. Polym. Sci. Polym. Chern. Ed. 18,621-628.
Chiu, W.Y., Carratt, G.M. and Soong, D.S. (1983) see Figure 2 Macromolecules 16, 348-357.
Choy, I.C. and Plazek, D.J. (1986) I. Polym. Sci., B. Polym. Phys. 24, 1303-1320.
Crowson, R.J. and Arridge, R.G.C. (1979) Polymer 20,737-746.
Dannenberg, H. (1963) S.P.E. Trans. 3,78-88.
de Gennes, P.-G. (1979) Scaling Concepts in Polymer Physics, Ch. V, Cornell University Press,
Ithaca, pp. 128-162.
Dillman, S.H. and Sefferis, J.C. (1989) I. Macromol. Sci.-Chem. A26(1), 227-247.
Dobas, I., Eichler, J. and Klaban, J. (1975) Collection Czech. Chern. Comm. 40,2989-2997.
Durand, D. and Bruneau, C.-M. (1983) Polymer 24,287-591 and 592-595.
Dusek, K. (1985) Brit. Polym. I. 17,185-189.
Dusek, K. (1986) Adv. in Polym. Sci. 75, 1-59. See pp. 9-12 for discussion of the morphology
of cured epoxy resins.
Dusek, K., IIIavsky, M. and LuMk, S.J. (1975) I. Polym. Sci. Polym. Symp. 53,29.
Elias, H.-G. (1976) Macromolecules Vol. 1 and 2. Plenum Press, New York, pp. 606-614.
Elliott, S.R. (1990) Physics of Amorphous Materials 2nd edn, Ch. 2, Longman Scientific and
Technical, Harlow, pp. 29-69.
Ellis, B., Bell, J., Chen, X.M. and Found, M.S. (1992) Paper in preparation.
Enns, J.B. and Gillham, J.K. (1983) 1. Appl. Polym. Sci. 2S, 2567-2591.
Entelis, S.G. and Tiger, R.P. (1976) Reaction Kinetics in the Liquid Phase, Ch. 1. Translated
from Russian, John Wiley, Chichester, pp. 1-25.
Eyring, H., Lin, S.H. and Lin, S.M. (1980) Basic Chemical Kinetics, Ch. 9, John Wiley, New
York, pp. 376-451.
Fava, R.A. (1968) Polymer 9,137-151.
Flory, P.J. (1941) 1. Amer. Chern. Soc. 63,8083-3100.
Flory, P.J. (1953) Principles of Polymer Chemistry, Ch. IX Cornell University Press, Ithaca,
pp. 347-398.
Flynn, J.H. (1989) Ene. Polym. Sci. and Eng. 2nd edn Suppl. pp. 715-723.
Flynn, J .H. and Wall, L.A. (1966) 1. Res. Natl. Bur. Stand. 70A, 487-523.
Frost, A.A. and Pearson, R.G. (1961) Kinetics and Mechanism 2nd edn, John Wiley, New
York, pp. 178-184.
Ghaemy, M., Billingham, N.C. and Calvert, P.D. (1982) 1. Polym. Sci. Polym. Lett. 20,
439-443.
Gilbert, D.G., Ashby, M.F. and Beaumont, P.W.R. (1986) 1. Mat. Sci. 21,3194-3210.
Goddu, R.F. and Delker, D.A. (1958) Anal. Chern. 30,2013-2016.
Gordon, M. and Ross-Murphy, S.B. (1975) Pure and Applied Chern. 43, 1-26.
Grillet, A.C., Galy, J., Pascault, J.P. and Bardin, 1. (1989) Polymer 30,2094-2103.
Grimmett, G. (1989) Percolation Ch. 1. Springer-Verlag, New York, pp. 1-24.
Guillett, J. (1985) Polymer Photophysics and Photochemistry see Ch. 3, pp. 52-70 and Ch. 6,
pp. 114-135 especially p. 132. Cambridge University Press, Cambridge.
Gupta, V.B., Drzal, L.T., Lee, c.y-c. and Rich, M.J. (1984-1985) 1. Macromol. Sci. B23,
435-466.
Hadad, D.K. (1988a) In Epoxy Resins, Chemistry and Technology 2nd edn. ed. May, C.A.
Hadad, D.K. (1988b) In Epoxy Resins, Chemistry and Technology, 2nd edn. ed. May, C.A.
Harran, D. and Laudourd, A. (1986) 1. Appl. Polym. Sci. 32,6043-6062.
Havlicek, 1. and Dusek, K. (1987) In Crosslinked Epoxies ed. Sedlacek, B. and Kahovec, J.
Walter de Gruyter, New York, pp. 417-424.
Horie, H., Hiura, H., Sawada, M., Mita, 1. and Kambe, H. (1970) 1. Polym. Sci. A1.S, 1357-
1372.
Hsich, H.-Y. (1982) 1. Appl. Polym. Sci. 27,3265-3277.
Huguenin, F.G.A.E. and Klein, M.T. (1985) Ind. Eng. Chern. Prod. Res. Der. 24, 166-171.
Ilavsky, M., Bogdanova, L.M. and Dusek, K. (1984) 1. Polym. Sci. Polym. Phys. ed. 22,265.
Jha, A. and Parker, J.M. (1989) Physics and Chemistry of Glasses 30, 220-228.
Kaelble, D.H. (1973) In Epoxy Resins, Chemistry and Technology, Ch. 5. eds. May, C.A. and
Tanaka, Y. Dekker, New York, pp. 327-37l.
Kaelble, D.H., Moacanin, J. and Gupta, A. (1988) Epoxy Resins, Chemistry and Technology
2nd edn, Ch. 6. ed. May, C.A. Dekker, New York, pp. 603-652.
Kamal, M.R. and Sourour, S. (1972) Society of Plastics Engineers IS, 93-98.
Kamal, M.R. and Sourour, S. (1973) Polym. Sci. and Eng. 13,59-64.
Kamon, T. and Furakawa, H. (1986) Adv. in Polym. Sci. SO, 173-202.
Keenan, M.R. (1987) 1. Appl. Polym. Sci. 33, 1725-1734.
Khabenko, A.V. and Dolmatov, S.A. (1990) 1. Thermal. Anal. 36,45-54.
Kim, S.L., Skibo, M.D., Manson, J.A., Hertzberg, R.W. and Janiszewski, J. (1978) Polym.
Eng. Sci. IS, 1093-1100.
Kong, E.S.W. (1986) Adv. in Polym. Sci. SO, 126-171.
Kovacs, A.J. (1958) 1. Polym. Sci. 30, 131-147.
Kovacs, A.J., Hutchinson, J.M. and Aklonis, J.J. (1977) The Structure of Non-Crystalline
Materials ed. Gaskell, P.H. Taylor and Francis, London, pp. 153-163.
Labana, S.S. (1986). Ene. Polym. Sci. and Eng. 2nd edn, Vol. 4, John Wiley, New York,
pp. 350-360.
Larson, R.G. (1988) Constitutive Equations for Polymer Melts and Solutions, Ch. 4.
Butterworth, Boston, pp. 93-127.
Lee, A. and McKenna, G.B. (1988) Polymer 29,1812-1817.
Lee, A. and McKenna, G.B. (1990) Polymer 31,423-430.
Lee, H. and Neville, K. (1976) Handbook of Epoxy Resins Appendix 4-1. McGraw Hill, New
York, pp. 4-35-4-56.
Lunak, S., Vladyka, J. and Dusek, K. (1978) Polymer 19, 931-933.
Macosko, e.W. and Miller, D.R. (1976) Macromolecules 9,199-206.
Macosko, C.W. (1985) Brit. Polymer J. 17,239-245.
Martin, J.E., Adolf, D. and Wilcoxon, J.P. (1988) Phys. Rev. Lett. 61,2620-2623.
Malkin, A.Yu and Kulichikhin, S.G. (1991) Adv. in Polym. Sci. 101,217-257.
Matsuoku, S., Quan, X., Bair, H.E. and Boyle, D.S. (1989) Macromolecules 22, 4093-4098.
McKenna, G.B., Santore, M.M., Lee, A. and Durran, R.S. (1991) J. Non-Crystalline Solids
131-133,497-504.
Mertzel, E. and Koenig, J.L. (1986) Adv. in Polym. Sci. 75,73-112. For discussion ofNMR see
pp. 92-112 and FT-IR see pp. 73-92.
Mestan, S.A. and Morris, C.E.M. (1984) J. Macromol. Sci. Rev. in Macromol. Chem. and
Phys. C24, 117-172.
Mijovic, J. (1990) J. Appl. Polym. Sci. 40,845-866.
Mijovic, J., Kim, J. and Slaby, J. (1984) J. Appl. Polym. Sci. 29, 1449-1462.
Mijovic, J. and Lee, C.H. (1989) J. Appl. Polym. Sci. 38,2155-2170.
Mijovic, J. and Wang, H.T. (1989) J. Appl. Polym. Sci. 37,2661-2673.
Mita, I. and Horie, K. (1987)J. Macromol. Sci. Rev. Macromol. Chem. andPhys. C27, 91-169.
Moehlenpah, A.E., Ishai, O. and Di Benedetto, A. (1969)J. Appl. Polym. Sci. 13,1231-1245.
Morel, E., Bellenger, V., Bocquet, M. and Verdu, J. (1989)J. Mat. Sci. 24,69-75.
North, A. M. (1964) The Collision Theory of Chemical Reactions in Liquids. Methuen, London.
Nielsen, L.E. (1969) J. Macromol. Sci., Rev. Macromol. Chem. C3(1), 69-103.
Ochi, M., Iesako, H. and Shimbo, M. (1985) Polymer 26,457-461.
Oleinik, E.F. (1986) Adv. in Polym. Sci. SO, 50-99.
Peyser, P. and Bascom, W.D. (1977) J. Appl. Polym. Sci. 21,2359-2373.
Pingsheng, H.E. and Chune, Li (1989) J. Mat. Sci. 24,2951-2956.
Pink, E. and Campbell, J.D. (1974) Mat. Sci. Eng. 15, 187-194.
Pinner, S.H. (1956)1. Polym. Sci. 21, 153-157.
Plazek, D.J. and Choy, I.e. (1989) J. Polym. Sci., B. Polym. Phys. 27,307-324.
Plazek, D.J. and Frund, Z.N. (1990) J. Polym. Sci., B. Polym. Phys. 28,431-448.
Pogany, G.A. (1969) J. Mat. Sci. 4,405-409.
Pogany, G.A. (1970) Polymer 11,66-78.
Prime, R.B. (1981) In Thermal Characterization of Polymeric Materials Ch. 5. ed. Turi, E.
pp. 435-569 see p. 445ff.
Rabek, J .F. (1980a) Experimental Methods in Polymer Science John Wiley, Chichester, Ch. 32,
pp. 527-542. (1980b) Ch. 20, pp. 299-331. (1980c) Ch. 15, pp. 221-253. (1980d) Ch. 34,
pp. 549-581.
Rabinowich, E. (1937) Trans. Far. Soc. 33, 1225-1233.
Rashid, H. U. (1978) Ph.D Thesis, University of Sheffield.
Rekhson, S.M. (1986) In Glass: Science and Technology, Vol. 3, Ch. 1, ed. Uhlmann, D.R. and
Kreidl, N.J. Academic Press, Inc, Orlando, pp. 1-117.
Riccardi, C.C., Adahbo, H.E. and Williams, R.J.J. (1984)J. Appl. Polym. Sci. 29,2481-2492.
Riccardi, e.e. and Williams, R.J.J. (1986a) J. Appl. Polym. Sci. 32,3445-3456.
Riccardi, C.C. and Williams, R.J.J. (1986b) Polymer 27,913-920.
Rogers, M.G. (1972) J. Appl. Polym. Sci. 16,1953-1958.
Rozenberg, B.A. (1986) Adv. in Polym. Sci. 75, 113-165.
Sandford, W.M. and McCullough, R.L. (1990) J. Polym. Sci. (Pt. B) Polym. Phys. 2S,
973-1000.
Santore, M.M., Duran, R.S. and McKenna, G.B. (1991) Polymer 31,2377-2381.
Scherer, G.W. (1986) Relaxation in Glass and Composites, Ch. 12, John Wiley & Sons, New
York, pp. 162-174.
Selby, K. and Miller, L.E. (1975) J. Mat. Sci. 10,12-24.
Senturia, S.D. and Sheppard, N.F. (1986) Adv. in Polym. Sci. SO, 1-47.
Serrano, D., Peyrelasse, J., Borred, C., Harran, D. and Monge, P. (1989) Macromol. Chem.
Macromol. Symp. 25,55--61.
Serrano, D., Peyrelasse, J., Borred, e., Harran, D. and Monge, P. (1990)J. Appl. Polym. Sci.
39, 679--693.
Sharma, D.K. and Soane, D.S. (1988) Macromolecules 21,700-710.
Shen, M.e. and Eisenberg (1966) Prog. Solid State Chem. 3, 407-481. See also Rubber
Chemistry and Technology 43,95-170.
Sheppard, N.F. and Senturia, S.D. (1986) Polym. Sci. and Eng. 26,354-357.
Smith, T.T. (1961) Polymer 2,95-108.
Sojka, S.A. and Moniz, W.B. (1976) J. Appl. Polym. Sci. 20, 1977-1982.
Sourour, S. and Kamal, M.R. (1976) Thermochimica Acta 14, 41-59.
Sperling, L.H. (1986) Introduction to Physical Polymer Science, Ch. 6, John Wiley, New York,
pp. 224-300.
Stanley, H.E. (1986) Enc. Polym. Sci. and Eng. 2nd edn, Vol 4, pp. 323-349.
Stauffer, D. (1985) Introduction to Percolation Theory Taylor and Francis, London (A 2nd edn
is scheduled for 1992).
Stauffer, D., Coniglio, A. and Adam, M. (1982) Adv. in Polym. Sci. 44, 103-158.
Steinmann, B. (1987) In Cross-Linked Epoxies eds Sedlacek, B. and Kahovec, J. Walter de
Gruyter & Co., Berlin, pp. 117-130.
Steinmann, B. (1989) J. Appl. Polym. Sci. 37, 1753-1776.
Stevens, G.C. and Richardson, M.J. (1983) Polymer 24,851-858.
Struik, L.e.E. (1978) Physical Aging in Amorphous and Other Materials. Elsevier,
Amsterdam.
Stutz, H., Illers, K.-H. and Mertes, J. (1990) J. Polym. Sci. (Pt. B) Polym. Phys. 28, 1483-
1498.
Stutz, H. (1992) Private communication.
Tanaka, Y. and Kakiushi, H. (1965) J. Polym. Sci. 3A, 3279-3299.
Tanaka, Y. and Mika, T.F. (1973) In Epoxy Resins, Chemistry and Technology Ch. 3, ed. May,
C.A. and Tanaka, Y. section IV, pp. 212-238.
Tobolsky, A.V. (1960) Properties and Structure of Polymer Wiley, New York, pp. 1-12.
Treloar, L.R.G. (1975) The Physics of Rubber Elasticity, 3rd edn, Oxford University Press,
London.
van Krevelen, D. W. and Hoftyzer, P.J. (1976) In Properties of Polymers, their estimation and
correlation with chemical structure, 2nd edn, Ch. 3, Elsevier, Amsterdam (see also van
Krevelen, D.W. (1990), 3rd edn).
Verchere, D., Sautereau, H., Pascauit, J.P., Riccardi, C.C., Moschier, S.M. and Williams,
R.J.J. (1990) J. Appl. Polym. Sci. 38,725-731.
Wang, X. and Gillham, 1.K. (1992) J. Coatings Tech. 64,37-45.
Ward, I.M. (1983) Mechanical Properties of Solid Polymers 2nd edn, Ch. 8. John Wiley,
Chichester, pp. 166--174.
Whorlow, R.W. (1980) Rheological Techniques Ellis Horwood Ltd, Chichester, pp. 123-191,
Appendix AI, p. 422.
Williams, G. and Watts, D.C. (1970) Trans. Far. Soc. 66,80-85.
Williams, J.G. (1979) J. Appl. Polym. Sci. 23,3433-3444.
Williams, J.G. (1980) Stress Analysis of Polymers 2nd edn. Ellis Horwood, Chichester,
pp. 142-143.
Wisanrakkit, G. and Gillham, J.K. (1990)J. Appl. Polym. Sci. 41,2885-2929.
Wisanrakkit, G., Gillham, J.K. and Enns, J.B. (1990) J. Appl. Polym. Sci. 41, 1895-1912.
Wolfe, S.V. and Tod, D.A. (1989) J. Macromol. Sci. Chem. A26(1), 249-272.
Yamini, S. and Young, R.J. (1980) J. Mat. Sci. 15,1814-1822.
Yu, W. and von Meerwall, D. (1990) Macromolecules 23,882-889.
Ziman, J.M. (1979) Models of Disorder Ch. 7, Cambridge University Press, Cambridge,
pp.246--285.
4 Additives and modifiers for epoxy resins*
S.J. SHAW
4.1 Introduction
In addition to the two main ingredients of an epoxy formulation, i.e. resin

and curing agent, numerous other formulatory materials are available and
have frequently been employed to modify the properties and characteristics
of epoxies, both uncured and in their cured form. In this chapter each of the
main types of modifying material will be discussed with examples given of
the most common types that have been employed together with the potential
benefits they impart.
Although this discussion will be approached via a series of convenient
headings it is important to recognise that many ingredients employed in
epoxy formulations could legitimately be discussed under more than one
heading. Thus, for example, some of the materials described in this text as
flexibilisers could also be considered as reactive diluents and vice versa. The
author apologises for any confusion this may cause.
4.2 Diluents
Diluents are usually employed as a means of reducing viscosity in order to

aid general processability as well as allowing for greater incorporation of
other formulatory ingredients such as fillers and as a means of improving
characteristics such as wetting and the incorporation of resins into various
fibrous reinforcements.
In addition to viscosity, it is important to recognise that other properties
will also be modified, the properties in question and the magnitude of
modification being dependent on both the type and quantity of diluent
employed. In addition many of the materials capable of exerting diluent
effects possess adverse toxicological characteristics, in many cases being
more severe than those found with both resin and curative. Great care
should therefore be taken in both selection and use.
Diluents that have been considered for use in epoxy resin technology can
be divided into two broad classes comprising those which can be considered
as either non-reactive or reactive.
* © British Crown Copyright 19921DRA. Published with the permission of the Controller of
Her Majesty's Stationery Office.
4.2.1 Non-reactive diluents

A number of materials which can be regarded as non-reactive have been
considered and used to provide viscosity reduction in epoxies. These include
aromatic hydrocarbons such as toluene or xylene which are capable of
reducing viscosity in a simple diglycidylether of bisphenol A (DGEBA)
resin system sufficient to permit a room temperature viscosity approxi-
mately 20% of its initial value with a diluent loading of about 5%. When
employed at this level, studies have shown that the level of impact on
physical and mechanical properties are not usually dramatic, indeed in some
cases the presence of the diluent has been shown to increase the extent of
cure in the formulation, presumably through the reduction in viscosity
providing a reaction medium more conducive to molecular mobility and thus
reactivity. At higher concentrations however, the properties of the cured
epoxy can deteriorate markedly and, due to their non-reactivity, diluents of
this type can lead to the presence of voids in the cured polymer, primarily if
the formulation is subjected to elevated temperature cure sufficient to allow
vaporisation of the diluent.
Other non-reactive materials which have been considered as potential
diluents include dibutyl phthalate, styrene and various phenolic
compounds.
In addition to the drawbacks mentioned above, since the compounds are
not chemically bound into the cross-linked network, cured formulations
containing these materials usually exhibit inferior chemical resistance.
Consequently, attempts at viscosity reduction with epoxies have generally
made use of materials capable of chemical reaction with the epoxy resin,
these generally being referred to as reactive diluents.
4.2.2 Reactive diluents

Two major classes of reactive diluent can and have been utilised
commercially, these being either epoxy-containing diluents or compounds
which derive their reactivity from functional groups other than epoxide.
4.2.2.1 Epoxy-based reactive diluents Two types of diluent can be

considered under this classification, i.e. (i) mono-epoxy compounds and
(ii) di/polyfunctional epoxies. With both classes the presence ofthe epoxide
group(s) allows the diluent to participate, with the resin and curing agent, in
the polymerisation and cross-linking reaction thus permitting the diluent to
become chemically bound into the cross-linked network. Thus, to a degree,
the use of reactive diluents can alleviate some of the debilitating effects
found with the non-reactive compounds.
Figure 4.1 shows some mono epoxy reactive diluents which have been
considered and indeed in some cases used to modify epoxy resin viscosity.
ADDITIVES AND MODIFIERS 119
Octylene oxide p-Butyl phenol glycidyl ether
Butyl glycidyl ether Cresyl glycidyl ether
o
(§:r'CH-cn, 1\
Styrene oxide Allyl glycidyl ether
Phenyl glycidyl ether Glycidyl ester of tert-carboxylic acid
Figure 4.1 Monoepoxy reactive diluents.
Although capable of exerting a substantial viscosity-reducing effect, these

materials essentially reduce the functionality of the reaction system leading
to a decrease in the cross-link density of the cured formulation. Conse-
quently important physical and mechanical properties are frequently
affected, particularly the ability to exert a 'high temperature capability'.
The extent to which both cured properties and initial viscosity are reduced
will be a function of both diluent type and concentration. Thus choice of
both parameters will clearly be crucial in order to derive the greatest
reduction of viscosity with minimal cost to other properties.
In an attempt to preserve physical/mechanical properties at elevated

temperatures, the use of polyfunctional epoxy diluents has been considered.
Although to a degree they can be regarded as epoxy resins in their own right,
their use in more conventional formulations, e.g. those employing a
DGEBA as the basic resin component, has demonstrated their ability to
provide viscosity reduction. Typical of the polyfunctional epoxy compounds
which have been suggested as reactive diluents are those shown in Figure
4.2. In most cases, since the funtionality of the system as a whole is not
greatly affected by the incorporation of these materials, cross-link density,
and therefore properties associated with it are not usually affected
detrimentally. Indeed some property enhancement has been shown to be
possible.
4.2.2.2 Non-epoxy based reactive diluents. Although epoxide-based

reactive diluents are employed in the majority of cases where viscosity
reduction is considered desirable, a number of compounds which derive
reactivity via non-epoxide groups have been considered. Most notable
amongst these have been triphenyl phosphite and lactone compounds such
as butyrolactone, the structures of which are shown in Figure 4.3. Both
materials have been shown to have effective viscosity-reducing
characteristics, although frequently at the expense of an 'elevated
temperature capability'.
4.3 Fillers
Apart from the resin and curative, fillers are possibly the most common
formulatory ingredient employed in the majority of epoxy formulations.
Literally hundreds of different filler types can and have been used to modify,
in one form or another, the properties and characteristics of epoxies, in
addition to reducing cost. Table 4.1 indicates in terms of advantages and
disadvantages, the various effects that filler incorporation can have.
Although fillers can be considered beneficial for many applications,
disadvantageous characteristics such as an increased density (and hence
weight) together with an increase in viscosity which is likely to influence the
processing behaviour of the formulation, obviously require serious
consideration.
Table 4.2 lists some of the more important particulate fillers which have
been employed in epoxy formulations. Although the list cannot be regarded
as exhaustive, it does indicate the range of characteristic modifications
which the various filler types shown can provide. It is of interest to briefly
consider some of these property modifications.
o
@(CH-CH, 1\
CH-CH2
"I
Butadiene dioxide Divinylbenzene dioxide
Diglycidyl ether Butanediol diglycidyl ether
Limonene dioxide Viny1cyclohexene dioxide
Diethylene glycol diglycidyl ether

Figure 4.2 Polyfunctional epoxy reactive diluents.
4.3.1 Physical/mechanical properties

The addition of particulate fillers generally results in reductions in strength
characteristics such as tensile and flexural strength. Fibrous fillers, however,
when employed at loading levels greater than a critical value, frequently
provide improvements in these strength parameters. As would be expected,
incorporation of fillers, particulate or fibrous, invariably produces
(©-+ 3
Triphenyl phosphite Butyrolactone

Figure 4.3 Non-epoxy based reactive diluents.
Table 4.1 Advantages and disadvantages of filler incorporation
Advantages Disadvantages
Reduced formulation cost Increased weight

Reduced shrinkage Increased viscosity
Improved toughness Machining difficulties
Improved abrasion resistance Increased dielectric constant
Reduced water absorption
Increased heat deflection temperature
Decreased exotherm
Increased thermal conductivity
Reduced thermal expansion coefficient
substantial increases in modulus, the magnitude being dependent on filler

type and loading level.
One specific application where fillers have been shown to impart
substantial strength improvements is with structural adhesive formulations
where the addition of fillers such as aluminium and alumina is common
practice (Lewis, 1988).
Fillers do not usually provide any significant enhancement of glass
transition temperature, Tg , or other measures of high temperature
distortion.
4.3.2 Thermal characteristics

Fillers allow the possibility of modifying various thermal characteristics of
epoxy resins including properties such as exothermic behaviour, thermal
conductivity and thermal expansion coefficient.
The exothermic characteristics of many epoxy formulations can give rise
to serious processing difficulties, particularly when the production of bulky
items is being considered. Since epoxies generally exhibit good thermal
insulating characteristics, extremely high temperatures are possible in the
middle of particularly large castings to the extent that component
volatilisation and even charring can result. The incorporation of fillers can
Table 4.2 Filler types and potential property modifications
Filler Property modification
Aluminium Machinability, impact resistance, thermal conductivity,

mechanical properties, dimensional stability
Alumina Abrasion resistance, electrical resistivity, dimensional
stability, toughness, thermal conductivity
Aluminium silicate Extender, pigmentation, dimensional stability, chemical
resistance
Aluminium trioxide Flame retardation
Arsenic pentoxide Thermal resistance
Barium sulphate Extender
Beryllium oxide Thermal conductivity
Calcium carbonate Extender, pigmentation, dimensional stability,
machinability, mechanical properties
Calcium sulphate Extender, dimensional stability
Calcium silicate Mechanical properties
Carbon black Reinforcement, pigmentation, thermal conductivity,
electrical conductivity, thermal resistance
Copper Electrical conductivity, thermal conductivity,
mechanical properties
Colloidal silica Thixotropy
Fibrous glass Impact strength
Graphite Lubricity, pigmentation, thermal conductivity, electrical
conductivity, abrasion resistance
Glass microballoons Density reduction
KaolincIay Extender
Lithium aluminium silicate Thermal expansion coefficient
Mica Electrical resistance, dielectric properties, chemical
resistance, toughness, moisture resistance, lubricity
Molybdenum disulphide Lubricity
Quartz Electrical properties, dimensional stability, extender
Sand Abrasion, thermal conductivity
Silica Abrasion resistance, electrical properties, extender,
dimensional stability, thermal conductivity, moisture
resistance
Silver Electrical conductivity, thermal conductivity
Titanium dioxide Pigmentation, dielectric properties, extender
Talc Extender
Zirconium silicate Arc resistance
substantially diminish exothermic heat generation by reducing the quantity

of resin in the formulation and increasing thermal conductivity so as to allow
more efficient removal of the exothermic heat. Indeed thermal conductivity
can be increased by a factor of approximately five with the addition of fillers
such as aluminium, alumina and copper. Generally an increase in filler
concentration increases thermal conductivity, the upper loading limit
usually being associated with the maximum permissable processing
viscosity.
Fillers, when incorporated in sufficient concentration, can also be
successful in reducing the thermal expansion coefficient of epoxies down to
similar levels to that obtained with many metals. This can be advantageous
in electronic and structural adhesive bonding applications.
4.3.3 Shrinkage
When epoxies undergo polymerisation and cross-linking, shrinkage occurs
throughout the cure process which can be damaging in a number of
applications ranging from electronics to adhesive bonding. Although
generally regarded as superior to other thermosetting polymers such as
phenolics and polyesters, further reductions in shrinkage beyond those
found for most unformulated epoxies is frequently considered necessary.
This can often be achieved through the use of various fillers and flexibilising
ingredients. Filler incorporation reduces shrinkage by simple bulk
replacement of resin with an inert compound which does not participate in
the cross-linking process.
4.3.4 Electrical conductivity

For many electronic and electrical applications, epoxy resin formulations
exhibiting electrical conductivity are required. Unformulated epoxies, like
the majority of polymeric materials, exhibit high levels of electrical
resistivity. The judicious use offillers can, however, increase the levels of
electrical conductivity quite significantly (Lee, 1988). The filler material
most commonly employed to meet this requirement is silver, employed
either in flake or powdered form. Occasionally other fillers such as copper,
carbon black or graphite have also been considered, primarily as lower cost
but reduced efficiency alternatives. The popularity of silver stems primarily
from the absence of an oxide layer; such layers generally imparting
electrically insulating characteristics. For this reason metallic fillers such as
aluminium are seldom considered for this application.
4.3.5 Viscosity
The incorporation of fillers into an epoxy invariably results in an increased
viscosity, the maximum loading of filler for most applications often being
restricted by the maximum permissible working viscosity. Generally fibrous
fillers exert greater viscosity-enhancing effects, on an equivalent weight
basis than particulates. With the latter filler, particle size usually exerts a
dominating effect with fillers of small particle size tending to increase
viscosity to a greater degree than corresponding fillers of greater particle
size. This can be attributed to the greater surface area of the former (Potter,
1970).
Some fillers, in particular various types of silica, are capable of exerting a
thixotropic effect which has been put to good use in structural adhesives and
formulations requiring anti-sag characteristics (Lewis, 1988).
4.3.6 Toughness
The usefulness of epoxy resins for many applications is often limited by their
comparatively brittle nature and susceptibility to catastrophic failure.
Consequently numerous investigations have been conducted over many
years in an attempt to provide processes and procedures which would allow
significant improvements in toughness, achieved at minimum expense to
other important properties and parameters such as modulus and Tg •
Probably the most successful means by which this has been achieved is by a
process known as rubber-modification and the subject will be discussed in
more detail later (section 4.6).
A number of investigations have shown that the incorporation of
particulate fillers such as silica, glass micro spheres and alumina trihydrate
can increase the toughness of various epoxy formulations (Moloney et ai.,
1983, 1984, 1987; Spanoudakis and Young, 1984). Perhaps of major
advantage here is that this can be achieved with an improvement in modulus;
although rubber modification generally provides a more efficient means of
toughness enhancement, this is usually achieved at the expense of other
important properties such as modulus.
Variables such as particle size, particle size distribution, particle surface
chemistry and particulate volume fraction have been studied by various
researchers which have indicated the most important variables necessary for
optimum toughness enhancement.
A mechanism based upon the concept of 'crack pinning' (Figure 4.4) has
been proposed to explain the toughness improvements brought about by
particulate reinforcement (Lange, 1970; Evans, 1972; Green et ai., 1979).
- - - - - - - - - - - - - - crack breakaway
crack pinning
fl)~ rigid particles
initial crack front
crack growth direction
Figure 4.4 Crack pinning mechanism for toughness improvements.

It proposes quite simply that a propagating crack front, when encountering

an inhomogeneity, becomes temporarily pinned at that point. An increase in
load increases the degree of bowing between pinning points resulting in both
new fracture surface and an increase in length of crack front. These
processes lead to an absorption of energy and therefore an improvement in
toughness (chapter 5, section 5.8.1).
4.4 Resinous modifiers
A wide variety of what could possibly be described as resinous materials

have been considered for use in combination with epoxies either as a means
of cost reduction or so as to impart property modifications. Typical
examples include combinations of epoxies with nylons, polysulphides,
polyvinyl formal and butyral, polyurethanes, styrene-butadiene co-
polymers, chlorosulphonated polyethylene, fluorinated polymers, silicones,
isocyanates, furfural resins and acrylics.
One particular critical aspect of epoxies mentioned on numerous
occasions in this chapter concerns brittleness, with the use of particulate
fillers, rubber modification and thermoplastics in particular having been
shown as fairly effective routes to toughness enhancement. In addition to
these approaches, 'alloying' epoxies with other materials of a resinous
nature has also been shown to be a particularly effective means of enhancing
toughness. Combinations of epoxies with, in particular, nylon, poly-
sulphides and polyvinyl formal/butyral have been the more traditional
approach to toughness improvement (Lewis, 1988). In particular with
nylon-epoxies, the levels of toughness achieved have been sufficient to allow
their use as structural adhesives exhibiting high lap-shear and peel strength
values at ambient and slightly elevated temperatures. However, due to the
presence of the extremely hygroscopic nylon constituent, these systems have
been shown to be vulnerable to the effects of atmospheric moisture which
has greatly limited their use (chapter 7, section 7.7.2).
The combination of epoxies with phenolics has been shown to produce
resins having a higher temperature capability than those obtained with
epoxies alone. Indeed structural adhesives systems based upon an epoxy-
phenolic chemistry have been shown to exhibit extremely attractive short-
term high temperature capabilities (Wake, 1982).
Shaw and Tod (1990) have recently combined epoxy resins with
bismaleimide prepolymers in an attempt to produce structural adhesives
exhibiting the prime attributes of the two systems, i.e. the high temperature
capability of the bismaleimide with the excellent process ability of the epoxy.
Reasonable success was achieved. Indeed resin systems supposedly based
on epoxy-bismaleimide chemistry have recently appeared commercially as
structural adhesives and as the basis of composite prepregs (Landman,
1984).
Since epoxies are capable of absorbing, in some cases, considerable
quantities of water which can have a severely debilitating effect on various
important mechanical properties, many investigations have been conducted
in an attempt to identify the various molecular and structural factors which
are of importance. As a result it is now possible to tailor epoxies with a lower
propensity for water absorption together with a lesser sensitivity to the
effects of absorbed moisture.
Many of the formulatory ingredients described in this chapter will exert an
influence on water absorption characteristics. Studies conducted by various
workers have shown that the incorporation of halogens into the network
structure of various epoxy systems can improve hydrophobic characteristics
considerably. In particular Griffith (1982) has developed fluorinated epoxy
resins together with specifically designed curing agents having the structures
shown in Figure 4.5. Investigations conducted by Shaw and co-workers
(1988) have demonstrated the pronounced hydrophobicity which these
o CF~CF3
I I /0,
/~
CHo-CH-CHo-O-C
- - I
0 C-O-CHo-CH-CHo
I - -
CF3 CF 3
C"F2,,+i
Fluoroepoxy resin
Siliconeamine curing agent
o
1/
C;@:FOC'o
I - /
HO- y \
C
CF3 0
Fluoroanhydride cUling agent
Figure 4.5 Fluoroepoxy resin and curing agent structures.

systems are capable of in comparison to the more traditional non-

halogenated epoxies. Although such studies have demonstrated that these
systems are capable of use on their own, Twardowski and Geil (1991) have
suggested that, due to their relatively low Tg values, blending of
fluoroepoxies with, for example, DGEBA resins would provide a means of
both incorporating hydrophobicity whilst maintaining reasonable levels of
mechanical performance. Similarly, studies conducted by Goobich and
Marom (1982) have shown that the incorporation of a brominated epoxy
additive to both tetraglycidyldiaminodiphenylmethane (TGDDM) and
DGEBA systems improves hydrophobicity whilst essentially maintaining
the level of important mechanical properties.
4.5 Flexibilising/plasticising additives
In addition to filler or elastomer incorporation, a further means by which the

inherently hard and brittle characteristics of epoxies can be alleviated is by
making use of plasticising/flexibilising additives. The former approach
traditionally makes use of long-chain, non-reactive molecules which do not
become incorporated into the cross-linked epoxy structure. It is here that
the distinction is usually made between plasticising and flexibilising
materials, plasticisers acting in a non-reactive manner whilst flexibilisers is
the term usually applied to materials that react with the epoxy system during
cure. Consequently they become incorporated into the backbone structure
which generally results in them achieving a greater effectiveness than the
plasticising ingredients.
4.5.1 Plasticisers
Plasticisers of the type conventionally employed in vinyl type polymers such
as PVC have not achieved wide usage in epoxy technology. In addition to
being essentially incompatable with epoxies they also suffer from the
traditional disadvantage exhibited by inert plasticising materials in being
prone to separation from the base epoxy. Even in circumstances where these
disadvantages are not manifested, the extent of flexibilisation achieved is
generally slight in comparison to the more efficient reactive systems.
4.5.2 Reactive flexibilising additives

In addition to reducing the basic hardness and rigidity of epoxies,
flexibilising ingredients are also able to confer improvements in terms of
reduced exotherm and, in certain circumstances, reducing shrinkage (Lee,
1988). The flexibilised nature of the cross-linked network can result in a
more strain tolerant system, better able to relieve internal stresses which can
develop in cured networks, particularly when employed in high volume
castings. They can improve adhesive joint properties such as lap shear and
particularly peel strength in addition to improvements in impact strength
and low temperature crack resistance. On the negative side, reactive
flexibilisers can yield cured systems that are less strong mechanically and
have reduced electrical and, in particular, chemical and solvent resistance in
comparison to their non-flexibilised counterparts.
Some of the reactive flexibilisers which have been considered for use in
epoxies will now be briefly discussed.
4.5.2.1 Polyamides. The polyamides typically employed in many epoxy

formulations are produced by a Diels-Alder reaction of the 9,12 and 9,11
isomers of linoleic acid to form the dimer, which, on further condensation
with an aliphatic polyamine, results in a polyamide. A typical structure is
shown in Figure 4.6. As indicated, the molecule contains not only amide
groups but also primary and secondary amine groups; the latter groups being
capable of participating in cross-linking reactions with the epoxy. They are
therefore capable of exhibiting flexibilising characteristics as well as being
curing agents in their own right. With DGEBA resins, polyamide
concentrations of approximately 25 to 50 phr are sufficient to provide cured
products exhibiting both toughness and impact resistance. Although higher
loading levels will result in further increases in flexibility, this will be
accompanied by reductions in, for example, Tg and properties associated
with it together with products exhibiting a weak, 'cheesy' consistency.
4.5.2.2 Carboxyl-terminated polymers. Possibly the most commonly

employed carboxyl-terminated polymers which have been employed in
epoxy resin technology are the carboxyl-terminated butadiene-acrylonitrile
rubbers. When employed in concentrations of less than 20 phr they usually
exert a toughening rather than a flexibilising effect (see section 4.6).
o
II
C-NH-(CHo)o -NH-(CH 2h-NH2
I -- -
(CHo) 7
I - 0
/CH II
CH 'CH -(CH2h -C -NH-(CH2h-NH-(CHo)2-NH2
1/ I ""
CH, /CH-CH 2 -CH=CH-(CHZ)4-CH3
C
I
( CHo)
I - 5
CH 3
Figure 4.6 Structure of a typical polyamide curing agentlflexibilising additive.
Figure 4.7 Polysulphide structure.
However, by judicious choice of elastomer concentration, the curing nature

of the fomulation (type of curative and cure conditions) as well as the
chemical nature of the elastomer, it is possible to introduce a high level of
flexibility into an epoxy using these materials. Improvements in flexibility
have also been achieved by the use of carboxyl-terminated polymers such as
carboxyl terminated polyisobutylenes (Lee, 1988).
4.5.2.3 Polysulphides. Polysulphides are essentially mercaptan-termi-

nated elastomeric materials having the structure shown in Figure 4.7. The
value of n in the structure can vary from about 2 up to 26, yielding a wide
range of properties and flexibilising capabilities, although the higher
molecular weight materials can impose serious processing difficulties due to
their high viscosities. The polysulphides are usually used at concentrations
of between 25 and 100 phr in order to provide fairly tough, impact-resistant
products. At even higher concentrations (>100 phr) formulations can
exhibit poor tear resistance together with a decrease in other important
characteristics such as hardness, shrinkage and electrical resistivity with
increased polysulphide loading. As with most attempts at flexibilisation, the
incorporation of polysulphides will generally result in substantial reductions
in Tg and properties associated with it. In addition they also exhibit an
objectionable odour which, fortunately, usually disappears after cure.
4.5.2.4 Polyglycol diepoxides. The general structure of the polyglycol

diepoxides is as shown in Figure 4.8. These materials are generally produced
by reacting the terminal hydroxyl groups of the appropriate glycol with
epichlorhydrin so as to obtain the diglycidyl ether. They are generally used
as blends with epoxies of the DGEBA or novolac-epoxy types, typically at
concentrations of between 10 and 30 phr. At these levels, properties such as
elongation, impact strength and occasionally tensile strength are increased
with, as is usually typical, reductions in Tg and heat distortion temperature
(HDT). Chemical resistance and electrical properties also tend to suffer as
the concentration of flexibiliser is increased (Lee, 1988).
Figure 4.8 Structure of a polyglycol diepoxide.

4.6 Elastomeric modification
Of all the approaches that have been considered and adopted in an attempt
to alleviate the brittle characteristics of epoxy resins, elastomeric
modification has possibly been the most successful, where quite dramatic
improvements in toughness have been achieved with, in many cases, only
modest reductions in other important properties. Due to the technological
importance with which elastomeric modification is now viewed, it will be of
interest and indeed importance to consider this subj ect in some detail. In this
account the various factors which have been shown to be of importance for
successful toughness enhancement, including factors such as elastomer
structure and chemistry, compatibility and morphology, will be discussed
together with an account of the main mechanism which has been proposed to
account for the effects of elastomeric modification.
4.6.1 Types of elastomeric modifiers

The various types of elastomeric materials which have been considered for
the rubber modification of epoxies can be catagorised thus:
1. Reactive butadiene-acrylonitrile rubbers
2. Polysiloxanes (Yorkgitis et al., 1984)
3. Fluoroelastomers (Mijovic et al., 1984)
4. Acrylate elastomers (Kirshenbaum et al., 1984).
Of these the first, the butadiene-acrylonitrile rubbers have been by far the
most widely studied and employed in the rubber-modified systems which
have achieved the most commercial success. Thus the majority of this
section will be concerned primarily with these materials.
Butadiene-acrylontrile rubbers, as the name implies, are composed of a
relatively low molecular weight backbone consisting of butadiene and
acrylonitrile groups with reactive groups in the terminal positions of the
molecule as indicated:
X-f-CHz-CH =CH-CHzh-fCHz-cHh-h x
I
CN
Numerous variants of these systems have been investigated with the main
variables being (a) acrylonitrile content (Rowe et al., 1970), (b) molecular
weight (McGarry and Wilner, 1968) and (c) nature of the terminal group
(Siebert and Riew, 1971).
The acrylonitrile portion of the molecule imposes a degree of polarity on
an otherwise non-polar structure. As will be explained later a compatibility/
incompatibility balance dictates to a large degree the success or otherwise of
attempts at toughness enhancement. This is largely controlled by the
relative polarities of the resin and rubber components, the latter being
influenced by the acrylonitrile content. Various morphological

characteristics and indeed mechanical properties have been related to the
rubber acrylonitrile level. For reasons also related to compatibility and
morphology, elastomer molecular weight has also been shown to be an
important variable.
Experience has shown that terminal reactivity is an equally important
requirement for effective toughness enhancement. It is now generally
recognised that good adhesion between both the elastomeric and matrix
components of a rubber-modified thermoset is of vital importance. This can
usually be achieved by ensuring a means by which the liquid rubber can react
with the thermosetting polymer, in this case an epoxy. The type of elastomer
chosen therefore, at least in terms of terminal functionality, will depend
upon the chemical characteristics of the matrix polymer. With epoxies a
number of terminal functionalities have been studied including epoxy,
hydroxyl, phenol, mercaptan, vinyl, amine and carboxyl with at least
reasonable success being achieved in the majority of cases (Riew et al.,
1976). However, possibly the greatest benefits have been obtained from
elastomers containing the carboxyl functionality (CTBNs).
Other elastomer types such as the polysiloxanes and polyacrylates are
usually only considered when application conditions are of such an extreme
nature so as to prevent employment of the butadiene-acrylonitrile systems.
For example, the polysiloxanes exhibit important advantages in terms of
both low temperature flexibility (due to a low Tg of about -120°C relative to
- 35°C for butadience-acrylonitrile systems) and excellent weatherability
and moisture resistance. In addition the unsaturated structure of butadiene-
acrylonitrile systems leaves them prone to thermal instability and thus
unsuitable for long-term use at elevated temperatures. The polysiloxanes,
acrylates and flu oro elastomers can be regarded as having superior thermal
resistance and would thus be considered more suitable for such conditions.
Although exhibiting these distinct advantages, work to date has shown the
clear superiority of the butadiene-acrylonitrile rubbers in promoting the
greatest toughness enhancement in epoxy resins.
4.6.2 Compatibility and morphology

It is now universally recognised that in order to promote a substantially
enhanced toughness phase separation and the deVelopment of a two-phase
morphology is critically important whilst maintaining or at least minimising
the deterioration in other important properties/parameters such as modulus
and Tg• For significant toughness enhancement the rubber must initially
dissolve and become dispersed on a molecular level in the resin but
precipitate when the epoxy begins to cross-link and thus form the required
two phase morphology with rubbery particles dispersed in and bonded to the
cross-linked epoxy matrix. Thus elements of both initial compatibility and
eventual incompatibility must be regarded as vital elements of the system.

This approach can be viewed in a more theoretical sense by considering the
thermodynamics of mixtures in general and solubility parameters in
particular.
The compatibility/incompatibility of any mixture of two substances is
controlled fundamentally by the Gibbs free energy of mixing, I1Gm given by
(4.1)
where I1Hm and I1Sm are the molar enthalpy and entropy of mixing
respectively.
By considering a cohesive energy density approach to molecular
interaction, compatibility can, to a degree, be characterised by the solubility
parameter, 0, the value of which for a particular material determines the
tendency of its molecules to attract its own species in preference to those of
a dissimilar species. This approach can be expressed quantitatively in the
equation,
(4.2)
where V is the molar volume of the mixture, 01 and 02 are solubility

parameters for phases 1 and 2 and (j>1 and (j>2 are the respective volume
fractions of the two phases. Molecules having similar chemical structures are
generally found to have similar solubility parameters resulting, in turn, in a
reduced tendency to demix. However since this approach only considers
molecular interactions resulting from dispersion forces, anomalies do occur,
particularly in systems where polar interactions and hydrogen bonding play
a prominent inter-molecular role.
Combining equations 4.1 and 4.2 yields the following equation providing a
more detailed view of the factors influencing I1Gm and hence miscibility
(4.3)
Although, as mentioned above, lacking in certain respects, equation 4.3

provides a clear insight into the factors that will playa prominent role in both
phase separation and morphology development. In order to develop the
required morphology (i.e. small elastomeric particles present in the cross-
linked resin matrix) good compatibility between epoxy and elastomer is
required upon mixing of the two phases. Thus an initial requirement is that
I1Gm should be negative. As shown clearly in equation 4.3 this would be
promoted by small differences in the 0 parameters together with a low molar
volume (i.e. low molecular weight). However, in order to avoid the
continued intimate dispersion of the elastomeric component and hence
allow flexibilisation of the cross-linked polymer, at some stage in the cure
process, I1Gm must become positive thereby leading to both phase
separation and two-phase morphology development. From a molecular
standpoint this could be achieved by changes in molecular structure

resulting in diverging values of <5 or by an increase in molecular weight or a
combination of both occurring during curing. With most, if not all, rubber-
modified epoxy formulations, it is the molecular weight build-up during cure
(resulting in an increased magnitude of V) that is believed to result in a
change in ~Gm from negative to positive and thus phase separation. Thus
equation 4.3 in a somewhat simplistic sense, indicates possibly the two main
molecular factors that will control the critically important process of phase
separation and morphology development, i.e. factors which will playa
prominent role in determining the properties of the cross-linked polymer,
these being chemical structure and molecular weight.
With the more commonly studied DGEBAlCTBN formulations, various
factors have been shown to influence morphology. The most notable of
these include rubber concentration (Riew et al., 1976), type and extent of
curing agent employed and cure conditions (Chan et al., 1984; Shaw and
Tod,1989).
As would be expected, increasing the quantity of elastomer in the system
results in an increase in the volume fraction of the elastomeric phase and, as
has been shown in various studies, also produces an increase in the size of
elastomeric domains. Elastomeric concentrations in excess of about 20% by
weight have generally been shown to exhibit a process known as phase
inversion where a reversal in morphological characteristics occurs resulting
in cured epoxy particles embedded in an essentially elastomeric matrix
(Bascom et al., 1975). Although many studies have shown a positive
relationship between rubber volume fraction and the extent of toughness
enhancement (Chan et al., 1984) as would be expected, mechanical
properties generally deteriorate rapidly at elastomer concentrations in
excess of this value (Bascom et al., 1975).
The molecular weight and acrylonitrile content of the rubber have also
been extensively studied. Generally, molecular weight reductions and
enhanced acrylonitrile content have shown similar effects namely reductions
in both elastomer particle size and volume fraction together with increased
tendencies towards entrapment of elastomer in the matrix and thus
flexibilisation.
The type of curing agent employed will playa major role in determining
the mechanical properties of a rubber-modified epoxy. Not only will it
strongly influence the structure of the cross-linked epoxy phase and
therefore its mechanical properties, it will also be capable of exerting a
substantial influence on morphology. This it can do by firstly altering the
compatibility/incompatibility balance between the elastomer and epoxy.
This effect can be particularly severe with curing agents which, by necessity,
need incorporation in large quantities e.g. anhydrides and certain modified
amine-based curatives. Perhaps of greatest significance, however, is the
curing agent reactivity. Substantial evidence is now available which
indicates that curing agents of a highly reactive nature are relatively

inefficient in promoting substantial toughness improvements. This is
generally attributed to what could be described as 'morphological
constraint', where the highly reactive nature of the curative agent results in
a relatively rapid onset of gelation which thus interfere with and prevents
full phase separation. Increased curative reactivity, as a result, is generally
found to produce systems exhibiting both low volume fractions of the rubber
phase and relatively small particle size. Since a high volume fraction is of
major importance for optimum toughness enhancement, the use of highly
reactive curing agents should, if possible, be avoided. Since many
commercial applications, driven by economic considerations, require the
rapid cure times provided by highly reactive curatives, this can cause some
shortfall in properties.
Recent research has shown that cure conditions can have as significant an
effect on the mechanical properties of a rubber-toughened epoxy as the
formulation variables highlighted above (Shaw and Tod, 1989). Once again
the influence of this parameter, particularly cure temperature, on
morphology is generally considered responsible. Although the effects of
formulation variables on morphology are now well recognised, the influence
of cure temperature is somewhat less predictable, due to its ability to
influence several factors responsible for affecting phase separation. For
example, an increase in cure temperature will promote both an
improvement in epoxy/rubber compatibility and an increase in the cure rate;
factors which will tend to retard the phase separation process leading to
relatively small values of both rubber volume fraction and particle size.
Conversely an increase in temperature would also result in a reduction in
viscosity which would enhance the phase separation process thereby having
the reverse effect on the morphological characteristics described above. The
final influence of cure temperature, therefore, will be dependent on the
relative importance of compatibility, viscosity and cure rate on the phase
separation process. Although some conflicting trends have been observed,
particularly with regard to rubber volume fraction, an increase in cure
temperature has been found to usually result in an increased rubber particle
size (Shaw and Tod, 1989).
At this point it is perhaps beneficial to briefly discuss the morphological
factors that are generally believed to control the toughness of rubber-
modified epoxies. Although this has been, and to a degree still is a somewhat
controversial area, evidence accumulated in recent years has focused upon
what would now seem to be the most critical parameters. Essentially the
debate is centred upon the relative importance of two morphological
characteristics, namely rubber volume fraction and particle size/particle size
distribution. From their early pioneering research, Sultan and McGarry
(1973) suggested that the key morphological factor responsible for optimum
toughening was particle size, with relatively large particles (> 1.um
diameter) promoting crazing within the epoxy matrix and smaller particles
(approximately 0.01 J-lm) leading to shear deformation. Craze generation,
in their view, was seen to be the most efficient means of toughness
enhancement and they therefore proposed that the formation of relatively
large rubber particles was the key to maximum toughness. Crazing is now, of
course, almost universally recognised as not being a significant deformation
mechanism with epoxies. So in this respect alone this theory can be regarded
as flawed. However, in one major respect involving the formation of
particularly tough rubber-modified epoxies, the role of particle size and, in
particular, particle size distribution has been and indeed continues to be
emphasised. In the 1970s Riew et al.(1976) found that by incorporating
bisphenol A into a DGEBA/CTBN/piperidine system, substantial improve-
ments in toughness could be achieved over and above that obtained from
the non-bisphenol A formulation. They noted in particular that this highly
beneficial effect was accompanied by a bimodal particle size morphology,
which they regarded as responsible for the enhanced mechanical
performance. Since this early work, similar observations have been found by
other workers employing the same approach and indeed the bimodal
particle size concept has been utilised in commercial rubber-modified epoxy
formulations, where the incorporation of both low and high molecular
weight CTBN rubbers have been found to produce the required morphology
together with impressive fracture performance data. Although such
evidence would appear to strongly support the particle size protagonists, it is
important to recognise that the parameters of interest, volume fraction and
particle size, are extremely difficult to vary in magnitude independently of
each other. Thus, for many formulations, although bimodal particle size
distributions may exist and indeed exert an influence on the properties of the
epoxy, it is likely that volume fraction could also vary in a significant
manner.
Although a relatively small number of researchers would still appear to
view particle size and size distribution as providing a not insignificant
contribution to toughness enhancement, most now recognise that the critical
morphological factor is rubber volume fraction. Generally the greater the
volume fraction of the phase-separated elastomeric component the greater
the improvement in toughness. Although this will partly depend on the
amount of rubber initially added to the formulation, the other factors
described previously, e.g. rubber molecular weight and acrylonitrile
concentration, curing agent reactivity etc. will also have an important
influence. It is particularly worth while noting that numerous studies have
shown that rubber phase volume fraction usually exceeds the volume
fraction of added rubber thus providing an insight into the composite nature
of the rubber domains. An anomalous effect such as this can only be
resolved if it is assumed that the rubber particles have a not insignificant
concentration of epoxy, either on the molecular level or as sub-inclusions
dispersed in the elastomeric phase.
4.6.3 Toughening mechanisms

Several mechanisms have been proposed to account for the toughness
enhancements generated by the incorporation of elastomeric particles into
glassy polymers such as epoxies. These include theories based upon
contributions arising from rubber tear, multiple crazing and shear yielding.
A toughening model developed independently by Kinloch et al. (1983)
and by Pearson and Yee (1983) based upon both cavitation and matrix shear
yielding is now generally recognised as the most consistent in terms of the
wealth of experimental data generated in recent years. Due to its
importance it is of interest to discuss this model in some detail.
Many studies have indicated a positive correlation between fracture
toughness and the extent of plastic deformation found on fracture surfaces.
This, together with only very limited evidence for crazing processes have
suggested a mechanism based upon yielding and plastic shear flow of the
matrix as the primary source of energy dissipation during fracture. Since
greatly enhanced plastic deformation has been almost universally associated
with rubber particle incorporation it is clearly necessary to focus attention
on the rubber particles themselves and in particular the distribution of stress
which exists around them when located in the vicinity of a loaded crack tip.
This is best achieved by consideration of the work of Goodier (1933) who
studied the case of an isolated spherical particle embedded in an isotropic
elastic matrix which is subjected to a uniform uniaxial tensile stress at points
remote from the particle. He derived equations which indicated that for a
rubbery particle, which possesses a considerably lower shear modulus than
the matrix, the maximum stress concentration occurred in the equatorial
regions of the particle with a value of approximately 1.9. However, allowing
for complicating factors such as the large volume fraction of rubber particles
which would invariably exist, together with the triaxial as opposed to the
uniaxial stress state which particles would encounter, a stress concentration
factor of approximately 1.6 would exist.
The interactions between a triaxial stress field in the vicinity of a loaded
crack tip and that existing around rubber particles in close proximity could
be envisaged resulting in the initiation of two important processes.
First, the development of a triaxial stress would lead to dilatation of the
matrix. This, in combination with triaxial stresses inherent in the rubber
particles due to differential thermal contraction effects resulting from the
cure process, would cause cavitation of the particles. Indeed it is commonly
found that elastomeric materials undergo cavitation quite readily under the
action of a triaxial stress field. This cavitation process can be viewed as
responsible for the stress whitening effects observed in the numerous studies
conducted on rubber-modified epoxies.
The second process concerns the initiation and growth of shear yield
deformation in the matrix. During loading the stress concentrations
developing at the rubber particle equators would act as sites for the initiation
of shear deformation. Due to the large number of particles present, the

degree of yielding generated would be substantially greater than would
otherwise occur in an unmodified material. However, particles would also
act as sites for shear yield termination which would therefore keep the
yielding localised.
It is possible and indeed likely that both of the above processes would
occur during the early stages of loading, but with rubber particle cavitation
probably being the most dominant. However, once initiated, cavitation
within the particles would lead to more pronounced void formation which in
turn would greatly enhance further shear yielding in the matrix. This would
be partly attributable to an increased stress concentration which would be
likely to accompany cavitation and void formation. However, of greater
importance, void formation would result in reduced constraint on the matrix
adjacent to voided particles, thus relieving the degree of triaxiality. This in
turn would lower the yield stress in the inter-particle regions and thus
promote further extensive shear yielding. This would cause extensive crack
blunting ultimately resulting in the increasing development of the plastic
zone at the crack tip. It is the enhanced development of this plastic zone
which results in the greatly improved fracture toughness exhibited by
rubber-modified epoxies in relation to their unmodified counterparts.
4.6.4 The hybrid modification approach

Although some fairly modest improvements in toughness have been
observed with particulate reinforcement (section 4.3.6) these have not
generally been of the same magnitude as that obtained with elastomeric
modification. However, as previously described, the substantial modulus
increase which generally accompanies toughness enhancement with
particulate fillers can be regarded a major advantage. As a result several
studies have considered a hybrid approach to toughening; making use of
both particulate reinforcement and elastomeric modification (Maxwell et
al., 1984; Kinloch et al., 1985). Much of this work has shown significant
benefits in terms of both substantial improvements in toughness and
modulus, which can be achieved by a combination of this type. It remains to
be seen whether these attractive features result in any significant
commercial usage.
4.7 Thermoplastic modification
The inclusion of thermoplastics into epoxy formulations can be regarded as

a more recent means of imposing property variations into cross-linked
epoxies and recent studies have shown that once again toughness
enhancement is the major driving force.
Attempts at imparting significant improvements in toughness with high
cross-link density epoxies such as those based upon the traditional aerospace
epoxy, TGDDM, by use of elastomeric modifiers has generally resulted in
failure. This has been attributed to the inability of the cross-linked epoxy
matrix to undergo any significant shear yielding in response to the presence
of the rubber particles. A somewhat different approach, using for example
thermoplastics, has therefore been seriously considered, particularly for the
more difficult to toughen multifunctional systems.
Attempts at thermoplastic modification have been conducted using a
fairly limited range of thermoplastic polymers including, for example,
polyethersulphone (Bucknall and Partridge, 1983) and polyetherimide
(Bucknall and Gilbert, 1989), the basic structures of which are shown in
Figure 4.9.
Bucknall and Partridge (1983) attempted to study the relationship
between the structure and mechanical properties of two types of poly-
ethersulphone-modified epoxy, one being trifunctional and the other based
upon the TGDDM resin. They found that morphology, studied using both
dynamic mechanical spectroscopy and scanning electron microscopy, was
dependent upon both the type and concentration of resin and the curing
agent employed. In certain circumstances two-phase morphological
characteristics were observed with a polyethersulphone-rich phase
dispersed in a cross-linked epoxy matrix. Because of both high modulus and
Tg values of the polyethersulphone employed (2.8 GPa and 210°C
respectively), large concentrations of thermoplastic generally had a minor
effect on these particularly important parameters. This, in fact, represents a
significant advantage over rubber-modification where reductions in these
parameters, particularly modulus, can occur. Unfortunately, in this study,
polyethersulphone
o 0
t<l§r@q:@°©t>-ol
II II
II
0
l8J I CH
3 0
II n
polyetherimide
Figure 4.9 Basic structures of polyethersulphone and polyetherimide.
polyethersulphone addition was found to have only a minor beneficial effect

on fracture energy with thermoplastic additions of up to 40 phr resulting in
less than a doubling in fracture energy.
Somewhat similar studies have been conducted by others including
Diamont and Moulton using polyetherimide (Diamont and Moulton, 1984)
and McGrath and co-workers (Hedrick et al., 1985; Cerere et al., 1986) who
employed hydroxyl- and amine-terminated sulphones as modifiers for
DGEBA epoxies. Although in both cases increases in toughness were
obtained, as in the work of Bucknall and Partridge, the magnitude of
improvement was small in comparison to that which could be expected with
elastomeric modification.
More recent work by Bucknall and Gilbert (1989) using polyetherimide
modification, showed that the incorporation of 30 phr of thermoplastic into
an anhydride-cured TGDDM resin resulted in an approximately threefold
increase in fracture toughness, K 1C , with modulus remaining virtually
constant.
Although the use of tough, thermoplastic polymers provides for
improvements in epoxy toughness, in comparison to rubber-modification
the level of improvement is often generally poor, particularly with relatively
low Tg systems. With high cross-link epoxies where rubber-modification
would probably provide a similar, relatively poor, level of toughness
enhancement as that provided by thermoplastics, the comparatively high
modulus and Tg values of the latter would be regarded as highly
advantageous. Thus in the majority of cases it would seem likely that
thermoplastic modification would only be considered worth while for
relatively high cross-link density epoxies under consideration for high
temperature applications.
4.8 Miscellaneous additives
In addition to the main formulation types discussed in this chapter, countless

other miscellaneous additives have been either considered or used in various
epoxy systems. Since space clearly prevents a detailed account of all these
ingredients, it is of interest to consider a few that have either been employed
or which recent studies have indicated have considerable potential.
The use of coupling agents in epoxy resin technology, particularly in
structural adhesive bonding applications, is now fairly well established.
Most of the published literature within this context has been concerned
primarily with organosilanes as pre-bond primers. Studies devoted
principally to one specific organosilane, y-glycidoxypropyltrimethoxy-
silane, have shown that this material can dramatically improve the moisture
resistance of epoxy adhesive bonded metallic joints (Kinloch et al., 1975;
Gledhill et al., 1990). Unfortunately, the use of a primer as a coating
prior to application of an adhesive adds an extra step to the bonding process.
As a result, the possibility of adding a coupling agent to the adhesive has
been considered and indeed is common commercial practice with some
epoxy adhesive manufacturers. Since coupling agents such as organosilanes
are generally considered 'surface active', i.e. they exert an influence on the
interfacial zone between the substrate and epoxy, a mechanism of migration
and interface activity would clearly need to apply with an approach of this
kind.
As well as adhesive bonding, organosilanes have also been employed to
enhance the properties of filled epoxy systems, particularly where resistance
to moisture is required.
Since most epoxies are capable of burning, for many applications flame
retardency is considered important. This can be achieved by incorporating
into the formulation halogen (e.g. bromine )-containing resin or curative
species (Tanaka, 1988), flame retarding fillers such as antimony trioxide
(Potter, 1970) or somewhat more esoteric additives such as amino alkyl
phosphates or halogenated fi-Iactones (Mika and Bauer, 1988).
Although the presence of a solvent in an epoxy formulation would be
regarded as a substantial handicap in many applications, they are an
important formulatory ingredient for one specific application, namely
surface coatings (Tess, 1988). Solvents which have been considered for this
purpose include methyl ethyl ketone, 2-ethoxyethanol, xylene, 2-ethoxy
ethyl acetate and toluene either alone or more generally in blends of various
solvents where greater solvating power has been demonstrated. Numerous
factors will require consideration in the appropriate choice of solvent(s).
These include solvent power and evaporation rate which can strongly
influence the final properties of a surface coating together with viscosity,
cost and of course the highly important subject of health and safety.
In previous sections in this chapter, the general question of resin
shrinkage upon cure has been briefly discussed, with particular emphasis on
the beneficial effects of ingredients such as fillers and flexibilisers. In recent
years, a somewhat novel approach, making use of the so-called 'expanding
monomer' concept has yielded beneficial effects. Research conducted by
Piggott and co-workers (1985) has shown that, by copolymerising with
epoxies compounds such as dinorbornene spiro ortho carbonate, the levels
of shrinkage encountered upon cure can be greatly diminished resulting in
substantial improvements in properties which would be affected by the
internal stresses resulting from shrinkage. In particular, improvements in
the toughness of epoxy-carbon composites have been shown to result from
such modifications.
Studies recently published have demonstrated a novel approach to
improving the moisture resistance of epoxies. Fisher et al. (1985), Hu et al.
(1987) and Kelly and co-workers (1988) showed that the amount of water
absorbed by a TGDDM epoxy cured with diaminodiphenylsulphone (DDS)
could be lowered significantly by so-called blocking agents when absorbed

into thin films of the cured epoxy. The agents employed were isocyanate
compounds and were believed to function by reacting with polar functional
groups (such as hydroxyls and secondary and tertiary amines formed during
cure) contained within the cross-linked structure. It is generally accepted
that such functional groups are responsible for water absorption via
hydrogen bonding mechanisms. The fact that this approach relied upon
absorption of the blocking agent into the cured epoxy essentially meant that
the modification could only be applied to very thin films. More recent work
by Lonikar et at. (1990) has attempted to overcome this problem by
incorporating the reagents into the epoxy formulation prior to cure. This
was achieved by employing 'masked' isocyanates, the reactivity of which
is suppressed at temperatures well above room temperature. At suitably
elevated temperatures the isocyanate loses its latency allowing it to react
with polar functional groups formed during cure of the epoxy, resulting in
significant decreases in equilibrium moisture absorption. As far as the
author is aware, blocking agents of the type described have not been
employed in commercial practice as a means of promoting enhanced
moisture resistance in epoxies.
References
Bascom, W.D., Cottington, RL., Jones, R.L. and Peyser, P. (1975) J. Appl. Polym. Sci. 19,
2545.
Bucknall, C.B. and Partridge, LV. (1983) Polymer 24,639.
Bucknall, C.B. and Gilbert, A.H. (1989) Polymer 30,213.
Cerere, J.A., Hedrick, J.L. and McGrath, J.E. (1986) 31st SAMPE Symp. 583.
Chan, L.C., Gillham, J.K., Kinloch, A.J. and Shaw, S.J. (1984) In Rubber-modified Thermoset
Resins, eds. Riew, C.K. and Gillham, J.K., Adv. Chern. Ser. 208, Am. Chern. Soc.,
Washington.
Diamont, J. and Moulton, R.S. (1984) 29th Nat. SAMPE Symp. 29,422.
Evans, A.G. (1972) Phil. Mag. 26, 1327.
Fisher, C.M., Gilbert, R.D., Fornes, R.E. and Memory, J.D. (1985) J. Polym. Sci., Polym.
Chern. ed. 23, 2931.
Gledhill, RA., Shaw, S.J. and Tod, D.A. (1990) 1nt. J. Adhesion and Adhesives 10,192.
Goobich, J. and Marom, G. (1982) Polym. Eng. Sci. 22, 1052.
Goodier, J.N. (1933) J. Appl. Mech. 55,39.
Green, D.J., Nicholson, P.S. and Emberg, J.D. (1979). J. Mat. Sci. 14,1657.
Griffiths, J.R. (1982) Chern tech 290.
Hedrick, J.L., Yilgor, I., Wilkes, G.L. and McGrath, J.E. (1985) Polym. Bull. 13,201.
Hu, H.P., Gilbert, R.D. and Fornes, R.E. (1987) J. Polym. Sci. Polym. Chern. ed. 25, 1235.
Kelly, B.K., Gilbert, R.D., Fornes, RE. and Stannett, V.T. (1988) J. Polym. Sci., B: Polym.
Phys. 26, 1261.
Kinloch, A.J., Gledhill, RA. and Dukes, W.A. (1975) In Adhesion Science and Technology,
ed. Lee L.H., Plenum, New York.
Kinloch, A.J., Shaw, S.J. and Hunston, D.L. (1983) Polymer 24,1355.
Kinloch, A.J., Maxwell, D. and Young, R.J. (1985) J. Mat. Sci. 20,4169.
Kirshenbaum, S.L., Gazit, S. and Bell, J.P. (1984) In Rubber-modified Thermoset Resins, eds.
Riew, C.K. and Gillham, J.K. Adv. Chern. Ser. 208, Am. Chern. Soc., Washington.
Landman, D. (1984) Polyimides. Synthesis, Characterisation, and Applications, Vol. 1. ed.

Mittal K.L., Plenum, New York.
Lange, F.F. (1970) Phil Mag. 22,983.
Lee, S.M. (1988) Epoxy Resins, Chemistry and Technology, 2nd edn, Ch. 9. ed. May C.A,
Dekker, New York, p. 783.
Lewis, AF. (1988) Epoxy Resins, Chemistry and Technology, 2nd edn, Ch. 7., ed. May C.A,
Dekker, New York.
Lonikar, S.V., Rungsimuntakul, N., Gilbert, R.D. and Fornes, R.E. (1990) J. Polym. Sci. A.
Polym. Chern. 28,759.
Maxwell, D., Young, R.J. and Kinloch, A.J. (1984) J. Mats. Sci. Letts. 3, 9.
McGarry, F.J. and Willner, A.M. (1968) Report R68-8, MIT Cambridge, USA
Mertz, E.H., Claver, G.C. and Baer, M. (1956) J. Polym. Sci. 22,325.
Mijovic, J., Pearce, E.M. and Foun, C.C. (1984) Rubber-Modified Thermoset Resins. eds.
Riew, C.K. and Gillham, J.K. Adv. Chern. Servo 208, Am. Chern. Soc., Washington.
Mika, T.F. and Bauer, R.S. (1988) Epoxy Resins. Chemistry and Technology, 2nd Edn, Ch. 4.
ed. May C.A., Dekker, New York, pp. 465-550.
Moloney, AC., Kausch, H.H. and Stieger, H.R. (1983) J. Mat. Sci., 18,208.
Moloney, AC., Kausch, H.H. and Stieger, H.R. (1984) J. Mat. Sci. 19,1125.
Moloney, A.C., Kausch, H.H., Kaiser, T. and Beer, H.R. (1987) J. Mat. Sci. 22,381.
Pearson, R.A. and Yee, AF. (1983) Polym. Mat. Sci. Eng. 49,316.
Piggott, M.R., Lam, P.W.K., Lim, J.T. and Woo, M.S. (1985) Compo Sci. Tech. 23,247.
Potter, W.G. (1970) Epoxide Resins, Iliffe, London.
Riew, C.K., Rowe, E.H. and Siebert, AR. (1976) Toughness and Brittleness of Plastics. Adv.
Chern. Ser. 154, Am. Chern. Soc., Washington.
Rowe, E.H., Siebert, AR. and Drake, R.S. (1970) Mod. Plast. 47, 110.
Shaw, S.J., Tod, D.A and Griffith, J.R. (1988) Adhesive Sealants and Coatings for Space and
Harsh Environments. ed. Lee L.H., Plenum, New York, p. 45.
Shaw, S.J. and Tod, D.A (1989) J. Adhesion 28, 231.
Shaw, S.J. and Tod, D.A. (1990) Adhesion '90, Plastics and Rubber Institute, London, 12/1.
Siebert, R.A. and Riew, C.K. (1971) Org. Coat. Plast. Chern. 31, 552.
Spanoudakis, J. and Young, R.J. (1984) J. Mat. Sci. 19,473.
Sultan, J.N. and McGarry, F.J. (1973) J. Polym. Eng. Sci. 13,29.
Tanaka, Y. (1988) Epoxy Resins Chemistry and Technology 2nd edn, Ch. 2. ed. C.A. May,
Dekker, New York, pp. 9-283.
Tess, R.W. (1988) Epoxy Resins Chemistry and Technology, 2nd edn, Ch. 8. ed. May C.A.,
Dekker, New York, pp. 719-782.
Twardowski, T.E. and Geil, P.H. (1991) J. Appl. Polym. Sci. 42,69.
Wake, W.C. (1982) Adhesion and the Formulation of Adhesives, 2nd edn, Applied Science,
London.
Yorkgitis, E.M., Trau, C., Eiss, N.S., Hut, T.Y., Yilgor, I., Wilkes, G.L. and McGrath, J.E.
(1984) Rubber Modified Thermoset Resins. eds Riew, C.K. and Gillham, J.K., Adv. Chern.
Ser. 208, Am. Chern. Soc. Washington.
5 Fracture behaviour of epoxy resins
W.J. CANTWELL and H.H. KAUSCH
5.1 Introduction
As a result of their superior mechanical properties, low shrinkage and

thermal stability, epoxy resins are finding increasing use in a wide range of
engineering applications. Over a twenty year period between 1965 and 1985,
epoxy resin consumption in Western Europe increased by over 500% to
reach approximately 132000 tonnes by the mid-1980s (Feuerhahn, 1986).
Since epoxies cost between three and fiteen times more than other
thermosets such as isophthalic polyesters, they tend to be used in more
advanced engineering applications where the increased cost of the product is
offset by their overall superior load-bearing capacity and/or chemical
resistance. In Western Europe the coatings industry represents the principal
user of epoxy resins followed by the electrical and construction industries
(Feuerhahn, 1986). Although much of the research published in the
literature relates to epoxies for use in advanced composites, this sector
represented only 4% of the total market in this year. In almost all of these
applications, properties such as high strength, stiffness and toughness are
fundamental requirements. The consequences of defects or local stress
raisers in these systems can vary from inconvenient, for example the case of
a somewhat unaesthetic damaged surface coating, to catastrophic such as
the case of a heavily delaminated load-bearing aircraft structure. A detailed
understanding of the deformation and failure mechanisms that occur in
these materials is therefore clearly necessary.
During the last thirty years a large amount of work has been undertaken
on epoxy resins. Although many of these studies have been both systematic
and comprehensive, many questions remain unanswered and the influence
of varying many basic material parameters is still poorly understood. A brief
glimpse of the extensive literature highlights the large number of widely
differing epoxy systems currently available. This is shown in Figure 5.1
where previously reported data from the literature are presented. Here, the
mode 1 fracture energy (see section 5.2.1) of the various epoxies is plotted as
a function of their glass transition temperatures.
Clearly, a range of systems is available varying from relatively tough, low
temperature epoxies for use in the construction industry for example, to
high temperature, although somewhat brittle, epoxies such as those
destined for use in the aerospace sector. The figure also highlights one of the
FRACTURE BEHAVIOUR 145
1200
•
1000
800 •
N
E •
~ 600
::: •
. •
\!I 400 •
• • ••
• ,•
• • ••
...-
• ••
200
•• ••
0
0 SO 100 ISO 200 250 300
Glass Transition Temp. (oel
Figure 5.1 The wide range of epoxy properties and associated glass transition temperatures
presently available.
great challenges to epoxy chemists, that is, a tough system with a good high
temperature capability.
In this chapter the fracture bahaviour of epoxy resins will be considered in
some detail. Since most of the recent advances in the understanding of the
fracture behaviour of these materials have been achieved as a result of the
application of fracture mechanics principles, most particularly linear elastic
fracture mechanics (LEFM) considerable use of this approach will be made
here. Attention will be given to assessing the influence of structure (cross-
link density etc.) as well as the test conditions (temperature, rate, etc.) on
the basic fracture properties of epoxy resins. In the later part of the chapter
the various methods currently employed to enhance the relatively brittle
nature of epoxies will be presented.
The following section will present a brief summary of some of the more
important LEFM principles necessary for characterizing the behaviour of
these materials.
5.2 Linear elastic fracture mechanics (LEFM)
Theoretical considerations have shown that the stress to cause cleavage

fracture in a brittle solid should be approximately one tenth of the Young's
modulus of the material. Assuming that the typical modulus of an epoxy
resin is of the order of 3 GPa one would expect to obtain a failure strength
of 300 MPa. Experimental evidence has clearly shown this not to be the case
with the fracture strength of epoxy resins rarely exceeding one third of this
value. This disparity was first recognised by Griffith (1920) who identified
the significant influence of small defects on the load-bearing properties of

a brittle isotropic material.
Modern-day linear elastic fracture mechanics is based on two inter-
relatable conditions for fracture, the energy analysis (G approach) and
localized stress field approach (K approach). The former supposes that
fracture will occur when sufficient energy is released by crack growth to
supply the requirement for creation of new fracture surfaces. LEFM is
applied to materials which obey Hooke's law where the measured strain is
proportional to the applied stress. Although all epoxies exhibit some
inelastic deformation around the crack tip, the fact that the bulk of the
remaining material behaves elastically permits the successful application of
this analysis.
5.2.1 The G approach

By applying an energy balance argument to a linearly elastic cracked body of
thickness B subjected to an applied load, it is possible to determine the
critical strain energy release rate Gc for fracture to occur
G = P~ Be
c 2B Ba
where e is the compliance of the cracked body for a given crack length a
and Pc the critical load for the onset of crack propagation. The above
equation represents the basis for most Gc calculations with the basic
requirement being a knowledge of the variation of specimen compliance
with crack length. The compliance-crack length dependency can be
determined either analytically or experimentally depending upon the
complexity of the test specimen.
It is important to note that all modes of loading encountered in
operational service can be decomposed into the three basic modes denoted
I, II and III presented in Figure 5.2. Mode I corresponds to simple cleavage
MODEm
MODEl MOOED
Figure 5.2 Modes of loading: mode I opening mode, mode II inplane shear, and mode III
antiplane shear.
or tensile opening, Mode II to in plane shear and Mode III to antiplane
shear. The fracture energies associated with I, II and III modes of
deformation are denoted GIc , G uc and G IIIc respectively. Mode I
deformation is technically the most important of the three since it represents
the most commonly encountered loading condition and generally has the
lowest associated fracture energy. Mode I-type loading is also much more
readily simulated in laboratory conditions . Typical mode I test specimens
are presented in Figure 5.3. Of these, the single edge notch (SEN) and single
/·/"--"'1iI
•
I 1 2W
•.
(a)
(b)
4W
(c)
(d)
Figure 5.3 Typical test geometries used for characterizing the mode I fracture toughness of
epoxies. (a) compact tension; (b) single edge notch; (c) single edge notch bend; (d) double
torsion.
edge notch bending (SENB) are particularly unstable geometries whereas

stable crack propagation can, under certain conditions, be achieved in the
double torsion (DT) and compact tension (CT) specimens.
In selecting the precise dimensions for the chosen specimen geometries, a
number of size criteria have to be first respected. Perhaps the most
important of these concerns the specimen thickness B. Such a thickness-
dependency relates to the state of stress at the crack tip. This can vary from
plane stress conditions in a thin plate to plane strain conditions in a much
thicker specimen. Since the stress at which a material yields is greater in a
plane strain field than in a plane stress field, a smaller plastic zone size is
observed under plane strain conditions. The effect of this smaller plastic
zone is to reduce the measured toughness of the specimen. In designing
components for use in load-bearing applications, it is therefore of utmost
importance to ensure that the minimum (plane strain) value is employed.
Solutions enabling the direct determination of Gc exist for the commonly
used specimen geometries such as those presented in Figure 5.3. Williams
(1984) adapted the strain energy analysis for the SENB specimen and
developed the following expression for GIc:
where U is the energy under the load-displacement curve, B the specimen

thickness, D the specimen height and <I> a calibration factor taking account
of the specimen geometry and crack length. The filled circles in Figure 5.4
represent a typical set of experimental data for Araldite B. It is clear that the
level of scatter is low and that the data can be fully characterized using the G
analysis. The slope of the curve gives a value for Gic of 151 J/m2 , a typical
o
1.5
;:;;l 0.5
Io
e Uniform Crack
Non·uniform Crack
I
0.0 .l..-~_-'--~_'--~_L..-~---'_~---'
0.0 0.2 0.4 0.6 0.8 1.0
2 '4
BD <I> (m x 10 )
Figure 5.4 Variation of energy U with geometric factor BD<P for Araldite B single edge notch
bend specimens. The influence of a non-uniform crack front is apparent.
value for a relatively brittle epoxy of this nature. The figure also highlights
one important point, that is that most epoxy resins are well suited to the
application of linear elastic fracture mechanics. However, when
undertaking fracture mechanics tests, it is very important that a straight
uniform pre-crack be obtained. Generally speaking, it is not difficult to
introduce shart pre-cracks into epoxy specimens. Difficulty is often
experienced, however, in assuring a straight and uniform crack front. The
open circles in Figure 5.4 correspond to tests on specimens in which the
crack front was not straight. From the figure it is clear that the non-uniform
cracks result in greater levels of scatter and an artificially high value of GIc-
When pre-cracking epoxy samples with a razor blade and hammer care
should be taken to ensure that the crack propagates as a result of one single
blow rather than a series which may result in a non-straight crack front.
5.2.2 The K approach

The second approach frequently used to characterize the fracture resistance
of epoxy resins is the critical stress intensity factor analysis. Here, the stress
field around a sharp crack in a linear elastic material can be defined by a
stress intensity factor K, (for the case of mode I loading). The general form
of the stress intensity factor is given by:
K, = YoVa
where Y is a geometrical factor taking account of the specimen size, a the
crack length and 0 the applied nominal stress.
It is assumed that fracture occurs when K, reaches a critical value termed
K,o a material property for a given test condition (temperature, rate etc.).
KIc for a finite plate is given as:
KIc = YOc Va
where Oc is the applied stress at the onset of fracture. In order to determine
Klc it is therefore necessary to know the geometrical factor Y for the
particular geometry employed. Such factors have been established for a
wide range of geometries containing cracks and are listed in a number of
handbooks (Rooke and Cartwright, 1976; Sih, 1973; Tada et al., 1973).
If the data presented in Figure 5.4 are analysed using the K approach, a
value for the fracture toughness of 0.64 MPa mO. s is achieved, this again
being a typical value for a relatively brittle epoxy.
For linear elastic fracture mechanics a simple relationship exists between
K and G. For Mode I fracture this relationship takes the form of
G, =
c
Kic
E for plane stress conditions
and
G'e = (~e) (1 - y2) for plane strain conditions
where E is the Young's modulus and y the Poisson's ratio.
5.2.3 Crack opening displacement

Several workers (Gledhill and Kinloch, 1976; Marshall et al., 1974) have
used the crack opening displacement criterion in order to characterize
failure in a number of different thermoplastic and thermosetting polymers.
The tensile stress applied to the material containing the crack causes a
displacement of the two crack surfaces resulting in a crack opening
displacement o. Failure is assumed to occur when the crack opening
displacement reaches a critical value Oe, where:
K2
Oe = _'_e for plane stress
ayE
Under conditions of plane strain, this value is multiplied by the factor
(1- y2).
Quoted values for oe vary from 1 to 20,um depending on the test
temperature and crack velocity (Gledhill et al., 1978).
5.3 Deformation mechanisms
Much work has been undertaken in order to characterize the mechanical

behaviour of epoxy polymers (Lilley and Holloway, 1973; Gledhill and
Kinloch, 1976; Morgan et at., 1979; Morgan, 1980, 1985; Yamini and
Young, 1980; Young, 1980; Narisawa et al., 1982; Bradley, 1989; Smith,
1989; Barbezat, 1990; Bradley et at., 1992; Fischer, 1992). Simple tests using
uni-axial tension, compression and flexure specimens have been used
extensively to quantify the effects of material variables, test temperature,
environmental conditions and loading rate on the mechanisms of deform-
ation and fracture in these materials.
A number of workers have reported evidence for crazing around stress
concentrations in cross-linked epoxies (van den Boogaart, 1966; Lilley and
Holloway, 1973; Morgan and O'Neal, 1977). van den Boogaart (1966)
reported seeing crazes around crack tips in Epikote 828 that was not
completely cured. Subsequent work by Lilley and Holloway (1973) on a
number of fully cured epoxies highlighted craze-like features ahead of the
crack tip. Morgan and O'Neal (1977) undertook a number of in situ fracture
tests on epoxy films in the scanning electron microscope. They concluded
that when loaded in tension, their DGEBA-DETA epoxy deformed and
FRACTURE BERA VIOUR 151
failed through a crazing process. The evidence presented is not convincing
since the supposed craze-like structures are much less well defined than
those observed in thermoplastics. A more recent detailed study by Glad
(1986), again undertaken using thin epoxy films, failed to identify any
evidence for crazing in cross-linked epoxies. He noted that all
thermoplastics with strand densities greater than 8 x 1025 m -3, deformed
through a shear yielding mechanism. The typical strand density of a cross-
linked epoxy is approximately 150 X 1025 m- 3 . He suggested that the
'crazes' observed by Morgan were artefacts resulting from the formation of
voids around dust particles or other defects. It is therefore generally
accepted that crazing mechanisms of the type observed in many thermo-
plastics do not occur in cross-linked epoxy resins. Instead, shear yielding
represents the principal micro-deformation process in these materials.
Figure 5.5a shows shear bands in a hydantoin-based epoxy tensile specimen
tested at 85°C. The shear bands initiated at a very small defect located close
to or at the surface of the sample. The presence of the highly deformed zones
served to redistribute the local stress field and reduce the stress-
concentrating effect of the defect. When a crack did initiate in this region, it
propagated in a slow stable manner as can be seen on the fracture surface
(Figure 5.5b).
The presence of localized shear-yielding in areas of high stress
concentration is therefore beneficial since it results in crack tip blunting, a
re-distribution of the local stress field and an increase in the measured
fracture toughness. Several workers have attempted to measure the local
plastic deformation at the tip of a sharp crack (Glad, 1986; Hibbs and
Bradley, 1987). Optical density measurements by Glad (1986) indicated that
~~r
+.. . . "
K
j
~..
j' '
,
(a) (b)
Figure 5.5 Shear bands in a hydantoin-based epoxy tested in tension at 85°C. (a) Just prior to
crack propagation and (b) after failure (Cantwell et aI., 1988),
=-.;
I- 16
""'
rIO
W"
"
Figure 5.6 Strain field at the tip of a sharp crack in 3501-6 epoxy resin determined by in situ
strain measurements in a scanning electron microscope (from Bradley, 1989).
the maximum extension ratio A. in the damage zone of a moderately cross-

linked DGEBA epoxy could be as high as 1.4. Hibbs and Bradley (1987)
developed a novel technique to quantify the crack tip strain field. Here, the
beam of a scanning electron microscope (SEM) was used to burn a series of
dots around a pre-crack in a compact-tension specimen. The specimen was
then loaded using an SEM in situ stage and the relative movement of the dot
pattern monitored using a continuous image processing system. The local
strain field was then determined by numerically differentiating the
displacement field data. A typical strain field for 3501-6 epoxy resin (a brittle
epoxy for aerospace applications) is shown in Figure 5.6. As would be
expected, the strain concentration immediate to the crack tip is significant.
It is interesting to note, however, that the maximum value of strain is
approximately 15% whereas the elongation at break of this resin when
measured in a standard tensile test is typically only 1.5%. This significant
difference almost certainly relates to the presence of small defects or local
variations in the cross-link density which are clearly going to be more
numerous in a larger volume. The results yielded by simple tensile tests
should therefore be interpreted with caution.
5.4 Modes of crack propagation
The inelastic deformation mechanisms that ·take place at the tip of a sharp
crack in an epoxy resin will clearly have a significant effect upon the manner
in which the crack subsequently propagates. Many workers have made
considerable use of the double torsion (DT) test geometry in an attempt to
characterize both the toughness as well as the mode and stability of crack
propagation in epoxy resins (Young and Beaumont, 1976; Yamini and
Young, 1977; Phillips et al., 1978; Scott et al., 1980). Extensive testing of a
number of pure and rubber-modified epoxies has highlighted three basic
modes of crack propagation, these being:
(i) Stable brittle propagation
(ii) Unstable brittle propagation
(iii) Stable ductile propagation.
5.4.1 Stable brittle propagation

The first of these, stable brittle propagation, results in a flat, featureless
fracture surface and a smooth continuous double-torsion load-displacement
curve (Figure 5.7a). This type of behaviour is thought to be typical of
classical brittle fracture (Kinloch, 1985). Gledhill et al. (1978) have shown
that this mode of failure can be successfully characterized by a unique value
of crack tip opening displacement CTOD. For a DGEBA/9.8 phr TETA
epoxy this value was approximately 0.9 .urn.
(a) DISPLACEMENT
- - INITIATION
- - ARREST
(b) DISPLACEMENT
Figure 5.7 Schematic representations of (a) stable crack and (b) unstable crack growth in a
double-torsion test specimen.
5.4.2 Unstable brittle propagation

Unstable brittle propagation results in the crack advancing in a jump-like
fashion often referred to as stick-slip behaviour (Figure 5. 7b). Since the load
at which fracture initiates and arrests in the DT specimen are not the same,
many workers quote initiation and arrest values of K( G), that is K'i (G'i) and
Kla(G1a)·
The transitions at which crack propagation changes from stable brittle to
unstable brittle or from unstable brittle to stable ductile depend upon a large
number of both material and test parameters (Young and Beaumont, 1976;
Gledhill et al., 1978; Yamini and Young, 1979; Scott et al., 1980; Young,
1980; Cherry and Thomson, 1981). Crack instability in these materials may
occur for a number of reasons. Firstly, certain test geometries are such that
the strain energy release rate increases with crack growth, that is, dG/da is
positive. A second possibility is that the fracture energy of the material
decreases with increasing crack length, i.e. dG,/da is negative. This might
occur, for example, as a result of deleterious environmental effects (Cherry
and Thomson, 1981). Another possibility is that the polymer's fracture
energy decreases with increasing crack velocity yielding a negative value for
dG,/da. Yamini and Young (1977) have suggested that the latter might
explain the tendency for unstable crack propagation in moisture-
conditioned amine-cured Epikote 828.
Many workers have identified the fundamental role of plastic flow and
subsequent crack-tip blunting around stress concentrations in epoxies resins
(Young and Beaumont, 1976; Yamini and Young, 1977; Phillips et al.,
1978). Localized plastic deformation is clearly a fundamental mechanism in
this respect and many of the observations presented in the literature have
been explained at least quantitatively in terms of this plastic flow process. In
general, stable brittle failure occurs at lower temperatures and higher rates
of loading. Under these conditions, the yield stress ofthe material is likely to
be high and the extent of plastic flow and therefore crack blunting low
(Barbezat, 1990). With increasing temperature and decreasing rate, failure
tends to occur in the stick-slip manner associated with unstable brittle
propagation. Here, the yield stress of the material is lower and greater
plastic deformation can take place at the crack tip. The latter therefore
blunts out and takes a more rounded appearance (Scott et al., 1980). When
the crack finally does re-sharpen, the release of energy is significantly
greater than that required to create a· fracture surface and the crack
accelerates, sometimes up to several hundreds of metres per second, and
may eventually bifurcate (Cantwell et al., 1988b).
Williams (1973) and Kinloch and Williams (1980) developed and applied a
crack blunting model to predict the effect of varying certain test parameters
on the mode I fracture toughness of certain epoxy systems. Here, it is
assumed that fracture occurs when a critical stress ate is attained at a certain
distance c ahead of the crack tip. Then, by combining the analysis for the
local crack-tip stress field at a distance, ahead of the crack tip (this being
equal to c):
1 + R.
o = ooya
--
,-------;<";;:-"
YY y(2,) ( 1 + ~, ) 312
where (! is the crack tip radius and 00 is the applied stress with the classical
fracture mechanics equation
it was shown that the critical intensity factor KJc could be related to the value
for a sharp crack KJcs by the relationship:
3/2
( 1 + R. )
2c
1+ R.
c
Figure 5.8 shows the variation of the relative fracture toughness parameter
KrJ KJcs with the square root of the crack tip radius for a series of specimens
containing natural cracks (open circles) and machined blunt cracks (closed
circles) (Kinloch et al., 1983). In the former, the crack tip radius was taken to
be equivalent to the calculated crack tip opening displacement.
12 0 Calculated from CTOD

• Measured
10 •
Theory •
..::'" 8 '-
--....
~
~
u
6 •
•
(J tc =350 MPa
2
c = 0.7 /-lm
0
0 5 10 15 20 25 30
Figure 5.S Variation of KId KIcs as a function of the square root of the notch tip radius, (!, for a
DGEBA resin cured with 5 phr piperidine. The calculated values of the critical stress Ote and
critical distance care 350 MPa and 0.7 microns, respectively (after Kinloch et al., 1984).
This crack blunting analysis implies that a stress of up to five times the
value of the material's yield stress is attained at the crack tip. Kinloch (1985)
suggested that this critical stress, ate> could be interpreted as a constrained
yield stress.
5.4.3 Stable ductile propagation

The third mode of failure, stable ductile propagation yields a smooth load-
displacement DT plot similar to that observed under stable brittle
conditions (Figure 5.7 a). This type of failure is synonymous with high energy
dissipation, a relatively rough fracture surface and, as a result, higher
toughness.
At high temperatures and very low strain rates, cracks frequently advance
through a tearing process in which much energy is dissipated in plastically
deforming the material and creating multiple finger-like markings. Kinloch
et al. (1986) proposed a model for this type of crack growth based on a
meniscus instability analysis. The original model considers the advance of a
meniscus between a less-dense and a more-dense fluid. By developing the
model and applying it to the specific test conditions in question, Kinloch
et al. (1986) calculated a value of 60.um for the critical wavelength of the
finger-like instabilities, this correlating well with the value of 68 ± 24.um
observed on the fracture surfaces.
5.5 Effect of test conditions
Epoxies, like all polymers, exhibit a viscoelastic response when tested under
both short- and long-term loading conditions. Two fundamental test
parameters that highlight this pronounced viscoelastic behaviour are test
temperature and loading rate.
5.5.1 Temperature
Tests undertaken at different temperatures have shown that the mode I
fracture toughness of untoughened epoxies increases, sometimes
dramatically, with increasing test temperature (Bradley et al., 1992). This
increase is clearly related to the deformation process taking place at the
crack tip. Increasing the temperature has the effect of reducing the yield
stress of the polymer which in turn results in a larger plastic zone size and
greater crack blunting. Bradley (1991) suggested that the degree of
temperature-dependency depends upon the relative difference of the test
temperature, T, and the glass transition temperature, Tg, of the polymer in
question. Tests undertaken at temperatures considerably below the Tg of the
polymer should always, therefore, result in brittle failure. By plotting the
fracture toughness, values GIc (or G q if the plain strain requirements were
not satisfied) of his DGEBA epoxies as a function of T - Tg he obtained a
relatively unique curve with all the data falling within a narrow band (Figure
5.9). In all cases, the initial increase in toughness was found to occur at
approximately 125°C below Tg with a dramatic increase occurring at about
60°C below Tg • Unfortunately, no such relationship was obtained for the
rubber-toughened counterparts of these resins.
An interesting and, as yet, unexplained temperature-related effect was
observed by Scott et al. (1980) following low temperature tests on a series of
amine-cured epoxies. Whilst undertaking DT tests in a nitrogen gas
environment they observed that at temperatures below -100°C crack
propagation became unstable with the value of G 1i rising rapidly with further
decreasing temperature. The authors concluded that this change in failure
mode could not be attributed to a simple crack blunting argument but might,
instead, result from increased energy absorption by molecular relaxation as
well as an interaction of the crack tip with the gaseous envionment near its
liquefaction temperature. Thomson (1981) suggested that any frozen water
present on the surface of the specimen might be able to permeate and
plasticize the resin around the crack tip. Young (1980) pointed out that since
N2 has been shown to influence the failure mechanisms occurring in
thermoplastics, it could have a similar effect on epoxy resins. He suggested
that a critical test would be to undertake such tests in a helium gas
environment, a medium which does not have any observable effect on
thermoplastic polymers.
[DJ
4.-----~--~----~----.---~----~
t.
A
C
..
3 0 E
•
•
C-' 1
o
o t. flo
D6 t. 0
OL-____~__~~~_D_t.~~O~O__~__~____~
-300 -200 -100 o
Figure 5.9 Mode I fracture toughness versus T - Tg (T = test temperature, Tg = glass

transition temperature) for three DGEBA epoxies. The Tgs of epoxies A (0), C (ll) and E (0)
are 200, 189 and 225°C respectively. The filled symbols represent lower bound Gq values which
do not satisfy either geometrical or maximum force requirements for plane strain (after
Bradley, 1991).
5.5.2 Loading rate

Reducing the loading rate also has the effect of reducing the yield stress of
the epoxy and therefore increasing the measured toughness (Kinloch, 1985).
The rate sensitivity of epoxies depends upon a number of parameters such as
the test temperature and the cross-link density. At room temperature,
epoxies with a higher degree of cross-linking, and therefore an inherently
stiffer structure, do not exhibit distinct rate-dependent characteristics
(Barbezat, 1990) but low cross-linked epoxies do (Barbezat, 1990).
Gledhill et al. (1979) undertook a series of long-term constant load (static
fatigue) tests on aluminium tapered double cantilever beam (TDCB)
specimens bonded using an amine-cured DGEBA epoxy. During the course
of testing, none of the specimens failed and, indeed, no crack growth was
observed. The test specimens were therefore unloaded after a prescribed
time and tested to rupture. The resulting fracture toughness data indicated
that G\c increased steadily with time under load being almost double its
initial short-term value after 106 seconds (Figure 5.10). The authors noted
that since the compressive yield stress of this material decreased linearly
with increasing logarithmic time-of-test, the formation of a large plastic zone
was easier at longer times resulting in increased crack tip blunting.
Hunston et al. (1984) recognized the inter-relation between rate and
temperature and attempted to express it in a more quantitative manner.
They determined the linear viscoelastic properties for a number of un-
modified and rubber-toughened systems over a wide range of frequencies
and temperatures. These viscoelastic properties were then treated by
applying a time-temperature superposition procedure to the experimental
data. The resulting master curve showed surprisingly little scatter with the
2.2
t;jj
c •
••
:a t;jj
2.0
.'"
~
c
oS :a
~
1.8
•
0:::
oS
~
c
~ .;::= 1.4
.
.s 1.6
• •
~ e •
't:!-= ~1.2
OJ
" ••
...
• •
"
1.0
0.8
0 1 2 3 4 5 6 7
Log (Test time) (s)
Figure 5.10 Static fatigue tests on a tapered double cantilever beam showing the increase in
fracture toughness with loading time (after Gledhill et al., 1979).
data for both the unmodified resins and the toughened systems falling on
what appears to be a unique curve. The shift factors determined from these
viscoelastic analyses were then applied to the fracture toughness data
yielding a GJc master curve.
This type of data reduction technique is potentially very powerful since it
permits the toughness of an epoxy to be predicted for any given ratel
temperature combination.
5.6 Microstructural effects
Most of the fundamental physical and mechanical properties of an epoxy

resin are determined by the nature and density of its cross-linked network.
Although a significant amount of work has been undertaken in an attempt to
relate the mechanical properties of these materials to their microstructure,
a fundamental understanding of the effect of varying basic parameters is
still lacking. Clearly, the molecular architecture of an epoxy will directly
influence its ability to carry load and absorb energy. A detailed discussion
regarding the influence of varying the network structure of an epoxy on its
subsequent mechanical behaviour is beyond the scope of this chapter. For a
more detailed appraisal the reader is directed to more extensive works
(Morgan, 1985; Fischer, 1992).
In general terms, the degree of cross-linking influences the glass transition
temperature of the polymer, its ability to undergo localized plastic
deformation and therefore its toughness characteristics at a given
temperature (Nielsen, 1969). Available experimental evidence suggests that
mechanical properties such as tensile strength, modulus and failure strain do
not exhibit any appreciable cross-link density dependence (Misra et at.,
1979). Epoxies that have high cross-link densities tend to have high Tgs but
are frequently prone to brittle failure. A low cross-link density yields a less
rigid network structure that is capable of undergoing greater amounts of
plastic deformation which in turn may result in a higher fracture toughness.
Bell (1970) as well as Misra et at. (1979) reported significant increases in
impact strength with increasing molecular mass between cross-links (lower
cross-link density). LeMay and Kelley (1986) established a direct
relationship between the square root of the average molecular mass of the
network strands and the fracture energy for crack arrest (Figure 5.11).
Although not plotted, their results also indicated that KJc at initiation
increased with increasing molecular mass of strands. Levita (1989) made
similar observations following fracture mechanics tests on a DGEBA-DDS
epoxy. Here, a 100% increase in fracture toughness was reported over a
range of molecular weights from approximately 300 to 3800 g/mole. The
experimental results also suggested a linear dependence over this range of
molecular weights.
2.8~-~--~--~--~-~--~
IS 2.6
~
....
<Il
<lI
t:
eo:
2.4
~
OJ)
o 2.2
...:l
2.0 L -_ _ ~_---'- _ _~_ _L-_~_ _---.J
~.4 2.8 3.2 3.6

log Me (g/mole)
Figure 5.11 The dependence of the crack arrest fracture energy G, on the average molecular
weight between cross-links Me for a series of stoichiometric DGEBAIDDS epoxies (after
LeMay and Kelley, 1986).
In a detailed study, Glad (1986) showed that the extension ratio, A, within
the plane stress damage zone in thin epoxy films could be directly correlated
to the maximum extension ratio of a single network strand, Amm by the
relationship:
A - 1 = 0.32(Amax - 1)
This appears to be the first time that the deformability of an individual epoxy
network strand has been related to the deformation process at the crack tip.
Fischer (1992) showed that the calculated crack tip opening displacement
could be directly related to the effective molecular mass between cross-
links. He also suggested that the width and length of the deformation zone at
the crack tip are directly proportional to the molecular mass of the network
strands.
Considerable debate has raged in the literature over many years regarding
the existence or non-existence of a very fine nodular structure within epoxy
resins (Racich and Koutsky, 1976; Morgan and O'Neal, 1977; Dusek et ai.,
1978; Mijovic and Koutsky, 1979). It has been suggested that due to the fact
that epoxies are normally made by the addition of a curing agent to a pre-
polymer, a two phase structure could occur (Morgan and O'Neal, 1977).
Mijovic and Koutsky (1979) argue that nodules correspond to high cross-
linked regions ranging in size from 6 nm to 10 Jim. Much of the evidence for
the existence of this fine second phase structure is based on observations
obtained from scanning electron micrographs and replication techniques. If
such a fine structure does indeed exist in these thermosetting polymers, it is
certainly conceivable that they would have a significant influence upon the
mechanical response of the material itself. Indeed, Mijovic and Koutsky
(1979) established a correlation between the morphology of a DETA-cured
DGEBA epoxy and their ultimate mechanical properties. They suggested
that the size of the nodules, resulting from the use of various curing-agent
concentrations, can be quantitatively related to the amount of plastic flow
occurring during crack initiation. Dusek et al. (1978) etched the fracture
surfaces of an amine-cured DGEBA and revealed globular structures some
20 to 40 nm in size. Similar structures were observed, however, following
etching of amorphous polystyrene and PMMA. Further, the small angle
X-ray scattering curves for the cured epoxy did not differ from those of
common amorphous polymers. The authors concluded that nodules do not
exist in simple amine-cured epoxies of this sort. They did point out,
however, that with more complicated systems a more pronounced
inhomogeneity might be caused by thermodynamic incompatibility or by
non-alternating mechanisms of the curing reaction. It is also possible that
the presence of such a secondary phase may be due to incomplete mixing of
the hardener and resin during manufacturing (Ellis, 1992).
5.7 Fractography of epoxy resins
Considerable information concerning the cause of fracture as well as the

damage mechanisms occurring during the failure process can be obtained
from a detailed examination of the fracture surfaces. Although a significant
number of articles have addressed this problem, the wide range of terms and
expressions currently employed to describe fracture surface characteristics
still result in considerable confusion and a lack of a detailed understanding.
Even a cursory examination of almost any epoxy fracture surface
highlights the presence of a number of distinct regions exhibiting different
features and textures. One of the most effective ways of understanding and
explaining the formation of such zones is to systematically vary test
parameters such as temperature and rate whilst measuring the variation of
crack velocity during failure. A number of techniques are available for
measuring the velocity of a crack in a brittle polymer, possibly the best of
these being the graphite gauge technique developed by Stalder (1985). Here
a very thin layer of graphite is sprayed around the anticipated crack
trajectory. Following this, two silver electrodes are painted on the specimen
parallel to the crack path. Now, since a linear relationship exists between the
relative crack length and the relative conductance, the determination of the
variation of potential with time permits the crack velocity to be measured at
any moment during fracture. This technique has been successfully applied to
double-torsion (Stalder, 1985), single edge notch (Cantwell et ai., 1990),
three point bend (Stalder, 1985) and compact tension test geometries
(Cudre-Maroux, 1988).
In Figure 5.12 two single edge notch fracture surfaces are shown along
with the associated crack velocity profiles . Figure 5.12a shows a smooth
fracture surface with few apparent markings. The corresponding crack
velocity profile shows that the crack accelerated to 230 mls and then
stabilized. The second fracture surface, Figure 5.12b, is much rougher in
appearance due to extensive crack bifurcation. The velocity profile again
shows a region of crack acceleration, however, in this case the crack reaches
430 mis, the limiting velocity in this material. Crack branching and therefore
a very rough three dimensional fracture surface occurs only when the crack
a
i
500
i>- 400
~b
I-
300
U
0
...J
fa
...
\iOI 200
':Ie
U 100
«
1:11:
u
5 10 IS 20 25 30 J5
0 1 TA CE (mm)
b
t
Figure 5.12 Fracture surfaces of single edge notch Araldite B specimens showing the variation
of crack velocity during failure . Top: a smooth fracture surface; bottom: a rougher surface due
to extensive crack bifurcation.
has reached the maximum velocity (determined by the sonic velocity of the
material). Since any excess energy within the specimen cannot be dissipated
by further accelerating the crack, the crack must bifurcate creating multiple
fracture surfaces and greater energy dissipation.
Under certain conditions, other features may be prominent on the
fracture surface of fracture mechanics-type specimens. One example is
parabolic markings similar to those presented in Figure 5.13. Parabolic
markings are associated with the initiation of secondary cracks in front of the
primary crack in the region of fast unstable crack propagation. The precise
shape and form of such markings depend upon the relative velocities of the
primary and secondary crack fronts. If both fracture fronts move with the
same velocity a hyberbola is formed whereas an ellipse is created if the
primary crack propagates more rapidly than the secondary (Doll, 1989). In
this case, the distance between the vertices of the markings is between 50
and 100 microns. Following impact tests on notched PMMA samples, Van
Noort and Ellis (1984) observed a lower density of markings with the
average distance between vertices being approximately 200 microns.
The above observations have been used to explain certain macroscopic
features on the fracture surfaces of unmodified and particulate-filled
epoxies (Cantwell et al., 1990). Figure 5.14 shows the fracture surfaces of a
hydantoin-based epoxy tested at a number of temperatures between 23 and
105°C (Tg of the resin = 125°C). A detailed examination of the failed
surfaces highlighted four zones: a defect (not always in evidence), a smooth
mirror-like region, a slightly less smooth area and a rough three-dimensional
zone. The mirror-like zone was found to correspond to the region of slow
crack growth where the crack advanced at a speed of the order of several
millimetres per minute. Fine lines oriented parallel to the direction of crack
propagation are sometimes observed in this zone. These lines or 'river-
markings' occur as a result of crack propagation on two slightly different
Figure 5.13 Parabolic markings on the fracture surface of a DGEBA epoxy (photo courtesy of
Lu Fan).
(a)
(c)
Figure 5.14 Tensile fracture surfaces of a hydantoin-based epoxy resin (type XB 2900) cured
with a formulated anhydride hardener (HY 925) as a function of test temperature. Four regions
have been identified: (1) a defect (not always in evidence) (2) a slow growth mirror-like region
(3) a smooth region where the crack accelerates in an unstable manner and (4) a rough three
dimensional area. (a) 23°C; (b) 50°C; (c) 85°C; (d) lO5°C.
planes. The intersection of these two planes creates a welt frequently of the
order of tens of microns in diameter. River markings are useful in as much as
they indicate the direction of slow crack propagation. These lines can then
be traced back enabling the defect or cause of failure to be identified.
Beyond the mirror region extends another flat zone exhibiting parabolic
markings similar to those reported above. It was established that this region
corresponded to the zone over which the crack was accelerating in an
unstable manner. Once the crack achieved its limiting velocity (typically 400
to 500 mls in an epoxy tested at room temperature) and was therefore no
longer capable of accelerating, it bifurcated creating multiple fracture
surfaces and a rough fracture zone. The velocity of sound within this
material is approximately 1650 mls suggesting that the limiting crack
velocity is some 25-30% of this value. The rough zone was not observed in
the specimens tested at 85 and 105°e suggesting that the crack had either
stabilized at a velocity below the limiting value or else that the required
acceleration zone was larger than the specimen diameter.
5.8 Toughening strategies for epoxy resins
The data presented in Figure 5.1 indicate that a large number of epoxies
offering an equally wide range of toughnesses are presently available.
Unfortunately, almost all of the epoxies for use in high temperature load-
bearing applications are brittle offering values of GIc of 300 J/m 2 or less. In
recent years, a considerable amount of work has been undertaken in an
attempt to enhance the toughness of these materials. These studies have
tended to concentrate on the addition of a second phase such as rubber
particles, thermoplastic particles or mineral fillers such as silica flour.
5.S.1 Mineral filler-modified epoxies

Work by Moloney et al. (1984) and Spanoudakis and Young (1984) has
shown that the modulus and toughness of an epoxy can be increased
significantly through the introduction of small glass spheres. The effect
of the incorporation of solid fillers on mechanical properties of epoxy resins
is summarized in Table 5.1 (Moloney et al., 1987). Clearly, the degree
of enhancement in mechanical properties depends upon a number of
parameters of which the volume fraction of filler as well as the surface
treatment are amongst the most important. Typically, the addition of 40%
by volume of glass beads results in a doubling of the Young's modulus and a
400% increase in KIc whilst incurring little or no loss in tensile strength
(Moloney et al., 1984; Spanoudakis and Young, 1984). Similar observations
have been made following tests on silica-modified epoxy resins (Cantwell
et al., 1990). Some controversy still exists regarding the precise mechanisms
responsible for the observed improvements in toughness in these materials.
Several authors have suggested that the presence of the rigid particles serves
to pin the advancing crack forcing it to bow out and hence increase the
fracture energy (Spanoudakis and Young, 1984; Moloney et al., 1983).
Table 5.1 Parameters detelmining the modulus, toughness and strength of

particulate filled epoxies (from Moloney et al., 1987)
Property Effect on:
Modulus Kc Strength
Increase volume filler Increase Increase - Constant at

matrix value
Particle size Constant Constant Decrease
Increase aspect ratio Increase Increase
Increase strength and Increase Increase Increase
modulus filler
Improved adhesion Constant Constant Increase
Tougher matrix Small decrease Increase Decrease
Evans et al. (1985) suggested that substantial debonding of the second phase
glass particles may occur around the crack tip creating microcracks leading
to a reduction in the magnitude of the stresses in the process zone. A model
based on an idealized constitutive law was developed which predicted the
effect of filler volume fraction on toughness with some success. More recent
work (Smith, 1989) has suggested that the presence of these solid particles
results in a significant stress concentration in the surrounding matrix. At
high loads, the local stress exceeds the yield stress of the matrix resulting in
the formation of localized shear bands similar to those presented in Figure
5.15. The presence of such shear bands serves to blunt the propagating crack
and thereby enhance the measured toughness. Under conditions of extreme
crack tip ductility, such as at high temperatures and low strain rates,
extensive particle-matrix debonding may occur. Under these circumstances,
the presence of such a myriad of micro-cracks in a zone immediate to the
primary crack reduces the inherent toughness of the material resulting in a
significant drop in KIc (Cantwell et al., 1990).
Low (1990) showed that residual stresses playa significant role in the
failure process of filled epoxy systems. These residual stresses arise from the
thermal expansion mismatch between the epoxy and the dispersed phase.
He showed that compressive radial and tensile tangential stresses are
generated at the interface when the filler has a lower thermal expansion
coefficient that the matrix. It was indicated that the presence of such stresses
tends to enhance interface debonding and encourages catastrophic unstable
fracture resulting in a flat fracture surface.
5.8.2 Thermoplastic-modified epoxies

A number of workers have examined the feasibility of toughening epoxies
using a thermoplastic second phase such as poly-ether-imide, poly-ether-
Figure 5.15 Shear bands in a 16% silica-filled epoxy resin tested in tension at 85°C (courtesy of
Josh Smith).
sulphone, polysulphone and poly-phenylene-oxide (Hedrick et al., 1985;

Bucknall and Partridge, 1986; Pearson and Yee, 1991). The advantage of
thermoplastic modifiers such as those reported above is that their
incorporation into epoxy resins does not result in the reduction in modulus,
reduced high temperature properties or poorer creep resistance associated
with their rubber-modified counterparts. The early studies resulted in
relatively small increases in toughness (Bucknall and Partridge, 1983;
Hedrick et al., 1985). More recent work by Pearson and Yee (1991) using a
number of different thermoplastic modifiers to toughen a low Tg epoxy
resulted in some cases in a toughness enhancement close to that achieved
using CTBN (Figure 5.16). Thin sections removed from the necked region
of a series of tensile specimens exhibited signs of extensive shear banding
when viewed under polarized light. Surprisingly however, sections removed
from the crack-tip regions of mechanics-type fracture specimens did not
indicate the presence of this form of localized inelastic deformation.
Instead, all of the samples showed signs of profuse microcracking, the exact
nature of which depended upon the type of modifier used. The authors
suggested that the presence of this damage zone at the crack tip served to
shield the main crack from the applied stress intensity factor thus improving
the measured toughness.
5.8.3 Rubber-modified epoxies

Extensive work has shown that the fracture toughness of epoxy resins can be
significantly enhanced through the incorporation of a second rubbery phase.
Although rubber has a low shear modulus, its bulk modulus is comparable to
2.5
...
!::'
eoi 2.0
0
Q.,
6 0
.'"'" 1.5
0
• •
...=
.c •
= 0
•
....
0
E-<
1.0 o CTBN
.a • PEI-PDMS
.....
Col
~
• PPO
0.5
0 5 10 15 20
Modifier Content (phr)

Figure 5.16 Toughness enhancement of an epoxy resin using thermoplastic particles and
CTBN. The epoxy used for the CTBN-modified system was DER331 (a DGEBA from the Dow
Chemical Co.) and that for the thermoplastic-modified materials was DER332 (also a DGEBA
from the Dow Chemical Co.) (after Pearson and Yee, 1991).
that of an epoxy and rubber inclusions are therefore capable of bearing

significant load within the material. Many workers have attempted to
identify the precise mechanisms responsible for the significant toughness-
enhancement associated with rubber modification. Kunz-Douglass et al.
(1980) suggested that rubber particles enhanced toughness by bridging the
crack and applying surface tractions which effectively reduce the stress
intensity factor at the crack tip. Unfortunately, the rubber bridging model
fails to predict the effect of varying test temperature. Since the tearing
energy of rubber increases as the temperature is reduced below room
temperature, the fracture toughness of rubber-modified epoxies should also
increase with decreasing temperature. However, the available experimental
evidence shows that this is not the case. Rubber bridging may indeed
contribute to the energy-absorbing process but it is unlikely to be the
principal mechanism.
Evans et al. (1986) attempted to quantify the energy dissipated in plastic
dilation of the matrix following debonding of the rubber particle. They
concluded that this mechanism may be important when it occurs local to a
stress concentration or crack. Kinloch (1987) suggested that two important
deformation processes occur in rubber-toughened epoxies. Firstly, the
presence of the rubber particle creates a stress concentration which acts as a
site for plastic shear deformation. It is not clear whether these shear bands
initiate at the equator of the particle where the tensile stress concentration
is a maximum or nearer the pole where the maximum shear stress concen-
tration occurs. Since there are a large number of these stress-concentrating
particles within the volume of the material, considerable energy dissipation
occurs. Secondly, the tri-axial stress state existing at the tip of the crack
causes the particles to fail and cavitate creating voids within the bulk of the
material. Under continued loading, these voids grow and dissipate energy.
Further, the cavitation process results in a reduction extent of stress-
tri-axiality at the crack tip which in turn leads to a reduction in the yield
stress and therefore greater yielding in areas of severe stress concentration.
Chen and Jan (1991) used a number of epoxy end-capped CTBN rubbers
to influence the cavitation process in a DGEBA epoxy. They showed that
end-capping CTBN in this way resulted in large cavitated zones and a
considerably higher fracture toughness.
Huang and Kinloch (1992) attempted to quantify the relative contri-
butions of particle bridging, shear yielding and void growth as a function
of temperature in a piperidine-cured DGEBA. Their analysis indicated
that, at low temperature, shear yielding represents the main energy-
absorbing mechanism whereas void growth and yielding are equally
important at elevated temperatures (Figure 5.17). It is interesting to note
that rubber bridging makes an appreciable contribution to the toughness of
the polymer at sub-zero temperatures.
The degree of toughness enhancement appears to depend strongly upon
the epoxy's ability to undergo plastic deformation and, therefore, upon its
Shear yielding
0.6
I
~
<l 0.4
_-___. "- Voiding
0.2 Rubber bridging
0.0 .J......~____~_.-~--'-~----J'--~-'-~--.L_"----'
-80 -60 -40 -20 o 20 40 60

o
Temperature ( C)
Figure 5.17 The relative contributions of shear yielding, rubber bridging and void growth as a
function of temperature in a CTBN-modified piperidine-cured DGEBA (data taken from
Huang and Kinloch, 1992).
microstructure. An early study by Meeks (1974) showed that the extent to

which toughening can be induced in a rubber-modified epoxy depends upon
the curing agent used. Anhydrides were shown to be the most useful in this
respect with only modest toughening being observed with many of the other
curing agents examined. By employing different cure schedules, Kinloch et
al. (1987) varied the degree of cross-linking in a CTBN-modified DGEBA
resin. Although the addition of rubber particles resulted in significant
increases in G re for both the low and high molecular weight resins, the
relative increase was greatest in the latter. A more detailed study by Pearson
and Yee (1989) clearly showed that the degree of toughness enhancement
depends strongly upon the monomer molecular weight of the resin
employed. For low values of molecular weight the increase in GIc was less
than 100% whereas at higher values a thirty-five fold increase was reported.
These observations were supported by scanning electron micrographs of the
fracture surfaces which showed little plastic deformation in the high cross-
link density material but extensive plastic flow and void enlargement in
lower cross-linked epoxy. Levita (1989) also identified a strong resin
molecular weight-dependency in CTBN-modified DGEBA-DDS (Figure
5.18), although his results suggested that KIc tends to level off at high
molecular weights.
5.8.4 Effect of particle size and volume fraction

The aforementioned studies as well as that of Bradley et al. (1992), have
highlighted the fact that high Tg epoxies such as those destined for use in the
I~
Neat Resin
10% CTBN
I
~
4 • •
~~ •
~ 3
~
•
'-'
~
2 0
0
~
0
1 •
0
0
0
0 1000 2000 3000 4000
Resin Molecular Weight
Figure 5.18 Fracture toughness vs resin molecular weight for neat and CTBN-modified
DGEBA-DDS (after Levita, 1989).
aerospace industry often prove extremely difficult to toughen. The highly

cross-linked nature of many of these epoxies limits their ability to undergo
significant localized plastic deformation and therefore respond positively to
rubber-toughening. Bradley et al. (1992) pointed out that the interparticle
distance may well be greater than the size of the process zone at the crack tip.
Consequently, the rubber inclusions might not interact with the crack
process zone and under such circumstances their effect would be only
marginal. The interparticle distance can be estimated from a simple two-
dimensional analysis:
d = r [( 2;rc )112 _ ~]
P P 3VF 2
where rp is particle radius, and VF the volume fraction of particles. This

distance can therefore be decreased by increasing V F for a given particle size
or decreasing rp for a given volume fraction. Unfortunately, increasing the
volume fraction of rubber particles often results in an increase in particle
diameter. Therefore, identifying and separating the effects of rp and VF are
often difficult. In an early study, Sultan and McGarry (1973) suggested that
particles around one micron in size encouraged crazing in epoxy resins
whereas those considerably smaller than this value provoked shear yielding.
It is generally accepted, however, that crazing does not occur in these cross-
linked thermosets. Kunz-Douglass and co-workers (1980) reported that
large rubber particles were not as efficient toughness-enhancers as smaller
one micron inclusions. Hwang et al. (1989) claimed that sub-micron particles
are less effective tougheners than those in the 1 to 5 micron range. The work
of Bascom et al. (1981) suggested that a bimodal distribution of particles
was more effective than a unimodal phase. More recently, Pearson (1990)
showed that 0.2 micron particles were much more effective for toughening a
DGEBA epoxy than either 2 or 50 micron particles. He suggested that large
particles merely act as bridging particles which provide only a modest
increase in toughness whereas small particles within the crack-tip process
zone cavitate, relieving the plane strain constraint thereby facilitating
yielding.
Kinloch and Hunston (1986) showed that the influence of volume fraction
on Grc is complex and depends upon the particular test conditions. By
applying a time-temperature superposition analysis similar to that reported
above (section 5.4.2), they showed that at low temperatures varying the
volume fraction of rubber particles in a piperidine-cured DGEBA epoxy has
only a slight effect on toughness. At higher temperatures, however,
doubling the volume fraction of particles had a similar effect on G1c ' They
concluded that the relationship between Grc and V F depends upon the ability
of the epoxy to respond to the presence of the rubber particles by
undergoing shear-yielding.
5.B.5 Hybrid systems

A number of workers have examined the feasibility of adding both rigid and
soft fillers to epoxy resins in order to achieve a significant enhancement in
both modulus and fracture toughness (Young et at., 1986; Smith, 1989).
Smith (1989) showed that the Krc of a CTBN-silica hybrid epoxy was up to
50% higher than the standard silica-filled system. The elastic moduli of the
hybrid and standard systems were, however, similar. Examination of the
crack-tip region (Figure 5.19) indicated that damage took the form of shear
yielding, deb on ding and cavitation, i.e. the deformation and failure
mechanisms observed in the standard CTBN and silica-modified materials.
Interestingly, rubber particle cavitation occurred preferentially at the poles
Figure 5.19 The crack tip region of a silicalCTBN-modified epoxy resin tested at 105°C (photo
courtesy of Josh Smith).
of the silica particles. These observations support the numerical analysis of

Guild and Young (1989) sho showed that the maximum stress concentration
in a rigid filler-modified epoxy occurs at the poles of the inclusion.
Clearly, considerable progress has been made in the area of epoxy
toughening and fracture suppression. More recently, a number of workers
have been assessing the potential of core-shell modifiers consisting of a hard
thermoplastic shell and a soft rubber core. The particles are generally
sufficiently small and closely spaced to interact with the crack tip plastic
zone, yielding appreciable toughness enhancement even in notoriously
brittle high Tg systems.
5.9 Conclusions
Epoxy resins have proven themselves to be excellent engineering polymers

finding extensive use in areas such as coatings, adhesives and matrices for
aerospace composites. A large number of research studies have enabled the
effects of varying both basic material parameters and operating conditions
to be studied and characterized. The mechanisms by which these materials
deform and fail are now well understood. However, more work is required
before a full understanding of the influence of the network structure on
many of the basic mechanical properties is achieved. The inherent brittle
nature of these thermosetting polymers has been suppressed by
incorporating a uniformly-dispersed second phase such as rubber particles,
thermoplastic spheres or mineral fillers.
Acknowlegement
WJC is grateful to the Fonds National Suisse for financing a sabbatical at the Czechoslovakian
Academy of Sciences, Prague, during which time part of the text was prepared.
References
Barbezat, M. (1990) PhD Thesis, Ecole Poly technique Federale de Lausanne.

Bascom, W.D., Ting, R.Y., Moulton, R.J., Riew, C.K. and Siebert, AR. (1981)J. Materials
Sci. 16,2657-2664.
Bell, J.P. (1970) J. Appl. Poly. Sci. 14,1901-1906.
Bradley, W.L. (1989) In Thermoplastic Composite Materials, ed. Carlsson L.A Elsevier,
Applied Sci. Pub.
Bradley, W.L. (1991) Proc. of the 8th Int. Conf. on Deformation, Yield and Fracture of
Polymers, PRI paper 36.
Bradley, W.L., Schultz, W., Corleto, C. and Komatsu, R. (1992) to be published in Rubber-
Toughened Plastics, ed. Riew, C.K.
Bucknall, c.B. and Partridge, I.K. (1983) Polymer 24,639-644.
Bucknall, C.B. and Partridge, I.K. (1986) Polymer Eng. and Sci. 26,54-60.
Cantwell, W.J., Roulin-Moloney, AC. and Kaiser, T. (1988a)J. Materials Sci. 23, 1615-1631.
Cantwell, W.J., Roulin-Moloney, A.C. and Kausch, H.H. (1988b) J. Materials Sci. Letters
7,976-980.
Cantwell, W.J., Smith, J.W., Kausch, H.H. and Kaiser, T. (1990)J. Materials Sci. 25,633-648.
Chen, T.K. and Jan, Y.H. (1991) J. Materials Sci. 26,5848-5858.
Cherry, B.W. and Thomson, K.W. (1981) J. Materials Sci. 16, 1913-1924.
Cudre-Maroux, N. (1988) PhD Thesis, Ecole Polytechnique Federale de Lausanne.
Doll, W. (1989) In Fractography and Failure Mechanisms of Polymers and Composites, Ch. 10,
ed. Roulin-Moloney, A.C. Elsevier Applied Science.
Dusek, K., Plestil, J., Lednicky, F. and Lunak, S. (1978) Polymer 19, 393-397.
Ellis, B. (1992) Private communication.
Evans, A.G., Williams, S. and Beaumont, P.W.R. (1985) J. Materials Sci. 20,3668-3674.
Evans, A.G., Ahmad, Z.B., Gilbert, D.G. and Beaumont, P.W.R. (1986) Acta Metall. 34,
79-87.
Feuerhahn, M. (1986) Kunstoffe 76,872-873.
Fischer, M. (1992) In Advances in Polymer Science 100, Springer-Verlag, Berlin, pp. 313-355.
Glad, M.D. (1986) PhD Thesis, Cornell University.
Gledhill, RA. and Kinloch, A.J. (1976) Polymer 17, 727-731.
Gledhill, R.A., Kinloch, A.J., Yamini, S. and Young, RJ. (1978) Polymer 19, 574-582.
Gledhill, R.A., Kinloch, A.J. and Shaw, S.J. (1979) J. Materials Sci. Letters 14,1769-1772.
Griffith, A.A. (1920) Phil. Trans. Roy Soc. A221, 163-198.
Guild, F.J. and Young, R.J. (1989) J. Materials Sci. 24,298-306.
Hedrick, J.L., Yilgor, Wilkes, G.L. and McGrath, J.E. (1985) Polymer Bulletin 13, 201-208.
Hibbs, M. and Bradley, W.L. (1987) Proc. Soc. Exp. Mechanics, Fall Meeting, Georgia.
Huang, Y. and Kinloch, A.J. (1992) J. Materials Sci. Letters 11, 484-489.
Hunston, D.L., Kinloch, A.J., Shaw, S.J. and Wang, S.S. (1984) in Adhesive Joints ed. Mittal,
K.L., Plenum Press, pp. 789-807.
Hwang, J.F., Manson, J.A., Hertzberg, R.W., Miller, G.A. and Sperling, L.H. (1989) Poly.
Eng. Sci. 29, 1466-1476.
Kinloch, A.J. (1985) In Advances in Polymer Science 72, ed. Dusek, K. Springer-Verlag,
Berlin, pp. 46-67.
Kinloch, A.J. (1987) In Rubber-Toughened Plastics ed. Riew, C.K. Advances in Chemistry
Series 222 American Chern. Soc. 1989 Ch. 3.
Kinloch, A.J. and Hunston, D.L. (1986) Polymer 5,1207-1209.
Kinloch, A.J. and Williams, J.G. (1980) J. Materials Sci., 15,987-996.
Kinloch, A.J., Shaw, S.J. and Hunston, D.L. (1983) Polymer 24,1355-1363.
Kinloch, A.J., Gilbert, D.G. and Shaw, S.J. (1986) J. Materials Sci. 21, 1051-1056.
Kinloch, A.J., Finch, C.A. and Hashemi, S. (1987) Polymer Comm. 28, 322-325.
Kunz-Douglass, S., Beaumont, P.W.R. and Ashby, M.F. (1980) J. Materials Sci. 15, 1109-
1123.
LeMay, J.D. and Kelley, F.N. (1986) In Adv. in Polymer Science, 78 ed. Dusek, K. pp. 116-
148.
Levita, G. (1989) In Rubber- Toughened Plastics ed. Riew, C.K. Advances in Chemistry Series
222 American Chern. Soc. 1989 Ch. 4.
Lilley, J. and Holloway, D.G. (1973) Phil. Mag. 8,215-220.
Low, I.M. (1990) J. Materials Sci. 25, 2144-2148.
Marshall, G.P., Coutts, L.H. and Williams, J.G. (1974) J. Materials Sci. 9, 1409-1419.
Meeks, A.C. (1974) Polymer 15, 675-681.
Mijovic, J.S. and Koutsky, J.A. (1979)1. Appl. Polym. Sci. 23, 1037-1042.
Misra, S.C., Manson, J.A. and Sperling, L.H. (1979) ACS Symp. Series 114,137-156.
Moloney, A.C., Kausch, H.H., Kaiser, T. and Beer, H.R. (1987)J. Materials Sci. 22,381-393.
Moloney, A.c., Kausch, H.H. and Stieger, H.R. (1983) J. Materials Sci. 18,208-216.
Moloney, A.C., Kausch, H.H. and Stieger, H.R. (1984) J. Materials Sci. 19, 1125-1130.
Morgan, R.J. (1980) In Developments in Reinforced Plastics 1, ed. Pritchard, G. App. Sci.
Pub., Ch. 7.
Morgan, RJ. (1985) In Adv. in Polym. Sci. 72, ed. Dusek, K. Springer-Verlag, Berlin,
pp.3-43.
Morgan, RJ. and O'Neal, J. (1977) J. Materials Sci. 12,1966-1980.
Morgan, RJ., O'Neal, J. and Miller, D.B. (1979) J. Materials Sci. 14,109-124.
Narisawa, I., Murayama, T. and Ogawa, H. (1982) Polymer 23, 291-294.
Nielsen, L.E. (1969) J. Macromol. Sci.-Revs Macromol. Chem., C3(1), 69-103.

Pearson, R.A. (1990) PhD Thesis, Univ. of Michigan.
Pearson, R.A. and Yee, A.F. (1989) J. Materials Sci. 24,2571-2580.
Pearson, R.A. and Yee, A.F. (1991) Proc. of the 8th Int. Conf. on Deformation, Yield and
Fracture of Polymers, PRI paper 40.
Phillips, D.C., Scott, J.M. and Jones, M. (1978) J. of Materials Sci. 13,311-322.
Racich, J. and Koutsky, J. (1976) J. of Appl. Polym. Sci. 20,2111-2129.
Rooke, D.P. and Cartwright, D.J. (1976) Compendium of Stress Intensity Factors, HMSO,
London.
Scott, J.M., Wells, G.M. and Phillips, D.C. (1980) I. Materials Sci. 15,1436-1448.
Sih, G.c. (1973) Handbook of Stress-Intensity Factors for Researchers and Engineers, Lehigh
Univ.
Smith, J.W. (1989) PhD Thesis, Ecole Poly technique Federale de Lausanne.
Spanoudakis, J. and Young, R.J. (1984) J. Materials Sci. 19,473-486.
Stalder, B. (1985) PhD Thesis, Ecole Polytechnique Federale de Lausanne.
Sultan, J.N. and McGarry, F.J. (1973) Poly. Eng. and Sci. 13,29-34.
Tada, H., Paris, P. and Irwin, G.R. (1973) The Stress Analysis of Cracks Handbook, Del
Research Corp., Hellertown.
Thomson, K.W. (1981) J. Materials Sci. Letters 16, 1435-1436.
van den Boogaart, A. (1966) Inst. of Physics Conf. Proc. London, pp. 167-175.
Van Noort, R. and Ellis, B. (1984) I. Materials Sci. Letters 3,1031-1034.
Williams, J.G. (1973) Stress Analysis of Polymers, Longmans, London.
Williams, J.G. (1984) Fracture Mechanics of Polymers, Ellis Horwood Ltd, Chichester.
Yamini, S. and Young, R.J. (1977) Polymer 18,1075-1080.
Yamini, S. and Young, R.J. (1979) J. Materials Sci. 14,1609-1618.
Yamini, S. and Young, R.J. (1980) J. of Materials Sci. 15,1823-1831.
Young, R.J. (1980) In Developments in Reinforced Plastics-l, ed. Pritchard, G. App. Sci. Pub.
Ch.9.
Young, R.J. and Beaumont, P.W.R. (1976) J. Materials Sci. Letters 11, 778-779.
Young, R.J., Maxwell, D.L. and Kinloch, A.J. (1986) J. Materials Sci. 21,380-388.
6 Electrical properties of epoxy resins
G.P. lOHARI
6.1 Introduction
Epoxy resins are widely used for coating and encapsulating electrical circuit
components and electronic devices, where they serve to isolate a device
from adverse environmental effects of atmospheric gases, moisture, current
leakage, solvents, micro-organisms, mechanical shock and vibrations. In a
more recent application, known as micro-electronics packaging, they
provide the required insulation between closely-packed, fragile electrical
elements with good adhesion to both the substrate and the elements. It is this
application that now requires that the changes in the electrical properties of
epoxies during their curing and post-curing be understood at a fundamental
level, and that any effects of ageing on these properties be studied in enough
detail that procedures for eliminating these effects may be developed.
Although dielectric measurements for studying the progress of
polymerization were made as early as 1934 by Kienle and Race (1934), who
investigated polyesterification reactions, extensive literature on the subject
has appeared only since 1968. An excellent review of these studies up to 1986
has been given by Senturia and Sheppard (1986). For the reader's
convenience they have included, in this review, the details of measurement
methods and a comprehensive bibliography with 118 references on the
dielectric studies of various pure epoxies and their composites.
Since 1986, studies focused on dielectric monitoring of the progress of
chemical reactions during the processing of various epoxies have been made
by several groups (Day, 1986; Lane and Seferis, 1986; Bidstrup et aI., 1989;
Gatro and Yandrasits, 1989; Lane etal., 1989; Sheppard and Senturia, 1989;
Mangion, 1990; Kranbuehl, 1991; Parthun and Johari, 1992a; Alig and
Johari, 1992; Tombari and Johari, 1992). They have related the change in
the electrical properties to the increase in the viscosity of the epoxy resin,
thereby deducing that the increase in the glass transition temperature with
time allows modelling of the isothermal curing process - a subject of
considerable technical importance - by the Vogel-Fulcher-Tamman
equation. A collection of papers related to these observations has been
published in a special issue of Polymer Engineering and Science (vol. 29,
no. 5, 1989).
A somewhat different but more general treatment is based upon the
premise that in all addition reactions the growth in the size of a macro-
molecular addition product slows the rate of the very chemical reactions by
which it grows, and that this continues until neither chemical reactions nor
molecular diffusion occur at a perceptible rate. In this way, Johari and
coworkers (Johari and Mangion, 1991; Parthun and Johari, 1992a) prefer to
envisage the curing of an epoxy resin as a negative feedback process between
chemical reaction~-- and molecular diffusion, which ultimately and
irre¥ersibly vitrifies or solidifies a liquid. Thermodynamic evidence for this
vitrification is now available from the careful heat capacity measurements by
Cassettari et al. (1992); there is an irreversible, abrupt decrease in the heat
capacity of an epoxy resin over a narrow range of time, which remarkably
resembles the (reversible) decrease in heat capacity observed on
supercooling a liquid and rubbery polymer through their glass transition
temperature, Tg • The time at which this decrease occurs becomes less when
the reaction occurs at a higher temperature.
This chapter reviews the developments, since 1986, in our knowledge of
both the dielectric effects during the curing process of epoxy resins and the
electrical properties of their cured state. Much recent progress has come
from work on amine-cured epoxies, but many of the principles that have
been discovered relate equally to other reactions of linear chain and network
polymerizations, a subject of our continuing studies.
This chapter is organized such as to provide first (section 6.2) general
features of changes in the physical properties during the curing of a
thermoset polymer, in particular an epoxy resin. This is followed in section
6.3 by the description of a general formalism for the electrical properties of
an epoxy resin as they change with the curing time, temperature and
frequency used for the measurement. Section 6.4 provides the changes in the
dielectric permittivity, and section 6.5 those in the ionic conduction during
the process of sol-gel-glass conversion. The time and temperature evolution
of the dielectric properties is given in section 6.6 which is followed by the
chemical kinetics and dielectric behaviour in section 6.7. Sections 6.8 and
6.9 provide a description of the sub-Tg relaxations in partially-cured and
fully-cured epoxy resins and the effects of post-cure and physical ageing on
their electrical properties. A summary of the electrical applications of epoxy
resins is given in section 6.10.
6.2 Physical changes during the epoxy curing
A variety of mechanical relaxation measurements on thin wire braids

impregnated with epoxies, carried out by Gillham and coworkers (for a
review see Aronhime and Gillham, 1986), and of viscoelasticity
measurements on pure epoxies by Choy and Plazek (1986) have shown a
general pattern of physical changes that occurs during the curing of an
epoxy. As an epoxy group opens to form one OR group and one chemical
bond with another molecule during the curing process, the viscosity
ELECTRICAL PROPERTIES 177
increases in a manner such that the original liquid or sol state is converted to
the gel state of an infinitely-connected network structure with mechanical
modulus of less than 106 Pa. Here the viscosity approaches a formally
infinite value, but localized molecular motion and diffusion of unreacted
component persist in the network structure. As the curing progresses, the
cross-links density increases with further reactions, the cross-linked network
structure densifies and its mechanical modulus approaches that of a vitrified
or glassy solid, i.e. about 1010 Pa. Gillham and coworkers (Enns and
Gillham, 1983) have deduced the transformation-time-temperature
diagrams of several epoxy resins by using such results obtained from their
torsional-braid analysis.
Once an epoxy resin has vitrified, chemical reactions continue to occur,
albeit at a much slower rate, which further raise its density, glass transition
temperature, Tg, and mechanical modulus - an effect known as post-cure. In
addition, a second type of spontaneous densification, which does not involve
chemical reactions, occurs when the high-energy, kinetically metastable,
amorphous structure tends, as in a chemical-stable glass, toward a lower-
energy, thermodynamically metastable structure of the corresponding
equilibrium state. Known as structural relaxation, it also increases both the
solid's density and mechanical modulus. Thus the effects of isothermal
ageing of a cured epoxy resin involve two processes; first a chemical one
that raises the cross-link's density during the early period of ageing, and
second a physical or structural one (relaxation) that does not alter the
chemical structure, but has the same effects on the density and mechanical
modulus as the chemical process, and which ultimately determines the
physical behaviour of a cured epoxy resin. Although the two processes
during isothermal ageing have qualitatively similar effects on the density
and mechanical modulus, they have quite different effects on the electrical
behaviour of epoxy resins (section 6.8).
6.3 Theoretical formalism for electrical properties
As an epoxy liquid and a proton-containing curing agent chemically react,

the isothermal growth of a macromolecule causes the molecular diffusion
coefficient to irreversibly decrease with the passage of time until the
reactions that allow the growth come to a virtual halt. The concomitant
changes in the molecular or structural state and chemical composition of the
liquid irreversibly alter its electrical properties, we deduce, in at least seven
ways, namely:
(i) A general decrease in the dc conductivity as the diffusion coefficient
or impurity ions in the epoxy resins decreases with increase in its
viscosity and any proton transfer along H-bonds in the epoxy mixture
is virtually eliminated by the formation of cross-links.
(ii) An increase in the molecular diffusion or relaxation time as a result

of which the dielectric permittivity measured at a fixed frequency
monotonically decreases toward the value corresponding to the
infrared region.
(iii) A change in the number of dipoles per unit volume, and therefore
in !:If, the contribution to permittivity from orientation polarization
as a result of both the chemical reactions that alter the dipole moment
and the volume contraction that raises the number density of dipoles.
(iv) A change in the dielectric relaxation function as the chemical
structure of the liquid changes and its viscosity and density increase.
(v) A change in !:lfin the contribution to permittivity due to infrared
polarization, as the vibrational frequencies of the various modes in
the structure change on cross-linking and densification.
(vi) A change in the optical refractive index or optical polarization as
the network structure forms and densifies.
(vii) The splitting of a unimodal relaxation function into a bimodal
relaxation function.
Thus in addition to the frequency and temperature dependence, the
complex dielectric permittivity, f* , of an epoxy resin is also time-dependent
and may be expressed by a general equation:
(6.1)
where the dielectric permittivity, f', and loss, f", contain contributions from
both the dipolar orientation polarization and dc or ionic conduction, W is the
angular frequency of the applied electric field and T and tc refer to the
temperature and time since the beginning of the curing process of a
thermoset polymer. (Terms in parentheses refer to independent variables in
such studies.)
In ion-containing materials, the measured permittivity and loss include
contributions from both the space charge effects and dc conduction which
are due to diffusion of ions and the dipolar polarization, such that
f'eW, T, tc) = (Zo sin (aJr/2) w-(1+a l )G2(w, T, tc)/C~ + fdip(W, T, tc) (6.2)
and,
fll(W, T, tc) = (ao(T, t)/weo) - (Zo cos (an/2)w- a)CZ(w, T, tc)/
Co + fdip(W, T, tc) (6.3)

where the first term on the right hand side of equation 6.2 and the second in
equation 6.3 refer to the contribution from a 'constant phase element' for an
interfacial impedance which is in series with the bulk properties of a
material. In these first terms in equations 6.2 and 6.3 Zo is the characteristic
impedance of the electrode-insulator interface, G is the measured
conductivity which is equal to the reciprocal of resistivity, Co is the geometric
ELECfRICAL PROPERTIES 179
capacitance of the sample, and a = 0.5, as Johnson and Cole (1951),
McDonald (1987) and Cole and Tombari (1991) have shown. In the last
terms of equation 6.2 and 6.3 Edip and Edip are the contributions to E' and E"
from dipolar reorientation alone. 00 in equation 6.3 is the dc conductivity
and eo the permittivity (= 8.8514 pF m -1) offree space.
For epoxy resins with relatively low ionic concentration or dc
conductivity, contributions from space-charge polarization are generally
insignificant and can be neglected, so that E' and E" may be written as,
(6.4)
and
(6.5)
where E'ion = (oo/weo) is the contribution to E" from dc conductivity. Here it
is clear that the permittivity of an epoxy resin during its curing process is
determined only by the dipolar orientation polarization but the loss is
determined by both the ionic or dc conduction and dipolar reorientation
process.
We first consider the change in the dc or ionic conductivity contribution,
E~o", during the curing of an epoxy resin. A number of the author's studies
on diamine-cured epoxies (Mangion and Johari, 1990c, 1991a, b, c; Johari
and Mangion, 1991; Parthun and Johari, 1992a, b) have shown that the
decrease in the dc conductivity of an epoxy resin during its curing is
described by a power law,
t (T) t (T)] x(T)
(J,(T t) = 0 (T t ~ 0) [ gel - c (6.6)
o 'c 0, c (T)
tgel
where oo(T, tc ~ 0) is the dc conductivity at the beginning of the curing
reaction at a temperature T, tgel is the time to reach the gelation point and
x is the critical exponent of the scaling equation used as a generalized
property function by Stauffer et al. (1982) and Djabourov (1988). (An
alternative equation oo(T, tc) = A exp (B/(to - t», which implies a
singularity at t = to, was used also by Johari and Mangion (1991), Mangion
and Johari (1991a, b, c) and Parthun and 10hari (1992a, b), but was found to
be a less satisfactory description of the measured conductivity than the
power law.) The significance of x in equation 6.6 lies in recognizing that for
a given time to reach gelation, its value determines the rate at which 00
decreases with time, i.e. the higher the value of x, the more rapid is the
decrease of 00 with the curing time. The description of the data by equation
6.6 allows extrapolation of the measured dc conductivity so that the time for
the gelation of an epoxy resin during both its isothermal and ramp-curing
can be estimated.
By combining equations 6.5 and 6.6, the dielectric loss during the curing of
an epoxy is given by:
E"(W T t) = ao(T,tc~O) [tgel(T)-tc]X(T) +E"· (W T t) (6.7)

, 'c (T) dIp , 'C
weo tgel
We now consider the dipolar contributions to E' and E". According to the
classical theory of dielectrics, the complex permittivity due to orientation
polarization alone in a material is given by
x 1'' exp - iwt ( - a:;t) ) dt (6.8)
where EO and Eoo are the limiting low- and high-frequency permittivities
respectively, t refers to the time for the decay of polarization for a chemical
structure that remains unchanged during the measurement time, or when
the curing reaction occurs too slowly in comparison with the time for
measurement. In equation 6.8 <l> represents a relaxation function whose
normalized form is given by
[<l>(t)]t, = exp [- (tlr)p]t, (6.9)
where f3 is an empirical parameter. For a Debye relaxation process, i.e. with
a single relaxation time, f3 is equal to unity. For the majority of viscous
liquids and solids, f3 is in the range zero to unity and here it is referred to as
Kohlrausch-Williams-Watts (KWW) function (Williams and Watts, 1970),
or stretched exponential parameter.
For a chemically and physically stable material, t and i in equations 6.8
and 6.9 have a strict meaning. It is the time for the observation of the decay
of polarization during which i, the average relaxation time, remains
invariant with time. For the curing process of a thermoset, i is not invariant
of time and therefore an approximation needs to be made. For frequencies
of 1 kHz or more which are often used in the studies of epoxy resins, t is of
the order of 1 ms or less and the effects from a change in both the chemical
structure and physical state of the epoxy resin during this period of 1 ms can
be justifiably assumed to be negligible at the usual curing temperatures
when the curing time is about one hour or more. This approximation allows
us to write equation 6.9 in a form used for a system with a time-invariant
chemical structure, as follows:
N*(iwi) = S: exp - iwt ( - aa~ )dt (6.10)
or
N'(wi)-iN"(wi) =.p( - aa~) (6.11)

where the normalized quantity N* is itself a function of the product of wand
r of the structural state of the epoxy resin at any given instant since the
beginning of the reaction, .Prepresents a one-sided Laplace transform and
10' and 10" are written in terms of N' and N" by:
(6.12)
and
(6.13)
where (100 - 10 = AIT with A as a constant according to the Curie Law.

(0 )
For a few epoxy resins whose 100 could be measured during the curing
process, Huraux and Sellaimia (1973), Sheppard and Senturia (1986),
Mangion and Johari (1991a,b) and Tombari and Johari (1992) have found
that the total decrease in 100 is by about 10% or less. This decrease therefore
has much less effect on 10* according to equation 6.8 in comparison with the
effect from the increase in r, which occurs by several orders of magnitude,
typically from 10- 9 s to 10+ 4 S. As a first approximation, therefore, consider
that both 100 and 10 remain constant with tc and equations 6.8 and 6.10 may be
00
combined to obtain
It is easy to see that equation 6.14 is phenomenologically invariant of wand

r as the choice of experimental variable. This means that any of the several
shapes of the relaxation spectrum or Cole-Cole plots (Cole and Cole, 1941)
can be obtained by varying either w or r. Thus during the curing of an epoxy
resin, which has been envisaged as a negative feedback between molecular
diffusion and chemical reactions (Johari and Mangion, 1991), the shape of
the plots of 10' and 10" against the reaction time resembles the dielectric
dispersion and loss spectrum (i.e. plots of 10' and 10" against the logarithmic
frequency) of a chemically and physically stable substance. However, for
epoxy resins where curing causes r to irreversibly increase, the detailed
shape of the complex plane plots of 10' and 10" is mathematically described by
a relaxation function, when measurements are made for a fixed frequency,
for which the form of equation 6.9 is to be written as:
(6.15)
where the r(tc) is now a pseudo-equilibrium average value of r at instant to
which is determined by using the limiting short-time and long-time values of
f*(iw, tJ with r as a variable of tc' The new parameter, y, has been named
the curing parameter by Mangion and Johari (1990b, c).
Before considering the application of this formalism, one further aspect
needs to be discussed here; that is that the dc conductivity of a time-variant
system such as the curing of epoxy resin is not easily measurable because of
the requirement that measurement be made as a function of decreasing
frequency till a plateau value at a low enough frequency is reached and that
during the period of measurement the chemical structure of the substance
remain unaltered. This requirement cannot be fulfilled for the curing of
epoxy resins. Therefore, it is necessary to devise an alternative procedure to
determine whether a fixed value of frequency used for the measurement is
low enough for e"(w, T, tc) to be equal to oo(T, tc)/weo within the
experimental error. We describe this procedure as follows.
The measured values of e' and e" of an epoxy resin may be transformed
into the complex electrical modulus, M*, formalisms by the equations:
M*(iw, T, tc) = (e*(iw, T, tc)r l = M'(w, T, tc) + iM"(w, T, tc) (6.16)
where
(6.17)
M' and Mil are the real and imaginary components of the complex electrical
modulus. When M* is entirely due to ionic conduction, it is related to
Maxwell or conductivity relaxation time, Ta , by the equation,
M*(iw, T, tc) = Mo( iW.TaC T , tc) ) (6.18)

1 + lWTa(T, tc)
where Ta = eoe%o, andMo = eo l . In equation 6.18, M* is invariant ofw and
Ta (or equivalently 00) as one's choice of variables and M' and Mil obey the
expression
[M"(W, T, tc)f + [M'(w, T, tc) - YzMo(T, tc)f = [YzMo(T, tc)f (6.19)
Accordingly, a complex plane plot of Mil against M' would be a semicircle
with a radius equal to YzMo and centres on the M' axis, provided w, T and
tc were such that no processes other than the ionic conduction contributed to
e*(iw, T, tc). If other contributions were present, the plot would deviate
from the semicircular shape. Such a complex plane plot can therefore be
used to determine the time during the curing process of an epoxy resin up to
which its measured conductivity for a fixed frequency is equal to its dc
conductivity.
6.4 Dielectric effects of sol-gel-glass conversion
Since the rate of chemical reactions during the curing process of an epoxy
resin is sensitive to temperature, the change in the electrical properties with
both the temperature and time become important variables in the curing
process. Therefore, measurements are often carried out as a function of
both temperature and time. As a typical example of the behaviour observed,
measurement of the changes in electrical properties during the course of
curing of a stoichiometric mixture of diglycidyl ether of bisphenol A
(DGEBA) and hexamethylenediamine (HMDA) at six different
temperatures are shown in Figure 6.1, where the real and imaginary
components of dielectric permittivity and electrical modulus are plotted
against the logarithmic curing time.
For short periods of curing, E' slightly decreases first towards a plateau
value and thereafter in a stepwise manner to E' of about 4.4 as the curing
time, to approaches infinity. This step shifts towards longer tc as the
isothermal temperature for curing is decreased. The corresponding value of
E" first decreases from a near plateau value to reach a local minimum, which
is followed by a peak. As tc ~ 00, E" decreases to reach a limiting low value of
less than 0.02, which corresponds to that of a vitrified solid at a high
temperature. As the isothermal curing temperature is decreased, the
minimum becomes shallower, the peak becomes higher and both the
minimum and the peak shift towards longer curing times.
The corresponding plots of M' in Figure 6.1 show an increase in M' with
curing time, that occurs in two steps towards an ultimate value Moo. Both
steps shift to longer times with decrease in the curing temperature, and the
first step becomes smaller in height than the second step, while the
corresponding first plateau becomes broader on a logarithmic scale. The Mil
plots in Figure 6.1 show two peaks whose positions shift to longer tc and the
width of the first peak increases.
IO..---------,r-------,------, 0.3.----------,-------,------,
,.. ::.:~Y'
.: .
.:
'
logro (curing time!s) logro (curing time!s)
Figure 6.1 The real and imaginary components of dielectric permittivity and electrical modulus
of the DGEBA cured with hexamethylenediamine measured for a fixed frequency of 1 kHz are
plotted against the curing time. The isothermal temperature for curing are: (1) 284.3 K,
(2) 296.5 K, (3) 304.2 K, (4) 312.3 K, (5) 324.3 K and (6) 336.5 K.
The theoretical formalism given in the preceding section here is more

appropriately considered by examining the shape of the complex plane plots
of E" and E' and of Mil and M', which are shown in Figures 6.2 and 6.3. The
shape of the E" against f' plots resembles that of the Cole-Cole plots of
chemically and physically stable materials with a relatively large ionic
conductivity. Nevertheless, it is advisable to maintain a distinction between
these plots, in which changes in the f* and M* occur when the dielectric
relaxation time irreversibly (and spontaneously) increases and each data
corresponds to a certain te , and Cole-Cole plots, where the changes in the f*
and M* are produced by varying the measurement frequency. The width of
the plot, i.e. (E'(te~ 0) - f'(te~ 00)), increases with decrease in the
curing temperature, as does the contribution to f" from ionic conductivity,
which, although large in the beginning of the curing process, becomes too
small to alter the shape of the complex plane plot as the curing proceeds.
The corresponding complex plane plots of M' and Mil measured for a
fixed frequency of 1 kHz in Figure 6.3 have the shape of a semicircle, with
6
DGEBA-HMDA
E"
3
"''''''''''~ ,
21----=--
. ..:. . J
O~--~L-----~----~~--~~
3 5 7 9 11
E'
Figure 6.2 The complex plane plots of e* for the DGEBA cured with hexamethylenediamine
measured for a fixed frequency of 1 kHz during the isothermal cure at the same temperature as
in Figure 6.1. (li) calculated values from the parameters, t'1.e = 4.85 and y = 0.32 for (1);
4.35 and 0.32 for (2); 3.90 and 0.32 for (3); 3.48 and 0.31 for (4), 3.00 and 0.30 for (5) and 2.45
and 0.30 for (6).
DGEBA-EDA
0.3
343.3K
\.0,' ....
." " " - - - - - i
334.6K
'y.. .., . . . .
M"0.2
....
... 324.2K
\,..••••••••• ' ••••••=.- - - - - - I
i······· ". . . \\.-.................. 313.7K
r---."\......... .. . . . .
0.1
305.3K
.................. ..... 296.2K
o 0.1 0.2 0.3

M'
Figure 6.3 The complex plane plots of M* for the DGEBA cured with ethylenediamine
measured for a fixed frequency of 1 kHz during its isothermal curing at the temperature
indicated. The semicircle is due to conductivity relaxation and the skewed arc due to dipolar
relaxation.
centre on the axis, which is followed by an apparently skewed arc. However,

as tc ---'; 00, progressively more deviations from the shape of the skewed arc
appear. A change in the temperature for curing also affects the shape of
these plots in that both the diameter of the semicircle and the width of the
skewed arc generally increase as the temperature for isothermal curing of
the epoxy resin is increased.
The shape of the complex plane plots of E* and M* indicate that at the
initial stages of an epoxy resin curing, it is the dc conductivity rather than the
dipolar reorientation which dominates the dielectric behaviour. As the
curing time increases, deviations from a semicircular shape of the Mil against
M' plot occur and a new shape of a skewed arc emerges, which is a reflection
of a dipolar relaxation process. Therefore, only part of the measured
conductivity data is mainly due to ionic conduction and this part lies at times
of cure shorter than the time tmin , at which a minimum appears in Figure 6.3.
The conductivity, a(tc) , is equal to the dc conductivity when t < tmin but
exceeds the true ao(tc) as tc approaches tmin , and this limitation should be
considered in the analysis of the kinetic effects during the epoxy resin curing.
Typical plots for the measured conductivity against the curing time for one
epoxy resin, DGEBA cured with 4,4' -diaminodiphenylmethane are shown
DGEBA-DDM
-6
,..,
~
I
S -7
IJ1
---
0
b.O
.Sl
-8
-9L-----~------~--~--L-----~
3.0 3.5 4.0 4.5 5.0

loglO (curing time/s)
Figure 6.4 The logarithmic plots of the measured conductivity against the curing time for the
DGEBA cured with 4,4' -diaminodiphenylmethane at temperatures indicated. The dashed line
was calculated from equation 6.6 or power law, and the full line from equation 6.20, or
singularity equation (after Mangion and Johari, 1991).
in Figure 6.4. The curves calculated from equation 6.6 for values of ao(~ 0),
x and tgel have been shown (by the dashed line) to indicate the adequacy of
equation 6.6 as a description of curing kinetics.
Another description, which is an alternative to the power law of equation
6.6 has also been used by Mangion andlohari (1991a, b, c) and Parthun and
10hari (1992a). Its plausibility lies in recognizing that the increase in
viscosity during the early stages of the curing process determines the ionic
diffusion according to the Stokes-Einstein equation. This is particularly so
in view of 10hari and Pathmanathan's (1990) argument that the power law of
equation 6.6, when applied to temperature-dependence of relaxation time,
is equivalent to the well-known Vogel-Fulcher-Tamman equation over a
narrow range of measurements. This alternative equation implies an
approach of ao(t) towards a singularity during the curing process according
to
ao(T, te) = A(T) exp {- B(T)/[to(T) - tem (6.20)
where to is the point of singularity or the time taken to reach a value
ao(T, te) = 0 and A and B are temperature-dependent empirical parameters
which determine the rate at which conductivity approaches the singularity at
to. The data calculated from equation 6.20 are shown as a solid line in Figure
6.4. Mangion and 10hari (1990b, c, 1991a, b, c) and Parthun and 10hari
(1992a, b) have found that the value of to obtained from equation 6.20 is
close to the time of cure when the E" peak for 1 kHz frequency measurement
appears, and which in turn is closer to the vitrification time than to the
gelation time. In addition, Parthun and 10hari (1992a, b) have found that for
DGEBA cured with a variety of diamines, the value of tgel of equation 6.6 is
virtually independent of the frequency used for the measurement, whereas
that of to of equation 6.20 systematically changes. Figure 6.5 for the DGEBA
cured with ethylenediamine shows the variation of a(tc) measured at
-7~----~~----~-------r----~
... DGEBA-EDA
'e -g
"
00
~-1O
..s
-11 o
-7~------------~------------~
-8
...
'e -g
"
00
~-1O
..s
-11
-7~--------------~------------~~
...,
e -g
a.l
'6'
~-1O
-11
-12L-----~------~------L-----~
2 3 4
Figure 6.5 The logarithmic plots of the measured conductivity against the curing time for the
DGEBA cured with ethylenediamine at 305 K. The frequency used for the measurement is
indicated. (0) data calculated from equation 6.6 or power law and full line from equation 6.20,
or singularity equation (after Parthun and Johari, 1992a.) Plot (1) is for measurements madefor
1 kHz frequency; plot (2) for 10 kHz; and plot (3) for 100 kHz frequency.
different frequencies against the curing time of the epoxy resin, where the
calculated tgel and to are marked by arrows.
The changes in £* of an epoxy resin during its curing, but after the dc
conductivity's contribution to £" has become negligible, are of some interest
here. These changes as seen in Figure 6.2 show that the fixed frequency
values of £' and £" of an epoxy resin during its curing resemble the £' and £"
plots of chemically and physically stable dipolar liquids and solids measured
at a fixed temperature but with varying frequency. This, as mentioned in
section 6.3, is a reflection of the fact that relaxation functions are invariant of
w or r as the choice of independent variable. Mangion and lohari (1990b, c)
have used equations 6.14 and 6.15 to calculate y, the curing parameter, by
using a modification of procedure given by Moynihan et al. (1973) and
Bendler and coworkers (Montroll and Bendler, 1984; Dishon et al., 1985;
Weiss et al., 1985), and Parthun and lohari (1992a,b) have calculated it by a
procedure given by Muzeau et al. (1991). The data calculated by Parthun
and lohari have been shown along with the experimental values in Figure
6.2. The agreement seen here demonstrates the adequacy of the formalism,
given in section 6.3, for the curing behaviour of epoxy resins.
Given that the measured dielectric data are satisfactorily described by
equation 6.15, it now becomes possible to calculate the relaxation time at
any instant during the curing of an epoxy resin. This is so because in the plots
of Figures 6.1 and 6.2, each data point corresponds to a unique value of
wr(te) in equation 6.14, which in turn corresponds to a unique value of N*.
Thus both the time-dependence of £* , and of r(te) can be calculated when !1£
and yare known. lohari and Mangion (1991) were first to show the adequacy
of such calculations, which was followed by the extensive work by Parthun
and lohari (1992a, b) who used a variety of curing agents to demonstrate the
variation of r(te) with the curing time of an epoxy resin. As a typical example
of this calculation taken from the work of the latter group, the calculated
values of dielectric relaxation time, and the permittivity and loss of the
DGEBA cured with propylenediamine have been plotted against the curing
time in Figure 6.6.
The range of relaxation time as determined by the above procedure is of
course limited by the insensitivity of N*(wr) to wr for a given value of yin
equation 6.15 but wr can also be varied by independently varying w which
then allows the determination of a wider range of r over a broader time scale
during the curing process. This is a remarkable feature in that it allows the
determination of virtually the entire change in r from picoseconds to
kseconds as an epoxy resin reaches its ultimately cured state. By using
frequencies of 50 Hz, 1 kHz and 100 kHz, Part hun and lohari (1992a, b)
have been able to obtain the curing time dependence of r of a number of
epoxy resins, a typical example of which is shown in Figure 6.7. The data
obtained from different frequencies of measurement overlap over a narrow
region of curing time, but all lie on the same curve which is now much
1r---------.---------,---------,
o 1
-1
-5
-6
-7~--------L---------~------~
10.--------.~~----~--------.
8
£'
·w
bO -1
..9
-2~------~--------~--------~
2 3 4 5
Figure 6.6 The relaxation time, the permittivity and loss of the DGEBA cured with
propylenediamine measured for a fixed frequency of 1 kHz is plotted against the curing time
during their isothermal curing at (1) 294.9 K,(2)303.8 K,(3)315.5 K,(4)323.7 K,(5)335.9 K
and (6) 346.0 K. (0) calculated, and (e) measured, data.
broader, covering a range of 10- 9 to 10 seconds. The single plot of the

relaxation time against the curing time seems to be a satisfactory
demonstration of the correctness of the formalism given in section 6.3.
6.5 Ionic conductivity and sol-gel conversion
In the production of diglycidyl ether of bisphenol A as in other types of

epoxies made from epichlorhydrin and bisphenol A, Hel is produced as a
DGEBA-PDA
-J
-5
-7
-gL-____ ~ ______ ~ ____ ~
3.0 3.25 3.50 3.75

Figure 6.7 The relaxation time of the DGEBA cured with propylenediamine at 324.0 K is
plotted against the curing time. The data correspond to measurements made at 50 Hz (0),
1 kHz (M and 100 kHz (D).
byproduct which is subsequently neutralized with an alkali. Even after

considerable effort in removing NaCl, certain amounts of sodium, chloride
and hydrogen ions remain as impurities in the commercially available
uncured epoxies. In addition the curing agents also contain impurity ions
such as ammonium and hydroxyl ions in amines and anhydrides and
hydrogen and hydroxyl ions in alcohols. Altogether these impurities
substantially contribute to the dc conductivity of an epoxy resin according to
the equation,
(6.21)
where ni is the concentration of ions with ionic charge Zi and mobility f-li' f-li is
often related to the inverse of viscosity according to Stokes' law and
assuming that ni remains unchanged during the curing process, a is regarded
as inversely proportional to the viscosity and its decrease to the increase in
the viscosity of the epoxy resin. This subject has been ably reviewed by
Senturia and Sheppard (1986) who have correctly pointed out that the
variation of a with viscosity is an oversimplification for an epoxy resin, for
here the viscosity approaches a formally infinite value at gelation, while
sufficient ionic and molecular degrees of freedom persist in the gelled state.
After a reanalysis ofthe work on this subject, they deduced that the decrease
in a with curing time is related to the difference (T - Tg), which itself
rapidly decreases as curing progresses. Thus after the maximum rate of
conversion which was taken to be at about 30-40%, the conductivity
becomes more sensitive to changes in Tg with the curing time. Its magnitude,
they concluded, provides no indication of the approach of gelation of epoxy
resin.
An alternative conclusion has come from the detailed studies of Mangion
and lohari (1990b, 1991a, b, c) and Parthun and lohari (1992a, b) who have
shown that the dc conductivity decreases according to a scaling law of
equation 6.6, and its value approaches zero at the gelation time of the epoxy
resin. However, this approach towards a zero value, they point out, is
interrupted by the contribution from ac conductivity which rises, with the
progress of cure, towards a peak value in a fixed frequency measurement.
They have argued that for the anticipated appearance of a singularity for a,
the time required to reach gelation should be independent of the frequency
used for the dielectric measurement, provided the change in the frequency
does not alter the relative contributions from other additional relaxation
processes. This is, of course, expected since the dc conductivity due to ionic
diffusion is invariant of the frequency for measurement, or E" inversely
scales with the frequency of measurement. They have therefore measured
the dielectric properties for a frequency of 50 Hz, 1 kHz, 10 kHz and
100 kHz as the curing ofDGEBA with ethylenediamine, propylenediamine
and hex am ethylenediamine progressed and have found that the gelation
time, tgeb obtained from equation 6.6 is independent of the frequency used
for the measurement.
A note of caution is necessary here, namely, that this new approach to
determining the time for gelation should not be regarded as undermining the
value of the now widely quoted (by Senturia and Sheppard, 1986) previous
efforts which concluded that no electrical 'events' accompany gelation, for
none does. The determination of gelation time by electrical measurement
described here relies on accurate determination of the dc conductivity and
the analysis of its variation with the curing time, which could not be done
until recently. It is also necessary to recognize the fact that the (event of)
approach towards formally a zero conductivity at gelation is prevented by
the significant contribution from ac or dipolar conductivity of the gelled
state. There is, of course, a need for confirmation of the value for gelation
time obtained from equation 6.6 by measurements using a suitable
mechanical relaxation procedure.
6.6 Time and temperature evolution of the dielectric properties
The fact that a single frequency measurement during the isothermal curing
can provide information on both the gelation time and the relaxation time is
valuable in determining how the various relaxation processes evolve on

curing an epoxy resin. In virtually all physical measurements at a certain
instant during the curing of an epoxy resin, the data obtained correspond to
relaxation characteristics of its structure at that instant of cure but for the
frequency used in the measurement. At any instant of cure this (fixed)
structure is expected to relax with at least a bimodal distribution of
relaxation times, known as a- and f3-relaxations (Johari, 1973, 1976), when
Tg of the thermoset is near or below its curing temperature, but with only one
distribution of relaxation times when Tg is much greater than the curing
temperature, i.e. under conditions when the a- and f3-relaxation processes
are merged. Thus as the curing progresses or tc increases, the structural state
of an epoxy resin changes from that of a liquid when the curing temperature
is much greater than Tg to a rigid solid when it is equal to or lesser than Tg •
For a fixed frequency measurement, as the curing progresses, changing
properties belonging to each of the continuously different structural states
would be observed as these states traverse with time from the ones with a
very short relaxation time of a fluid, say 10- 9 s, to the ones with a very long
relaxation time, say 104 , of its vitrified state. This is illustrated in Figure 6.8
where the anticipated increase in the relaxation time for the a- and 13-
relaxation processes with decreasing temperature are illustrated for
measurement made at different instants to, t1 , t2 , etc. of cure. For simplicity,
the plots of f3-relaxation processes are drawn to have the same slopes and to
merge at a frequency of measurements of about 107 Hz. The pre-
exponential factor is necessarily kept the same for all plots. In Figure 6.8 the
measured E' and E" at 1 kHz frequency for a given value of y would initially
correspond to a structure and relaxation at a point T~ 1 on curve to. With the
passage of time, and in a continuous manner, this would be followed with
those properties which correspond to the structure and relaxation times for
points at T; 1 on curves t1> t2 , ... , etc., and ultimately on curve tVi!' the
vitrification time. As tvil is reached, all contributions from the a-relaxation
process may not yet reach their minimum value if the frequency of measure-
ment corresponds to the f3-relaxation rate of the network structure. In
section 6.4 we surmised that the deviations of E" and E' from a stretched
exponential decay function in the plots of Figures 6.2 and 6.3 as tc ~ 00 are
due to the contribution from f3-relaxation, and earlier studies (Mangion and
Johari, 1990c) have shown that the strength of f3-relaxation initially
increases with time during the post-curing at the expense of the height of
another sub- Tg, y-relaxation, whose strength decreases. Simultaneously,
the a-relaxation process shifts to lower frequencies and thus its contribution
to E" decreases. In measurements made at a fixed instant after tvil , where
post-cure occurred sufficiently slowly to allow the measurement of the
spectrum, as shown in earlier studies (Mangion and Johari, 1990a), a peak
in the frequency spectrum of E" is found and a shift in the position of
the f3-relaxation peak towards high frequencies with increase in the
+8.---------------------------------,
I
(Tcure)-l
-16~------------------------------~
(temp)-l
Figure 6.8 An illustration ofthe change in relaxation time for the a- and p-relaxation processes
of the network structure at a given instant during the curing of an epoxy resin. For simplicity,
the plots for the p-process are drawn to have the same slope and to merge with the a-process at
a frequency of 107 Hz. The pre-exponent for all plots has been kept the same.
temperature. Alternatively, if isothermal curing was done at the same

temperature but different frequencies were used for measurement of the
thermoset's dielectric properties, it could be deduced from Figure 6.8 that at
tc > tvit , the measured E" is frequency-dependent, showing a peak whose
position would shift to a higher frequency with increase in the temperature
of the isothermal cure.
Both from the theoretical and experimental, or technical, points of view,
it is instructive to determine how ~E(= E'(t ~ 0) - E'(t ~ 00)), y, tgelo and
tpeak(E") change with change in the temperature of the isothermal cure. Plots
of these quantities against the curing temperature in Figure 6.9 clearly show
that their magnitude monotonically decreases with the increase in the curing
temperature. The observed decrease is qualitatively consistent with the
expectation that an increase in temperature decreases the number of dipoles
per unit volume through the Kirkwood-Frohlich equation (Frohlich, 1949)
and therefore E'(t ~ 0) decreases. An increase in temperature also
decreases the optical refractive index and the infrared pol ariz ability , and
therefore, E' (t ~ 00) decreases, but this decrease is much less than that in
6r---------r---------,---------,18
DGEBA-PDA
4 12 "'"
'",..ro
x
>< "
,...,
0 i
-"'-
2"'-"'-'"
\.
'" o OQ"'"
<l e.
P?
2 6 ~
.~
0 0
275 300 325 350
T/K
Figure 6.9 The temperature-dependence of the dielectric parameters obtained from 1 kHz
measurement during the isothermal curing of DGEBA with propylenediamine.
jj.f = (f'(t -> 0) - f'(t -> 00)), y = curing parameter, tgel = gelation time and tpeak (f")
= time when the relaxation time of the epoxy resin is equal to 159 ms. (1) jj.f; (2) y; (3) t gel ;
(4) tpeak (f").
c'(t -':> 0). The decrease in tgel> and tpeak(c") with increasing temperature is
obviously a reflection of faster chemical reaction rates with increase in the
curing temperature, that convert the sol to a gel and gel to a vitreous state.
The decrease in y with increase in the curing temperature is unexpected in
that for chemically and physically stable substances the parameter for the
stretched exponential relaxation function, {3, approaches unity as the
temperature is increased. In this respect, the curing parameter, y, differs
from the K-W-W parameter, {3, and this difference justifies both its
distinction from {3, and its significance as a kinetic parameter for a kinetic
process. In technical terms, the decreasing value for y implies that c"(tc)
becomes more spread out during the curing period and that c" of the cured
sample is greater for a higher curing temperature than for a lower curing
temperature, if contributions from the {3-relaxation process remained
constant. Note that the use of y in equation 6.15 does not imply that its
Fourier transform into the frequency domain is meaningful.
6.7 Chemical kinetics and dielectric behaviour
Let us now consider whether a useful relation between the chemical kinetics
measured in terms of the extent of reaction and the electrical properties of
an epoxy resin can be found. Attempts for seeking such a relation require
that each of the seven effects noted in section 6.3 be considered, as both the
dynamic and static properties from near zero frequency to optical
frequencies are affected by the chemical reactions. Although in principle the
magnitude of these effects can be determined both theoretically and
experimentally, the effort required for doing so is enormous. So, as a
compromise, a procedure that includes only the most prominent effects can
provide a useful relation between the chemical kinetics and dielectric
properties of an epoxy resin, and these effects are changes in the
conductivity, a, and the average dipolar relaxation time, r, of an epoxy
resin.
A number of studies up to 1986, which have been ably reviewed by
Senturia and Sheppard (1986), have related the resistivity, or the inverse of
conductivity, to the extent of epoxide conversion, a. By assuming that a is
inversely proportional to log (a), they have provided a model for the curing
kinetics expressed by the equation,
(8 log a/8th = - K(1 - a)m (6.22)
where a is the reciprocal of resistivity measured for a fixed frequency, K is
the rate constant and m is the empirical reaction order parameter whose
value varies substantially depending upon the epoxy resin, the experimental
method and the temperature range.
In a fundamentally different approach Mangion and Johari (1991a, b)
have concluded that the dipolar relaxation time, rather than the
conductivity, is related to the reaction kinetics. Their studies have shown
that the dipolar relaxation time increases with the progress of reaction and
that the logarithmic plot of r against tc is sigmoidal in its shape up to r = 103_
104 s, beyond which measurement of r is prevented by the limit on the time
itself which is required for such measurements. Parthun and Johari (1992a)
have analytically examined the dependence of r on the curing time and have
shown that the shape of the plot of the extent of epoxy conversion against
logarithmic curing time is related to that of the corresponding plot of the
average dielectric relaxation time, instead of the resistivity of an epoxy resin
measured for a fixed frequency. This relation is given by an empirical
equation:
r(T, tc) = r(T, 0) exp [San(T, tJ], (6.23)
a = [(In r(T, tc) - In r(T, O»/S]lIn (6.24)
where S = In(r(T, oo)/r(T, 0» is a constant for a given epoxy and n is an

empirical parameter. The factor S normalizes the plot of In (r(T, t» against
tc' This normalization allows, for a direct comparison of the plot of
In r(T, tc), and the plot of a, against log (tc), by allowing the values of
S-1 In (r(T, tc)/r(T, 0)) to vary between zero and unity. The parameter n
o~----------------------~
DGEBA-EDA
-2
-4
~
Ol
o
- -6
-8
•
•
-10~--~--~----~--~--~--~
o 0.2 0.4 0.6
(X
Figure 6.10 The relaxation time calculated from the dielectric behaviour of DGEBA cured
with ethylenediamine at 296.2 K (.) compared against the prediction from equation 6.22. Full
line is calculated from equation 6.23 with r(T,O) = 6 ns, r(T,ao) = 52 sand n = 1.95.
alters the shape of the sigmoidal plot so that reT, tgel ) can be made to
correspond with aCT, tgel ».
When n = 1, the extent of cure is directly
proportional to In (r(T, te».
The measurement in the MHz frequency range by Tombari and 10hari
(1992) has confirmed that for the curing of epoxy resin DGEBA with
ethylenediamine at 296.2 K, equation 6.23 or 6.24 provides a satisfactory
description of the calorimetrically measured extent of reaction and the
dipolar relaxation time with r(O) = 6.0 ns, r( 00) = 52 sand n = 1.95. A
typical plot showing such an agreement is given in Figure 6.10. There is a
need for further measurements of dielectric and calorimetric behaviours of
the epoxy for demonstrating the usefulness of this relation.
6.S Curing and the high-frequency relaxation process
Careful studies of the dielectric properties in the frequency range 0.01-

20 MHz by Tombari and 10hari (1992) have led to observations of the
dielectric loss spectrum of chemically and structurally arrested states, at
different time periods, for DGEBA-ethylenediamine epoxy resin during its

curing at 296.2 K. They demonstrate that these spectra are similar to the
corresponding plots generally obtained for structurally stable liquids and
solids and yield the same value of r as do the fixed frequency measurements
made by Mangion and Johari (1991a). One such spectrum in its normalized
form which is based on their studies is shown in Figure 6.11. In addition to
the similarity with the spectra of structurally stable liquids and solids, Figure
6.11 shows a remarkable feature, namely, the presence of a second
relaxation process which comes into evidence after a curing period of 1.25 h,
and a decrease in the strength of relaxation as the curing progresses. Thus
what appears to be a single relaxation peak at the beginning of the curing
splits into two peaks which separate progressively more from each other as
the curing progresses and as the viscosity of the epoxy resin increases. These
relaxations are referred to as a- for the low-frequency and f3- for the high-
frequency processes.
At first glance, it would appear that, in Figure 6.11, f3-relaxation is the new
process in that it seems to be the one that emerges and separates from the
a-process. However, a spectrum normalized with respect to the f3-relaxation
0.4
DGEBA-EDA
0.3
..-..
8
w
,
00.2
~
:--
-w
0.1
~
~
0
-4 -2 0 2 4
10g(<.O't)
Figure 6.11 The normalized plots of dielectric loss against frequency for the DGEBA-
ethylenediamine mixture at several fixed instants during its curing at 296.2 K. The instant of
measurement for each plot in ks is (e) 0.9; (+) 2.7; (D) 4.5; (.) 6.3; (0) 8.7; (l\.) 11.1; (83)
13.5; (lSI) 15.9.
peak, which is not shown here, would indicate that it is instead the a-process
which emerges and progressively separates from the {3-process as it shifts
towards a lower frequency. For clarity, these features are illustrated in
Figure 6.12, where the variation of the relaxation rates for the a- and
{3-processes is shown against the reaction time at a fixed temperature. The
plot illustrates a pattern of the development of relaxation processes and
their separation when an epoxy resin undergoes isothermal curing. As the
reaction proceeds, the relaxation process with a unimodal distribution of
times acquires a bimodal distribution of the a- and {3-relaxation processes,
each showing a peak for a fixed frequency of measurement. This feature of
splitting of the relaxation process, we note, bears a remarkable resemblance
to that observed on supercooling a liquid whose a-relaxation process also
progressively separates, as reversible molecular clustering raises the liquid's
viscosity, from the {3-relaxation process in a time-temperature plane (Johari,
1976).
Recent studies have shown an additional remarkable feature of the sub- Tg
relaxation in epoxy resins, namely the development of the slower or high
temperature {3-relaxation process at the expense of a faster or low
temperature, y-process. This is shown in Figure 6.13 where plots of the E"
against the temperature after different curing times are drawn. In this figure
/
/
Iu-process
/
/
-9
log(time/s)
Figure 6.12 An illustration of the change in relaxation time with increase in the reaction time of
an epoxy resin, as the a- and t3-processes evolve at a fixed temperature.
18 1 2
14 DGEBA-DDS
-w 10
2
I
0
~
-w
bD
.3 -1
-2
75 175 275 375
-w 4
X
'"0......
2
0
75 175 275
T/K
Figure 6.13 The dielectric loss of the sub- Tg relaxation regions (i.e. y and (3) and the a-relations
regions ofthe DGEBA cured with 4,4' -diaminodiphenyl sulphone. The bottom figure is for the
enlargement of the corresponding regions of the y- and f3-relaxation processes. Plots 1 to 6 in
both figures refer to an accumulated equivalent ageing time of 0.2, 1.7,3.5,5.6,9.0 and 15 ks,
respectively, and plots 7 to 9 to 24, 45 and 84 ks all at a temperature of 398 K.
each curve corresponds to an accumulated equivalent curing time at 398 K,

a concept which takes into account both the curing which occurs at certain
heating and cooling rates and that which occurs on isothermal hold at a
certain temperature (Mangion and Johari, 1991a, b). The decrease in the
height of the a- and y-relaxation peaks and the increase in that of the
j3-relaxation peak for the DGEBA cured with 4,4' -diaminodiphenyl
sulphone (Figure 6.14) clearly show the evolution of the high-temperature
sub- Tg relaxation processes as an epoxy resin progressively cures.
500~--.---.---~--~----~--~--~---,
400 DGEBA...,...DDS
Q
300
200
100
0 II
0
.....
X 9
S
-w
7
l' .~Y
I~~~~~~----------~
1
2 3 4 5 6
log(AECT/s)
Figure 6.14 The peak value for f" for the a-, {3-, and y-relaxation processes during the curing
of the DGEBA with 4,4'-diaminodiphenyl sulphone (solid notations) and with 4,4'-
diaminodiphenyl methane (open notations) is plotted against the accumulated equivalent
curing time (log (AECT» at 398 K.
6.9 Ageing effects on electrical properties
The vitrification of an epoxy resin generally occurs before the chemical

reactions that raise its viscosity reach completion, i.e. that unreacted
components persist in the cured epoxy resin. Therefore, when an in-
completely cross-linked state of a macroscopically rigid, approximately fully
cured epoxy resin is kept at a temperature below its Tg , localized rotational
and translational diffusions of its reacting molecules or chain segments tend
to form (ultimately) a stoichiometric product (Choy and Plazek, 1986;
Plazek and Frund, 1990). This chemical process increases the number
density of cross-links and thereby the Tg of the cured epoxy resin. For cases
where an increase in the number of cross-links increases the number of
dipoles which can undergo hindered rotation, the ageing of the amorphous
network structure is expected to increase, as a result of chemical reactions,
the dielectric relaxation strength of a sub- Tg relaxation process. Since this
increase competes against the decrease in the dielectric relaxation strength
that generally occurs on spontaneous densification or structural relaxation
towards a lower-energy, thermodynamically metastable structure of an
amorphous solid (Tool and Eichlin, 1931; 10hari, 1985; Plazek and Berry,
1986; Scherrer, 1986), the net change in the electrical properties of a cured
epoxy resin is expected to depend on the relative magnitude of the effects of
the chemical reaction and the structural relaxation.
Sidebottom and 10hari (1990) and Mangion and 10hari (1990c, 1991b)
have provided careful measurements of this effect for the DGEBA cured
with 4,4'-diaminodiphenyl methane or DDM. Figure 6.15 shows typical
results of a study in which the cured epoxy was aged for a total period of
about 500 hat 403 K, about 25 K below its calorimetric T g, and its dielectric
properties were measured at predetermined intervals. The cured epoxy
resin showed only one sub- Tg relaxation peak at 268 K, and the height of this
peak first increased with increase in the ageing time, attained a peak value
between the period of 48 and 168 h and thereafter decreased. An analysis of
this peak in terms of the stretched exponential relaxation function was done
by Sidebottom and 10hari (1990) and the dielectric strength tle of the sub- Tg
relaxation process was determined. This is also plotted against the ageing
time in Figure 6.15.
In the chemical reactions of curing of epoxy resin with a diamine, the
epoxy group opens to form one OH group and one covalent bond between
450 I I
2.0
DGEBA-DDk _0-
0_0
Tg 0_0 -
0- 0 -
425 0""
!
0
~
b"
E-< 400 l- 1.5 t>
t..c(ca!c) x 3 _--..... '"
~-~
/r,~~
r,/
375
I
t..c(meas) "\
I
----- 1 1 1.0
350
0 2
!oglO[ time/hi
Figure 6.15 Plots of the glass transition temperature, Tg , and dE = (Eo - Eoo), of DGEBA
cured with 4,4' -diaminodiphenylmethane against the time for isothermal ageing at 403 K. Data
for T (0) are taken from Plazek and Frund (1990). (L',) the dEvalues as measured, (.) the dE
g values as calculated. The scale for dE is on the right.
the end carbon atom of the epoxy molecule and the terminal atom of the
curing agent. The volume swept out by the hindered rotation of the OH
group about the C-O bond is much smaller than the volume swept out by
rotation of the epoxide-resin group about the newly formed C-N covalent
bond. Therefore, the increase in dE at the early stage of ageing in Figure
6.15 seems at first sight to be mainly due to an increase in the number of OH
groups formed as a result of chemical reactions, and in fact, the mechanical
relaxation peak corresponding to the j1-relaxation process in epoxy resins
observed by Ochi et al. (1987) has been attributed to the motion of hydroxyl
groups. However, a variety of studies on nonstoichiometric compositions
of cured epoxy resins by Mangion et al. (1992b) have shown that the height
of the dielectric j1-relaxation peak remains the same both for stoichio-
metrically starved (2 moles DGEBA-O.75 moles DDM) and saturated
(2 moles DGEBA-1.25 moles DDM) epoxy resin, although the number
of -OH groups formed on curing in the former case is much less than in the
latter case. This suggests that the dielectric j1-peak is not due to the motions
of only the -OH group in the network structure; other groups also move and
contribute to the j1-process.
Sidebottom and 10hari (1990) have calculated the increase in the dE of the
sub-Tg relaxation which is based on the premise that the number of OH
groups formed is equal to the number of cross-links in the network structure
and that the latter can be satisfactorily determined from the observed
increase in the glass transition temperature using the configurational
entropy theory by DiMarzio (1964). They found that the calculated dE is one
third the value evaluated from the experiment as shown in Figure 6.15.
Therefore, the formation of an OH group as a result of chemical reactions in
the amorphous solid state of a cured epoxy resin is only partly responsible
for the increase in dE. Other local dipolar segments also undergo rotational
diffusion. These dipolar segments are likely to be in the vicinity of the OH
group, but their exact chemical constitution cannot be unambiguously
identified by dielectric measurements alone. It is probable that the
formation of OH groups creates localized regions of relatively loose
molecular packing, where molecular or segmental diffusion persists.
One of the remarkable features of the change in the electrical properties
of cured epoxy resins is that when the Tg of the network structure approaches
a constant value, with increase in the ageing time, as is seen in Figure 6.15,
the dE of the sub-Tg relaxation process begins to decrease. This effect is
qualitatively similar to that observed for molecular liquids (Johari, 1973,
1976) and linear chain amorphous polymers (Pathmanathan et al., 1989;
Muzeau and Johari, 1990). It now seems sufficiently well-established that
the nature of changes in the electrical properties of cured epoxy resins over
a very long period of ageing is similar to that for other amorphous polymers,
although over a short period of ageing the nature of these changes in the two
cases contrast with each other.
6.10 Electrical applications of epoxy resins
Epoxy resins are now increasingly used in the various areas of electronics, a
use that is partly due to the versatility of the procedures by which they can be
adapted in a controlled manner for encapsulating, potting, thin-film coating
and embedding or packing electronic circuits or light emitting diodes, and
partly due to their suitable electrical properties. One of the earlier
monographs on the subject is Plastic Coatings for Electronics (Licari, 1970).
Since then a number of edited books (May, 1988; Lai, 1989; Seraphin et al.,
1989) containing articles on the subject of electrical applications of epoxy
resins have become available. Amongst these, perhaps the most
comprehensive account, with 190 references, is given by Lee (1988) in
chapter 9 of the second edition of Epoxy Resins, Chemistry and Technology
(ed. May, 1988). This may be consulted for details of the characteristics of
epoxy resins for electronic applications and procedures for their use.
Amongst the primary reasons for the use of epoxy resins in electronics is
their dielectric permittivity which should be 3-6 at ambient temperatures
and low frequencies, their dissipation factor, tan 6, of 10-3 to 10- for 60 to
1000 Hz frequency, their dielectric strength of about 120-180 kV/mm and
the volume resistivity of 1019_10 12 ohm/m, with relatively low sensitivity to
temperature. In addition, an epoxy resin should possess a number of other
physical properties in order to make it suitable for electrical and electronic
applications, such as good adhesion to metals and silicon, low permeability
Table 6.1 Typical epoxy resin compositions for 100 parts by weight of DGEBA epoxy resin
Application Inorganic filler Curing and other reagents

(by part) (by part)
Low temperature potting Glass beads (18) Butyl glycidyl ether (2)
epoxy (0.125mmdia) Diethylamino propyl amine (2)
Flame-resistant epoxy Chlorendic anhydride (90)
Miniaturized circuits epoxy Allyl glycidyl ether (10)
Polysulphide polymer (10)
2,4,6-Tris-( dimethylamino-
methyl)phenol (10)
Low exotherm potting epoxy Silica (100) Phthalic anhydride (30)
Flexible potting epoxy Polysulphide polymer (50)
Triethylene tetramine (10)
Electrical encapsulating- Silica (5) clay (75) Diethylene triamine (12)
compound
Machinable compound Powdered aluminium (200) Allyl glycidyl ether (10)
Diethylene triamine (11)
High heat distortion Silica (100) m-Phenylene diamine (14)
(casting)
For low exotherm (casting) Powdered aluminium (80) Polysulphide polymer (25)
Pumice (70) Dimethylaminomethyl phenol
(10)
or absorption of water (typically 0.1 % to 0.25%), greater stability against

chemicals and solvents to prevent both a decrease in the surface resistivity
and corrosion. There are other requirements that epoxy resins fulfil in order
to be used in electronic packaging, namely, their low viscosity prior to curing
at a sufficiently low temperature, low stress resistance that develops
between the electronic components and the epoxy resin, low heat of reaction
and high thermal conductivity to present possible damage to the device
under protection, and minimum effects of ageing. All of these take on a
particular importance in the use of epoxies for encapsulating electronic
implants such as a heart pacemaker where DGEBA epoxy cured with
diethylenetriamine is used. Table 6.1 provides a partial list of typical epoxy
resin compositions according to their applications as given by Lee (1988).
Other compositions can be obtained on request from the various producers
of epoxy resins.
Acknowledgements
I am grateful to Drs M.B.M. Mangion and E. Tombari and Mr M.G. Parthun for allowing me
here to include figures which have resulted from their research collaboration with me.
References
Alig.1. and Johari, G.P. (1992) Macromolecules 25.

Aronhime, M.T. and Gillham, J.K. (1986) Adv. Polymer Sci. 78,83-115.
Bidstrup, S.A., Sheppard, N.F. Jf. and Senturia, S.D. (1989) Polym. Eng. Sci. 29,325-328.
Cassettari, M., Salvetti, G., Tombari, E., Veronesi, S. and Johari, G.P. (1992) 1. Polym. Sci.
B. Polym. Phys. (in press); (1992) JRNCOVO Cimento 14D, 763-766.
Choy, I.-C. and Plazek, D.J. (1986) 1. Polym. Sci. B. Polym. Phys. 24, 1303-1320.
Cole, R.H. and Tombari, E. (1991) 1. Non-Cryst. Solids 131-133, 969-972.
Cole, K.S. and Cole, R.H. (1941) 1. Chem. Phys. 9,341-351.
Day, D.R. (1986) Polym. Eng. Sci. 26,362-366.
Dimarzio, E.A. (1964) 1. Res. Natl. Bur. Std. 65A, 611--617.
Dishon, M., Weiss, G.H. and Bendler, J.T. (1985) 1. Res. Natl. Bur. Stand. 90,27-39.
Djabourov, M. (1988) Contemp. Phys. 29,273-297.
Enns, J.B. and Gillham, J.K. (1983) 1. Appl. Polym. Sci. 28,2567-2591.
Frohlich, H. (1949) Theory of Dielectrics Oxford University Press, Oxford.
Gatro, J. and Yandrasits, M. (1989) Polym. Eng. Sci. 29,278-284.
Huraux, C. and Sellaimia, A. (1973) C.R. Acad. Sci. Paris 277 (Serie B), 691--695.
Johari, G.P. (1973)1. Chem. Phys. 58, 1766--1770.
Johari, G.P. (1976) Ann. N. Y. Acad. Sci. 279, 117-140.
Johari, G.P. (1985) lou. de Physique C8, 46, 567-572.
Johari, G.P. and Mangion, M.B.M. (1991) 1. Noncryst. Solids 131-133, 921-929.
Johari, G.P. and Pathmanathan, K. (1990) Phil. Mag. B. 62,225-228.
Johnson, J.F. and Cole, R.H. (1951) 1. Amer. Chem. Soc. 73,4536-4540.
Kienle, R.H. and Race, H.H. (1934) Trans. Electrochem. Soc. 65,87-107.
Kranbuehl, D.E. (1991) 1. Noncryst. Solids Part J, 131, 134,930-934.
Kranbuehl, D.E., Delos, S., Hoff, M., Haverty, P., Freeman, W., Hoffman, R. and Godfrey,
J. (1989) Polym. Eng. and Sci. 29,285-289.
Lai, J.H. (ed.) (1989) Polymers for Electronic Applications, CRC Press, Boca Raton.
Lane, J.W., Khattak, R.K. and Dusi, M.R. (1989) Polym. Eng. Sci. 29,339-346.
Lane, J.W. and Seferis, J.C. (1986) Appl. Polymer Sci. 31, 1155-1167.
Lee, S.M. (1988) In Epoxy Resins: Chemistry and Technology, ed. May, C.A CRC Press, Boca
Raton.
Licari, J.J. (1970) Plastic Coatings for Electronics McGraw-Hill, New York.
McDonald, J.R. (ed.) (1987) Impedance Spectroscopy Wiley, New York.
Mangion, M.B.M. (1990) Ph.D. thesis, McMaster University, Hamilton, Ont. Candada.
Mangion, M.B.M. and Johari, G.P. (1990a) 1. Polym. Sci. B., Polym. Phys. 28,71-83.
Mangion, M.B.M. and Johari, G.P. (1990b) 1. Polym. Sci. B. Polym. Phys. 28, 1621-1639.
Mangion, M.B.M. and Johari, G.P. (1990c) Macromolecules 23,3687-3695.
Mangion, M.B.M. and Johari, G.P. (1991a) Polymer 32,2747-2754.
Mangion, M.B.M. and Johari, G.P. (1991b) 1. Polym. Sci. B. Polym. Phys. 29,437-449.
Mangion, M.B.M. and Johari, G.P. (1991c) 1. Polym. Sci. B. Polym. Phys. 29, 1117-1125.
Mangion, M.B.M. and Johari, G.P. (1991d) 1. Polym. Sci. B. Polym. Phys. 29, 1127-1135.
Mangion, M.B.M., Wang, M. and Johari, G.P. (1992a) 1. Polym. Sci. B. Polym. Phys. 30,
433-444.
Mangion, M.B.M., Wang, M. and Johari, G.P. (1992b) 1. Polym. Sci. B. Polym. Phys. 30,
445-454.
May, C.A (ed.) (1988) Epoxy Resins: Chemistry and Technology Marcel Dekker, New York.
Montroll, E.W. and Bendler, J.T. (1984) 1. Stat. Phys. 34, 129-162.
Moynihan, C.T., Boesch, L.P. and Laberge, N.L. (1973) Phys. Chern. Glasses 14, 122-125.
Muzeau, E. and Johari, G.P. (1990) Chern. Phys. 149,173-184.
Muzeau, E., Perez, J. and Johari, G.P. (1991) Macromolecules 24,4713-4723.
Ochi, M., Yoshizumi, M. and Shimbo, M. (1987) 1. Polym. Sci., B. Polym. Phys. 25, 1817-
1827.
Parthun, M.G. and Johari, G.P. (1992a) 1. Polym. Sci. B. Polym. Phys. 30,655--667.
Parthun, M.G. and Johari, G.P. (1992b) Macromolecules 25,3149-3155.
Parthun, M.G. and Johari, G.P. (1992c) Macromolecules 25,3254-3264.
Pathmanathan, K., Cavaille, J.Y. and Johari, G.P. (1989) 1. Polym. Sci. B. Polym. Phys. 27,
1519-1528.
Plazek, D.J. and Berry, G.C. (1986) In Glass Science and Technology, Vol. 3, ed. Kreidl, N.J.
and Uhlmann, D.R. Academic Press, London.
Plazek, D.J. and Frund, J. Jf. (1990) 1. Polym. Sci. B. Polym. Phys. 28,431-448.
Scherrer, G.W. (1986) Relaxation in Glass and Composites Wiley, New York.
Senturia, S.D. and Sheppard, N.F. Jf. (1986) Adv. Polym. Sci. 80, 1-47.
Seraphim, D.P., Lasky, R. and Li, C-Y. (ed.) (1989) Principles of Electronic Packaging
McGraw-Hili, New York.
Sheppard, N.F. Jf. and Senturia, S.D. (1989) 1. Polym. Sci. B. Polym. Phys. 27,753-762.
Sidebottom, D. and Johari, G.P. (1990) Chern. Phys. 147,205-211.
Soualimia, A., Huraux, C. and Dispax, B. (1982) Macromol. Chern. 183,1803-1810.
Stauffer, D., Coniglio, A and Adam, M. (1982) Adv. Polym. Sci. 44, 103-158.
Tombari, E. and Johari, G.P. (1992) 1. Chern. Phys. (in press).
Tool, AQ. and Eichlin, C.G. (1931) 1. Amer. Ceram. Soc. 14,276-308.
Weiss, G.H., Bendler, J.T. and Dishon, M. (1985) 1. Chern. Phys. 83, 1424-1427.
Williams, G. and Watts, D.c. (1970) Trans. Faraday Soc. 66, 80--85.
7 Epoxy resin adhesives*
S.l. SHAW
7.1 Introduction
The use of structural adhesives in the manufacture of load-bearing

components has grown extensively in recent years. This can be attributed
to a number of desirable qualities which adhesive bonding allows in
comparison with more traditional joining techniques such as riveting and
welding. These include:
(i) Allowance of a relatively uniform stress distribution, resulting in
improved fatigue performance.
(ii) The ability to join dissimilar substrate materials which, due to their
dielectric nature, minimises the possibility of electrolytic corrosion
between dissimilar metals.
(iii) Allows the joining of thin-gauge metals to each other, in particular
honeycomb assemblies, resulting in the availability of lightweight
structures exhibiting high strength to weight ratios.
(iv) Allows both increased design flexibility and the ability to fabricate
complex shapes.
(v) The possibility of reduced production costs in comparison to welding
and riveting.
Although there is a range of chemically different structural adhesives
(Wake, 1982), those based upon epoxy resins have, over recent years,
earned a reputation for combining both high load-bearing characteristics
together with ease of processing, and can be regarded as workhorse
products. A number of highly favourable characteristics can be regarded as
responsible for this popularity.
(i) They exhibit excellent adhesion to most metallic alloys and various
other substrate types.
(ii) They are capable of operation at temperatures up to approximately
IS0°C for both short- and long-term applications.
(iii) They are highly versatile in the sense that a wide range of processing,
cure and property characteristics can be achieved.
(iv) They cure by reaction mechanisms which do not result in the
generation of volatile by-products, e.g. water. Thus processing is
* © British Crown Copyright 1992IDRA. Published with the permission of the Controller of
Her Majesty's Stationery Office.
EPOXY RESIN ADHESIVES 207
relatively easy, without the necessity for high applied pressures during
the bonding operation.
(v) They exhibit good wetting properties when applied to well prepared
surfaces and exhibit relatively low shrinkage during cure.
A study of the literature concerning epoxies reveals the vast number of
formulation variables which can be employed for many applications
including adhesives (Lee and Neville, 1967; Potter, 1970; May, 1988). Many
different types of epoxide resins exist ranging from the essentially
workhorse epoxy, the diglycidylether of bisphenol A (DGEBA) to the more
complex systems based upon tri- and tetrafunctional resins (see chapter 1).
Perhaps rather more daunting, a large number of curing agents can be
employed to convert the epoxy prepolymer to a crosslinked network, with
a large number of these being capable of use in adhesive formulations
(chapter 2). In addition, adhesive formulations can contain and benefit from
further additives including fillers, toughening agents, coupling agents etc.
(chapter 4), as well as being available in liquid, paste, film or carried film
form, thus demonstrating the wide variation in epoxy adhesive types
available for the potential user.
In this chapter, an attempt will be made to discuss the factors pertinent to
the successful use of epoxy-based adhesives. Within this context,
consideration will be given firstly to the theories of adhesion, i.e. why
materials stick together including the important aspect of substrate wetting.
Substrate surface pretreatments, generally considered vital for adhesive
bonding with all adhesive types, including epoxies, will be briefly discussed
together with an account of the test methods available for assessing
important mechanical properties. Formulation aspects pertinent to epoxy
adhesive applications will also be briefly reviewed together with accounts of
the properties and, of particular importance, durability characteristics.
Finally some examples of where epoxy-based adhesives have been
successfully employed will be discussed.
7.2 Theories of adhesion and wetting phenomena
Kinloch (1980) has defined an adhesive as being a material which, when

applied to substrate surfaces, can join them together and resist separation.
To provide a fairly wide and balanced account of the various facets of
epoxy resin adhesive technology it is of interest to outline briefly the various
factors and theories which underpin this definition. This will involve a
discussion of two major factors, namely theories of adhesion and wetting.
The former will hopefully provide the reader with an insight into the major
question: why do materials stick to one another, whilst the latter will address
the important subject of wetting which has been described by Sharpe and
Schonhorn (1964) as the single most important factor likely to influence the
strength of an adhesive joint.
7.2.1 Theories of adhesion

Four main theories have been proposed to account for the phenomenon of
adhesion: (a) mechanical interlocking, (b) electrostatic, (c) diffusion and
(d) the adsorption theory.
7.2.1.1 Mechanical interlocking. The mechanical interlocking theory, as

the title indicates, quite simply proposes that adhesion between an adhesive
and substrate is primarily dependent upon mechanical keying of the
adhesive into substrate surface irregularities. The fact that flat surfaces can,
in the presence of an adhesive, exhibit good adhesion clearly denies any
universal applicability. Many examples exist where a mechanical inter-
locking component clearly provides a major contribution to adhesion
(Wake, 1982). For example, those familiar with carpentry will be well aware
of the long established view that roughening of timber prior to 'glueing'
provides a much stronger joint. Later in this chapter (section 7.3.3.2), the
influence of various surface treatments on the adhesive bonding of
aluminium alloys will be discussed. Here surface topography will be shown
to be of importance, thus suggesting a significant contribution to adhesion
by an interlocking component. Thus, although mechanical factors are
generally accepted, in certain cases at least, as providing an improvement in
adhesion and thus joint strength, it is important to recognise that it takes no
account of the dominant factors which must clearly exist, on a molecular
level, at an interface between an adhesive and substrate.
7.2.1.2 Electrostatic theory. The electrostatic theory of adhesion,

expounded initially by Deryagiun and co-workers (1955, 1957, 1969) is
based on the proposed existence of an electrical double layer across an
interface. Although it is possible to demonstrate the existence of electrical
charges within an adhesive joint, the most common example of this being the
observation of electrical discharge across the gap when a flexible adhesive
tape is rapidly removed from a substrate, it is now generally recognised that
such phenomena do not provide a significant contribution to adhesion.
Indeed Kinloch (1980) has suggested that this type of observation probably
arises from, rather than results in, high levels of adhesion.
7.2.1.3 Diffusion theory. The diffusion theory of adhesion was initially

proposed by Voyutskii (1963) in the then Soviet Union. The main idea here
is that adhesion between two phases occurs as a direct result of intermixing
of the two contact substrates at the molecular level. Obviously for this to
occur, molecules existing within the interfacial zone must (a) have, or under
certain conditions be capable of exhibiting, high molecular mobility and (b)
exhibit a sufficiently high degree of compatibility. The latter requirement is
unlikely to be fulfilled for the majority of bonding applications where an
extraneous adhesive is employed to bond thermoplastic substrates since the
required levels of compatibility, particularly at the high molecular weights
present in most thermoplastic compositions, would be insufficient for bond
formation via a molecular diffusion process. Having said this, a diffusion
mechanism is generally considered responsible for the self adhesion, or
autohesion, exhibited by many unvulcanised elastomers, with natural
rubber being a well known example. In addition the solvent welding of
thermoplastics, where the substrates are treated with a suitable solvent
followed by contact and pressure/heat application, is generally believed to
be due to a diffusion mechanism (Titow, 1978).
Although Voyutskii has not claimed a diffusion-related mechanism for the
adhesion phenomena exhibited in epoxy-metal bonding, other Soviet
researchers have suggested the need for the 'diffusion' of adhesive into the
rough porous oxide layers found on many metallic alloys such as aluminium.
Although such a 'diffusion' process is clearly necessary so as to maximise the
required degree of overall area contact between adhesive and substrate,
diffusion must be regarded as an intermingling of the two components of an
interphase on the molecular level. Consequently, diffusion as described
above, cannot be regarded as being of any significance in the bonding of
structural materials such as metals and composites with epoxies or indeed
any other type of structural adhesive.
7.2.1.4 Adsorption theory. Of all theories proposed to account for the

phenomenon of adhesion, the so-called adsorption theory is the one which
has achieved more general acceptability within the adhesion fraternity and
which can be said to offer the greatest relevance to adhesive bonding with
epoxies (Hunstberger, 1967; Kinloch, 1980; Wake, 1982).
The basic tenet of the adsorption theory is that, provided sufficiently
intimate intermolecular contact is achieved between two materials, they will
adhere because of surface force interactions between the atoms in the two
contacting materials. The intermolecular forces responsible for this
behaviour include van der Waals' forces which can include the universal
dispersion forces that result from randomly fluctuating dipoles, to the
stronger intermolecular forces associated with permanent dipoles and
hydrogen bonds. Indeed it has been suggested that the excellent adhesive
properties exhibited by epoxies is due in large part to pendent hydroxyl
groups which exist within the molecular structures of most epoxies and
which are considered capable of interacting, via hydrogen bond formation,
with hydroxyl groups frequently found in abundance on most metal oxide
surfaces under normal conditions (Potter, 1970).
Mention must also be made here of the process of chemisorption whereby
adhesive application to a substrate leads to the possible generation of
primary covalent bonds. Although adhesion arising from so-called
secondary force interactions can alone result in very reasonable joint strength
characteristics (and indeed can usually be regarded as the dominant factor

in adhesion) the additional presence of covalent bonds across the interface
is often regarded as providing enhancements in load-bearing capacity.
In addition, as discussed later in this chapter (section 7.7.4.2), primary
covalent interfacial bonding is generally believed, in certain circumstances,
to offer substantial advantages in terms of environmental stability.
Finally, it is of interest to reconsider briefly the previously employed
phrase 'intimate intermolecular contact' since it provides probably the key
to successful adhesion and indeed in a sense provides the very reasons why
adhesives, such as epoxies, are frequently required for the joining of
materials. All intermolecular forces responsible for adhesion are very short
range, exerting influences over A distances only. All practical surfaces, no
matter how hard attempts have been made to smooth and polish, are
exceedingly rough. Thus, as eloquently described by Allen (1979), when two
practical surfaces are brought together, this can be regarded as akin to
inverting the Alps over the Himalayas. In other words, only very limited
contact is achieved to an extent that only a very small proportion of the two
surfaces are in true intermolecular contact. This, of course, is where an
adhesive, having the ability to flow and achieve contact with the majority of
the substrate surface comes into play. Under these circumstances,
intermolecular force interactions exist throughout the majority of the
contact area thus providing the dramatic improvements in load-bearing
capacity.
7.2.2 Wetting
Possibly one of the single most important factors likely to influence the
strength of an adhesive joint is the ability of the adhesive to wet and spread
spontaneously on the substrate surface. This most important characteristic
can be quantified in terms of the contact angle which a liquid (adhesive)
forms when placed in contact with a solid surface as shown in Figure 7.1.
Considered initially by Young (1805) he related contact angle to the surface
energies of the two contacting materials, as indicated in equation 7.1 which
has been given his name.
sotid
Figure 7.1 Contact angle 0, of a liquid (adhesive) in contact with a solid surface.
Ysv = YSL + YLV cos () (7.1)

where Ysv = surface free energy of the substrate in contact with the vapour
of the liquid, YSL = interfacial free energy and YLV = liquid (adhesive)
surface free energy.
Since a contact angle of 0 will clearly be a desirable objective, equation 7.1
can be employed to demonstrate a simple wetting criterion, namely that for
the case where an adhesive will spontaneously spread on a substrate, then
Ysv;::: YSL + YLV (7.2)

If the interfacial free energy, YSL, is ignored, which mayor may not be of
importance, the criterion demonstrated in equation 7.2 can be simplified
still further to state that spreading of adhesive on substrate will occur when:
Ysv ;::: YLV (7.3)

i.e. when the surface free energy of the substrate is greater than that of the
intended adhesive. Although highly simplified, the above offers, at the very
least, an indication of the surface energy factors which are likely to be of
importance for the successful utilisation of a structural adhesive. Indeed, the
need for surface treatments with most substrate materials prior to adhesive
bonding, be they metals, composites or plastics, can be attributed, partly
at least, to the criterion indicated in equation 7.3, as will be discussed later
in this chapter (section 7.3).
Although the surface tension of a liquid, YLV, can be measured using well
established techniques, Ysv, the surface free energy of a solid, is more
difficult to determine. Although attempts have been made to assess Ysv
values for solids close to their melting points, such values can be misleading
in adhesion-related studies since the wetting of a liquid adhesive on a solid is
usually a room temperature process, or at least conducted at temperatures
far removed from the melting point of the substrate. In recent years,
however, contact angle values exhibited by a range of organic liquids on
various substrate surfaces have been used to provide surface free energy
values (Kaelble, 1971; Sherriff, 1976).
As far as adhesion and wetting are concerned, solid substrate surfaces can
be grouped into those exhibiting low surface free energies which would
include fluorinated and hydrocarbon polymers having values of
approximately 20 mJ . m -2. Of greater magnitude would be polymers
exhibiting some considerable polar character followed by substrates such as
inorganic materials and metals exhibiting values in excess of 500 mJ . m -2.
Thus bearing in mind equation 7.3, adhesives such as those based on
epoxies, which exhibit surface free energies of approximately 45 mJ. m- 2 ,
would not be expected to wet and adhere strongly to relatively low energy
surfaces such as polyethylene. However, the reverse would be expected with
high energy solids such as metals and it is generally found that epoxy
adhesives show better performance with metals than with organic polymers.
Indeed with the latter, surface treatments designed to impart polar surface
characteristics as a means of substantially increasing surface free energy are
generally regarded as mandatory. Interestingly, the criterion demonstrated
by equation 7.3 does in fact suggest that the application of molten poly-
ethylene to a solid cross-linked epoxy should result in the thorough wetting
of the latter by the former since polyethylene is now acting as the adhesive
and the epoxy as the substrate. In other words the surface energetics are
now reversed and experimental studies have in fact confirmed the validity
of equation 7.3 in this case (Kinloch, 1980).
Although metals and metal oxides, in theory, are regarded as occupying
the high surface energy end of the spectrum, in practice reality can be
remote from this theoretical utopia. The very factors which promote wetting
and good adhesion in the presence of an adhesive, also promote the adsorp-
tion of atmospheric contaminants such as hydrocarbons and moisture which
in turn can convert the substrate surface from one of high surface energy to
one of low surface energy. The importance of this phenomenon has been
demonstrated by the work of Gledhill et al. (1977) who studied the effect of
relative humidity on the wettability of mild steel surfaces. They found, as
shown in Figure 7.2, that the presence of atmospheric moisture, even at low
relative humidities, was sufficient to bring about a dramatic reduction in
surface free energy from what would be expected for a totally dry,
uncontaminated surface (>500 mJ . m -2). In addition, increasing relative
humidity from 7 to 88% was shown to result in a gradually decreasing value
of Ysv, suggesting that the ability of the s!Irface to allow wetting by an epoxy
eo
\
70
60
..
~
j
50
>
~
40
300
15 50 75 100
RH (%)
Figure 7.2 Influence of relative humidity on the surface free energy, Ysv, of a grit-blasted mild
steel surface.
adhesive would depend upon the relative humidity prevalent at the time of
bonding. Joint strength values obtained from mild steel substrates bonded
using an epoxy adhesive under two relative humidity environments
demonstrated this to be the case, with joints bonded in the lower humidity
environment exhibiting substantially greater strength values.
In addition to the thermodynamic factors discussed above, wetting as a
kinetic process is also of importance, since although surface energetic
considerations may indicate the potential establishment of intimate
molecular contact between an adhesive and substrate, the kinetics of wetting
may be the crucial parameter. For example, a rapid curing epoxy, or one
having a particularly high viscosity, may not be able to fulfil the wetting
potential implied solely by thermodynamic considerations.
Topographical factors can also influence wetting. Wenzel (1963) has
shown that surface roughness can exert a strong influence on the contact
angle an adhesive exhibits on that surface. He demonstrated this argument
mathematically thus:
cos ()' = r cos () (7.4)
where r is defined as a roughness factor given by the ratio of the actual area
to the projected area of the surface, and ()' and () are the contact angle values
for rough and smooth surfaces respectively. Thus, with a smooth surface,
exhibiting () less than 90°, roughening would result in a reduction in ()' with
an apparent improvement in wetting. However, if () were greater than 90°
then an increase in surface roughness would have the reverse effect, with ()'
appearing to increase still further.
7.3 Substrates and surface pretreatments
As mentioned previously, adhesion is a phenomenon which exists within a

few molecular layers of the contacting materials. It is therefore necessary to
produce substrate surfaces which will be conducive to efficient inter-
molecular interactions so as to promote both adequate initial joint strength
and longevity in the intended service environment.
The degree to which surface preparation will be necessary will depend
upon several factors including in particular the level of strength desired,
longevity requirements (dependent upon service environment), level of
contamination and the nature of the substrate. Numerous surface treat-
ments exist for a wide range of substrates differing in complexity, efficiency
and the nature of their effects on substrate surfaces. In this account an
attempt will be made to describe briefly the main approaches to surface
treatment which have been employed. This will be followed by more
detailed accounts of methods investigated and employed with some of the
more typical structural materials.
Surface treatments can be divided into three broad areas, these being
associated with solvent cleaning, mechanical abrasion and chemical
treatment.
7.3.1 Solvent cleaning

Solvent cleaning is concerned solely with the removal of organic con-
taminants (greases, oils, etc.) from the surface. These contaminants will
reduce surface free energy and thus render the surface prone to poor
adhesive wetting. Solvent treatment can be approached in several ways
ranging from solvent wipe to ultrasonic vapour degreasing. The former,
which simply involves wiping the substrate surface with a clean cloth or
paper, although simple, suffers from the fact that the organic contaminants
are essentially more evenly dispersed on the surface rather than removed.
Although solvent immersion can be regarded as an improvement, solvent
purity becomes vitally important since treated surfaces will only be as clean
as the solvent. Vapour degreasing is the preferred variant. With this
technique the disadvantages of the previously described methods are
essentially removed since in this case substrate surfaces are continuously
bathed in distilled uncontaminated solvent. Thus complete removal of
organic matter is virtually assured. A more sophisticated form of solvent
cleansing, ultrasonic vapour de greasing , utilises ultrasonic sound waves,
provided by transducers incorporated into the degreasing tank, as a means
of promoting rapid agitation of solvent which provides for efficient removal
of organic matter.
Although solvent treatment is capable of removing organic contami-
nation, it has no influence whatsoever on the oxide layer present on most
metallic alloy surfaces. For many situations, removal of the perhaps
relatively weak inherent surface oxide layer is regarded as important and in
these circumstances more rigorous treatments such as mechanical abrasion
and/or chemical treatment are necessary.
7.3.2 Mechanical abrasion

The mechanical abrasion approach to surface treatment can be regarded as
incorporating processes by physicaVmechanical means without having
any significant impact on the chemistry of the surface. Paramount in this
category are treatments involving dry abrasion, i.e. sanding of the substrate
surface with abrasive paper, and grit blasting. Both techniques require
solvent de greasing prior to abrasive treatment in order to prevent transfer of
organic contaminants to the abrasive medium.
7.3.3 Chemical pretreatment
The third and perhaps the most important category, at least in terms of
typical aerospace materials such as the aluminium and titanium alloys,
concerns chemical pretreatments. With these alloys the prime objective is to
(a) remove the existing, perhaps inherently weak and unstable oxide layer
and (b) generate a new oxide layer conducive to both good adhesion and
stability, particularly to hot/wet environments. As will be described, major
advances have been achieved in recent years in developing efficient surface
treatments for both aluminium and titanium alloys.
Although epoxy-based adhesives have shown themselves capable of
bonding a wide variety of substrates, they have received more widespread
use in bonding traditional structural materials such as various metallic alloys
and, more recently, fibre-reinforced composite materials. Consequently in
this necessarily limited review, discussion of surface pretreatments will be
restricted to these more common materials.
7.3.3.1 Steel alloys. For the purposes of this discussion it is convenient to

divide steel alloys into those vulnerable to rusting on the one hand and
stainless steels on the other. With regards the former, extensive experience
over many years has shown the simple solvent degrease/grit blast/degrease
procedure to be most suitable for the majority of non-stainless steels.
Stainless steel alloys all contain, in addition to the main metallic
component, relatively large amounts of chromium the presence of which in
surface regions promotes the desired level of passivation. A number of
studies have been conducted aimed at developing surface pretreatments for
adhesive bonding. Work conducted by Allen and Alsalim (1977) has shown
that any strong reducing acid will etch stainless steel surfaces with, in
particular, sulphuric acid (10%) and hydrofluoric acid (4%) producing the
most beneficial results in terms of improvements in joint strength beyond
that which could be expected with a simple solvent degrease pretreatment.
Using this approach the deposition of a carbon deposit following the etching
process necessitated the need for a post-etch de smutting procedure. Nitric
acid and chromic acid were found suitable for this purpose. In addition work
by Smith and Haak (1979) showed that additional treatments involving
chromic acid etching and anodising in nitric acid provided substrate surfaces
beneficial to adhesion and adhesive bonding.
7.3.3.2 Aluminium alloys. As a result of their strength, low density and

corrosion resistance, aluminium alloys have achieved a virtual monopoly in
the aircraft industry. For many years aluminium, whether pure aluminium
cladding, an alloy, or an unclad alloy, was pretreated for adhesive bonding
using a chromiC/sulphuric acid etching process. This treatment has,
however, lost favour in adhesive-bonding applications because of problems

that were encountered concerning poor durability of adhesive joints in
warm/moist environments, this being particularly relevant with the
relatively new low temperature curing, toughened epoxy adhesives. In
comparison the more traditional phenolic-based adhesives have continued
to exhibit excellent durability characteristics with aluminium subjected to
the chromic/sulphuric acid etch (Schliekelmann, 1979; Albericci, 1983).
In an attempt to provide superior surface pretreatments, Boeing began a
programme in the late 1960s aimed at examining the efficacy of an anodising
process employing phosphoric acid. Greatly superior performance was
achieved using this technique resulting in its commercial exploitation in 1974
(McMillan, 1979). It is now widely employed in aerospace applications in
the United States. Contrary to this practice, chromic acid anodising has been
the preferred pretreatment in Europe where it has been used successfully for
many years.
With both anodising techniques, surface treatment would involve, in
broad terms, a preliminary degrease in, for example 1,1,1-trichloroethane
followed by an alkaline etch, the latter designed primarily to remove
inherent oxide. This would be followed by etching in chromic/sulphuric acid
followed in turn by the appropriate anodising procedure. Interdispersed
between these main procedures would be various rinsing processes followed
finally by drying and primer application.
7.3.3.3 Titanium alloys. Although aluminium alloys are currently the

most widely used alloys in aircraft construction, the need for a higher
temperature capability in, for example, supersonic aircraft applications, has
resulted in titanium achieving a small measure of utilisation. As with
aluminium, the desirability of adhesive bonding has resulted in numerous
studies being conducted on titanium with a view to the development of
efficient surface treatments. Most notable amongst these have been those
based upon chromic acid/fluoride anodising (Moji and Marceau, 1976), an
alkaline/hydrogen peroxide etch (Cotter and Mahoon, 1982) and a sodium
hydroxide anodising process (Poole, 1985).
7.3.3.4 Fibre-reinforced composites. Composites based upon glass,

Kevlar and particularly carbon-reinforced polymeric composites offer
substantial advantages over traditional structural aerospace materials in
terms of reduced weight together with potential improvements in specific
strength and modulus. As a result of inherent problems associated with the
joining of composite materials using conventional fastening techniques,
adhesive bonding has been regarded as potentially a highly desirable
alternative to mechanical joining.
In surface treatments the prime objective is, in part, similar to that
previously described for metallic alloys, i.e. to produce a clean surface free
of major extraneous contaminants. In addition, however, the polymeric
component of fibrous-reinforced composites is, due to is organic character,
prone to moisture uptake particularly when exposed to highly humid and
warm environments. Studies have shown that absorbed moisture can result
in the generation of voids in adhesive bondlines resulting in significant
reductions in both initial strength and fatigue resistance (Sage and Tin, 1982;
Parker, 1983). Thus drying at elevated temperatures prior to bonding so as
to yield a water content of less than 0.4 % is generally considered advisable.
Apart from this, surface treatments are relatively simple in comparison to
those employed with metallic substrates with a carefully controlled grit blast
or even hand abrasion generally being considered suitable for the attain-
ment of both a high initial joint strength and adequate long-term durability.
7.3.4 Primers
In addition to the surface pretreatments described above, the use of
substrate surface primers is frequently recommended for the adhesive
bonding of many metallic alloys. Primers can be employed to fulfil a number
of requirements. These can include:
(i) Protection. Freshly prepared metallic surfaces are prone to contami-
nation by atmospheric contaminants including water, hydrocarbons,
etc. The virtually immediate coating of the pretreated surface with a
primer can protect that surface from atmospheric contaminants by
basically exchanging a surface of very high free energy, and thus prone
to contamination, for one of more moderate surface free energy but
highly compatible with the adhesive. After priming, the surfaces are
less prone to contamination resulting in a substantial relaxation in
the necessary time interval between surface pretreatment and
bonding. Such primers are frequently solvent-based systems with
solids similar to the intended adhesive.
(ii) Wetting enhancement. Adhesive systems used for metal bonding are
often composed of films carried on cloth supports such as polyester
or nylon. Because of their precise formulation, film adhesives often
suffer from having only a short period of time at the appropriate
bonding temperature before cure inhibits flow in a system which
already has an inherently high viscosity. The application of a primer
solution to a metallic substrate prior to adhesive bonding can there-
fore overcome the potential wetting problems which would clearly
be associated with the use of such adhesives.
(iii) Corrosion inhibition. Corrosion-inhibiting primers, generally com-
prising an epoxy together with strontium chromate in a solvent
base, are generally recommended for use with aluminium alloys
following anodising pretreatments (Noland, 1976). This combined
approach to surface pretreatment has been shown to provide the

highest degree of moisture protection to bonded aluminium alloys,
particularly when bonded using relatively low cure-temperature
epoxy adhesives.
(iv) Durability enhancement. Under certain circumstances, water can
cause the displacement of adhesive from a substrate producing a
debond resulting in turn in a substantial reduction in load-bearing
capacity. This effect has been shown to occur with steel joints bonded
using simple epoxy adhesives (Gledhill and Kinloch, 1974). Studies
using in particular organosilane compounds as the basis of primers,
have shown that the application of organosilane solutions to substrate
surfaces prior to adhesive bonding can result in substantial improve-
ments in adhesive-substrate interfacial stability in the presence of
moisture (Gledhill et al., 1990). Durability related aspects such as
this will be described in greater detail later in this chapter (section
7.7).
7.4 Methods oftest
The mechanical testing of structural adhesive joints can be divided into two
broad classifications namely (i) small coupon scale tests bonded and tested
to well-defined specifications and (ii) evaluation at a structural level with
either a complete structure or a sub-assembly being assessed. The former
approach provides a means of assessing various adhesive joint qualities
associated with material components, surface treatments and bonding
procedures, allowing the generation of large amounts of data at relatively
low cost. On the other hand structural level testing provides much more
limited information at greater cost and is generally employed as a means of
verifying design principles originally based, to a degree, on preliminary
coupon type tests. Since tests at the structural level would generally be
conducted on designs specific to a particular application, and therefore
unique to that application, this section will be devoted to the small, coupon
type test techniques which can be applied to epoxy adhesives in addition to
other generic types.
Before briefly describing some of the main test methods which have been
employed, it is of interest to mention the three main loading modes which
adhesive joints can encounter in service and which test methods are
designed to simulate. These are shown in Figure 7.3.
Although this approach is generally used in a fracture mechanics context
(see chapter 5), it is useful to employ it here. The opening mode I situation,
usually obtained in joints subjected to cleavage or peel, can be regarded as
the most harmful loading mode which an adhesive joint can encounter.
Conversely, mode II, an in-plane shear mode, represents the most desirable
I I[
Figure 7.3 Modes of loading. I, tensile-opening; II, in-plane shear; III, tearing, antiplane
shear.
means of load transfer with most adhesives, adhesive joints being optimised
to maximise shear properties. Mode III, which represents a tearing mode of
failure has been little studied and will not be considered further.
7.4.1 Conventional test techniques

Numerous adhesive joint configurations are available for evaluating the
performance of adhesives, surface treatments and bonding processes, etc. In
this necessarily limited account only the more common and most widely
utilised of these configurations will be discussed. More extensive coverage
can be found in the literature (Anderson et al., 1977; Arnold, 1981;
McMillan, 1981; Wake, 1982; Adams and Wake, 1986).
7.4.1.1 Lap-shear specimen. The lap-shear specimen, in its various

forms (Figure 7.4) is probably the most commonly employed method for
evaluating adhesive joint strength. Although the name implies a predomi-
(0)
Cb)
'" (c)
Figure 7.4 Typical tensile lap-shear joint geometries (a) single lap-shear joint; (b) double lap-
shear joint; (c) scarfjoint.
nantly mode II shear stress application, this is a major oversimplification.

As a result of its geometry, complications arising from, in particular,
adhered bending can result in the generation of substantial cleavage stresses
at the specimen ends in addition to a non-uniform shear stress distribution.
Thus, to employ fracture mechanics terminology, the lap-shear con-
figuration can be regarded as exhibiting combined mode 1111 behaviour with
significant peel and shear stress contributions. In spite of this complication,
the lap-shear joint owes its popularity to ease of both preparation and test in
addition to the fact that, in certain respects, its joint geometry simulates the
combined mode IIII loading situation experienced by bonded joints in
service. It thus allows an insight into the adhesives which are likely to
perform well when subjected to this combined stressing situation.
7.4.1.2 Peel. Peel testing of adhesive joints is probably second only to lap-
shear in popularity, this arising from both ease of manufacture and test. In
addition, although exhibiting certain disadvantages, peel tests yield
qualitative information regarding the ability of an adhesive joint to
withstand the most damaging applied stress which an adhesive joint may be
subjected to in service.
Three main types of joint configuration have been employed. These are
generally referred to as the T-peel, Bell peel and the climbing drum method
(Figure 7.5). The T-peel is probably the most widely used and is prepared by
bonding together two adherents of similar thickness which, after cure, are
pulled apart in the manner indicated. The Bell peel and climbing drum
techniques involve the peeling of an adherend at constant radius around a
steel roller. Both methods involve considerable energy absorption in the
adherends so that the peel values obtained from these two techniques can
be, and generally are, considerably higher than those obtained, for example,
from the T-peel where deformation in the adherends is considerably less.
Due to the difficulties inherent in assessing both applied and failure
stresses, results from peel tests are generally reported as linear values,
e.g .N. mm .
-1
7.4.1.3 Other conventional test methods. A wide range of other joint

geometries has been employed in assessing adhesive joint behaviour.
These include butt-joints, napkin ring specimens and various modifications
thereof. One further specimen worthy of note which has achieved some
measure of prominence in recent years is the so-called thick adherend
specimen (Figure 7.6). As indicated, the specimen is machined from two
25.4 mm thick aluminium plates 50.8 mm wide by 352.8 mm long, pre-
viously bonded together. A 25.4 mm overlap is machined into the specimen
in the manner shown. This type of specimen has been employed to measure
the shear properties of adhesive bonds, this being largely possible due to the
thickness of the adherends restricting adherend bending. This in turn
(0)
(b)
Ie)
Figure 7.5 Typical peel joint configurations. (a) T-peel; (b) climbing drum; (c) Bell peel.
inhibits the generation of peel loads of the type found in other joint
configurations such as the standard lap-shear specimen.
7.4.2 Fracture mechanics approach

Since structural adhesives such as those based on epoxies frequently fail by
progressive crack propagation, failure criteria based upon both the initiation
and propagation of flaws inherent in the joint can be considered relevant.
Consequently fracture mechanics, where theory enables a mathematical
evaluation of the conditions in which flaws will propagate under stress, has
recently been considered as a means of assessing the load-bearing
characteristics of adhesive joints.
The study of fracture mechanics can be conveniently divided into two
inter-relatable parts. The first, arising from the initial work of Griffith
,
- t==/ ==~~d?j~~r----__._?l-
____________________
~t:= --J~ ______________
Figure 7.6 Thick adherend specimen.
(1920) quite simply proposes that fracture occurs when sufficient energy
(denoted by the term G) is released from the stress field by growth of the
crack to satisfy the requirements of the new fracture surfaces. The second,
proposed initially by Irwin (1964), is based on the premise that the stress
field at the crack tip can be defined by a parameter called the stress intensity
factor, K, and states that fracture occurs when the value of K exceeds a
critical value. Although both approaches have been employed in the study
of adhesive joint fracture, the energy balance approach has received most
attention.
In recent years several specimen geometries have been designed with a
view to measuring the fracture resistance of structural adhesives (Kinloch
and Shaw, 1981a); most studies being devoted to epoxy-based adhesive
systems. Some of these designs are indicated in Figure 7.7. Probably the
most popular specimen employed to assess mode I fracture energy, G'e>
has been the contoured double cantilever beam specimen. Designed and
initially employed by Mostoroy and Ripling (1966), this specimen is
contoured in such a way as to provide a constant compliance. Since at a given
applied load the value of fracture energy remains independent of crack
length, G can be readily determined without knowledge of crack location; a
feature which can be extremely useful where difficulty is encountered in
determining the precise location of the crack tip. A number of variations of
the above geometry have been designed and employed.
In addition to mode I, specimens have been developed for measuring
mode II (in-plane shear) and combined modes 1111 loads. With regards the
last, the scarf-joint (Figure 7.8) has been designed to impart a combination
of both mode I and mode II load conditions on the adhesive layer; the value
of () dictating the precise load combination (Trantina, 1972). The so-called
independently-loaded mixed-mode specimen, also shown in Figure 7.8, was
designed to allow independent application of cleavage and shear loads so as
to provide both modes I and II or a controlled combination of the two
(Bascom et al., 1977).
Mode III loading has only received minor attention with regards adhesive
joint fracture studies.
7.4.3 Environmental testing

One of the most hostile environments that an adhesively bonded structure
can experience in service is one consisting of a high relative humidity
(a)
/ /
9 /
~
(c)
Figure 7.7 Typical mode I fracture mechanics adhesive joint specimens. (a) Parallel double
cantilever beam; (b) contoured double cantilever beam; (c) double-torsion.
together with a high ambient temperature. Consequently investigations

aimed at assessing the ability of bonded joints to withstand such harsh
environments are reasonably common. Two approaches are generally
employed. First, to submit bonded joints to exposure trials in 'real'
environments, including in particular potentially damaging tropical
Ca)
(b)
o
Figure 7.8 Typical mode II fracture mechanics adhesive joint specimens. (a) Scarfjoint;
(b) independently loaded mixed-mode joint.
exposure. Second, to undertake accelerated tests with adhesive joints

subjected to condition extremes beyond those that would be expected in
actual service. Although not capable of providing an accurate measure of
joint longevity, accelerated exposure trials allow, at the very least, insights
into the bond components (adhesive, interface, oxide layer, etc.) which
could be vulnerable to potentially harsh service conditions, as well as
providing a fairly simple and convenient means for investigating ways of
improvement.
Most of the joint configurations discussed above have been used, with
varying degrees of success, in accelerated exposure trials. One particular
specimen not described thus far and worthy of mention here is a double
cantilever beam variant which has become known as the Boeing wedge test
(Marceau et ai., 1977). This specimen is qualitative in nature and has as its
main advantage both simplicity in preparation and use together with the
speed with which information can be provided, particularly concerning
environmental resistance. As indicated in Figure 7.9, the specimen is simply
stressed by inserting a wedge into the bondline, thus requiring no external
load application and enabling joints to be subjected to controlled
environments. Durability assessments are made by relating crack extension
measurements to time and the locus of failure can be observed. The wedge
test is now commonly employed as a comparative test technique for
Figure 7.9 Boeing wedge test adhesive joint specimen.
assessing factors such as surface preparation, adhesive type, etc. (Marceau

and Scardino, 1975; McMillan, 1979).
7.5 Epoxy adhesive formulation
Structural adhesives based on epoxies can be formulated to fulfil a wide

range of applications together with cure and performance requirements.
Indeed the number of formulatory variables can be regarded as immense
providing the formulator with considerable scope, as well as difficulty,
for achieving a specific set of requirements. Not only will characteristics
be dependent upon the two main formulatory variables, i.e. resin type
and curative but will also depend upon the selection of a wide range of
available modifiers such as fillers, diluents, flexibilisers, toughening agents,
etc. However, although a large number of epoxy-based adhesives are now
available commercially, it will be likely that the majority of these will be
based upon a surprisingly small number of the many available ingredients.
In this necessarily limited account, an attempt will be made to outline the
main formulatory materials available with examples given of the most
common types employed together with the benefits they bestow. A more
detailed account of the various types of additives and modifiers available for
use with epoxy resins generally can be found in chapter 4.
Epoxies used in adhesive formulations can range from very simple, low
molecular weight, primarily aliphatic resins to complex multi-functional
predominantly aromatic types. However, by far the most common type
employed is the diglycidyl ether of bisphenol A (DGEBA), having the
following structure:
This material can, in fact, be regarded as the standard workhorse resin of

the epoxy range and will be found in the majority of epoxy-based structural
adhesive formulations. Epoxies of this type are available in varying
molecular weights and can range from moderately viscous liquids (low
values of n) to solids at room temperature (high n). Liquid and paste

adhesive grades are generally composed of relatively low molecular weight
epoxies. Higher molecular weight variants are frequently employed in film
adhesives where a degree of structural integrity is required under use
conditions.
In addition to DGEBAs, other resin variants are available, each capable
of exploiting specific service requirements. These include interesting
examples of epoxies being alloyed with other resin types including nylon-
epoxies, polysulphide-epoxies and epoxy-phenolics. The former are based
on a blend of epoxy with a soluble grade of nylon and have been shown to
produce structural adhesives exhibiting toughness and a remarkable
resistance to cleavage forces thus resulting in excellent peel and lap-shear
strengths. Unfortunately the propensity to absorb large concentrations of
moisture, thus rendering them susceptible to environmental degradation in
particularly harsh hot/moist atmospheres, has somewhat diminished the
euphoria which initially greeted their introduction. In addition blended
resins such as these, together with polysulphide-modified variants are
limited to a high temperature ceiling of approximately 70-80°C. However,
epoxy-phenolic adhesives, produced by reacting a high molecular weight
epoxy with a phenolic resin, are capable of operation at considerably higher
temperatures, particularly for short periods of time.
Choice of curing agent can be equally as important as choice of resin and
the adhesive formulator would need to consider several factors in choosing
the appropriate curative regime for a specific formulation. These factors
would include application conditions (particularly service temperature)
together with cure and other processing requirements. Although literally
hundreds of potential curing agents have been identified, only a relatively
small number have been employed in adhesive formulations. By far the most
common are the amines which can be categorised, chemically speaking,
into the aliphatic primary and secondary polyamines, tertiary amines,
polyamides, heterocyclic amines and aromatic amines (chapter 2).
Aliphatic amines of various types form the basis of most commercial room
temperature-cure epoxy adhesives. Not surprisingly these are generally two-
component systems. Although simple aliphatic amines such as diethylene-
triamine (DETA) and triethylenetetramine (TETA) have been and still are
to a limited extent employed in room temperature cure formulations,
amines such as diethylene glycolbispropylamine, together with those based
upon polyamide structures are nowadays more typical. The latter type offers
the advantage of imparting improved flexibility to an otherwise rigid and
possibly somewhat brittle cured resin. In addition, because of their reduced
functionality in relation to aliphatic amines such as TETA, they can be used
to good effect in equal proportions with epoxies, thus minimising potential
mixing errors generally associated with curatives necessarily employed in
low concentrations. This, together with lower volatility and toxicity in
comparison with most other amines has made their use popular, particularly
as the basis of the twin-tube epoxy adhesive kits now so prevalent in the
domestic market.
Although two-part adhesive systems were originally the norm, one-
component epoxy adhesives are now readily available, exhibiting long shelf-
life characteristics, particularly when subjected to refrigeration or frozen
storage. This clear advantage is achieved by the use of latent curing agents,
with one particular amide, dicyandiamide (DICY), being the most common
curative employed in one-component epoxy adhesive systems. DICY is
generally employed in concentrations of approximately 5 phr (parts per
hundred of resin by weight) and results in the rapid polymerisation and
cross-linking of an epoxy on heating the formulation in excess of 150°C.
In order to bring cure temperature down to more manageable levels, DICY-
cured formulations generally require the use of a catalyst. Film adhesives, by
necessity also employ latent curing agents so as to allow the required storage
stability together with the requirement of being able to undergo film pro-
cessing procedures without undergoing any significant cure.
Other curing agents which have been, and to a limited degree are still
employed, include anhydrides and aromatic amines. These curing agents
can be employed in hot-cure adhesive formulations and are capable of
providing cured adhesives exhibiting higher temperature capabilities than
the other curing agents discussed.
After resin and curing agent, fillers are the most common additive in
epoxy structural adhesives where their incorporation can result in some not
inconsiderable improvements in behaviour. Notable advantages include
their ability to reduce shrinkage, thermal expansion coefficient, exothermic
characteristics and, of course, cost. Their ability to reduce shrinkage can in
fact be regarded as impressive. Some unfilled epoxy formulations are
capable of undergoing about 5% shrinkage on cure which can exert a sub-
stantial influence on joint performance. The addition of filler in the appro-
priate concentrations can reduce shrinkage and result in quite a dramatic
improvement in properties. In addition, a reduction in coefficient ofthermal
expansion can also reduce significantly the thermal expansion mismatch
which would exist between a substrate and adhesive, thus minimising the
generation of internal stresses during the cure process.
Filler incorporation has also been used to good effect to influence the flow
characteristics of epoxy adhesives. Under certain circumstances thixotropy,
a characteristic of particular practical importance in many paste adhesive
formulations can be obtained by a suitable choice of filler. Such
formulations offer the advantage of a fairly limited degree of flow when
subjected to a moderate applied shear; the adhesive reverting to virtually
zero flow when shear is removed.
In addition to these fairly broad effects, somewhat esoteric fillers can be

employed to enhance or provide specific properties. As an example, the use
of silver-filled epoxies has been employed in electrically conductive bonding
applications. Although reasonably effective in this application, this can
clearly be regarded as an example of filler addition not resulting in reduced
formulation cost.
Fillers commonly employed in epoxy adhesive formulations include silica,
aluminium and alumina powders.
Other possible additives include diluents, flexibilisers and toughening
agents. Since these are discussed in some depth in chapter 4, they will not be
mentioned further here. However, brief discussion oftoughening agents is
worthy at this point since they can, to a degree, be regarded as the one
formulatory ingredient which has put the word 'structural' into the term
'structural epoxy adhesive'.
Epoxies, in common with many other thermoset polymers are prone to
brittle behaviour which can result in relatively poor adhesive joint
performance, particularly where significant peel forces exist. Toughening
agents, most notably those based upon liquid reactive rubbers such as
carboxyl terminated butadiene acrylonitrile rubbers (Ting, 1988) together
with the use of thermoplastic-based modifiers such as polyethersulphone
(Partridge and Bucknall, 1983) have been employed to excellent effect to
dramatically increase toughness and improve resistance to peel forces; this
generally being achieved with minimal effect on other important properties
and parameters such as modulus and glass transition temperature, Tg •
Finally, a brief word regarding epoxy adhesive films. Adhesives of this
type are generally available in both supported and unsupported forms; the
former containing scrim carrier, generally based on nylon or polyester. Film
adhesives of this type, as well as being relatively easy to handle, also offer
more controlled regulation of bond thickness in comparison to their liquid/
paste counterparts. They are generally employed in relatively large area
bonding applications.
7.6 Properties of adhesive joints
The properties of an adhesive joint will depend upon several factors

including, for example, the mechanical and physical characteristics of the
bulk adhesive material, the degree of adhesion between the adhesive and
substrate and the design of the joint. Factors likely to influence adhesion
such as surface pretreatment have already been discussed and will therefore
not be considered further. The highly important subject of adhesive joint
design can also be considered beyond the scope of this chapter. This section
will therefore be devoted to a consideration of the bulk properties of epoxy
adhesive materials which in turn will be followed by some examples of
typical adhesive joint properties which could be expected from a range of
epoxy-based structural adhesives.
7.6.1 Bulk properties of epoxy adhesive

The properties likely to be of importance in determining the strength of a
polymer, such as an epoxy, can be demonstrated by the use of the Griffith
equation
_ (2EG c )1/2
Of- -- (7.5)
na
where Of is the stress at fracture, E is modulus, Gc fracture energy and a
regarded as a critical flaw size. As indicated, the Griffith equation shows
clearly the direct positive relationship between stress at fracture, or the
'strength' of the material and the two other key properties of modulus and
fracture energy. Although flaw size a can be regarded as a source of
complication, the Griffith equation provides a clear indication as to the bulk
mechanical properties which are likely to influence the load-bearing
capacity of an adhesive material. Clearly high values of both modulus and
fracture energy, all other factors being equal, are likely to enhance the load-
bearing characteristics of a cured epoxy polymer.
The fracture energy, G Jc , of all epoxy formulations will depend critically
upon the precise formulation details, e.g. the type of resin and curing agent
employed, together with other factors such as cure conditions. An
examination of the literature provides a clear indication of the range of
fracture energies obtainable by simply varying choice of resin and/or curing
agent (Young, 1980; Kinloch and Young, 1983). Values varying between
approximately 50 to in excess of 600 J . m -2 have been obtained. Cure
conditions, in particular cure temperature, have also been shown to
promote large variations in fracture energy. For example work conducted by
Shaw and Tod (1989) showed that by varying the temperature of cure
between 120 and 160°C over cure times ranging from 2 to 6 hours fracture
energy GJc varied between 180 and 360 J.m -2 (Figure 7.10); the higher GJc
values occurring at the higher cure temperatures. Likewise, modulus has
also been shown to be dependent on the nature of both the resin and curing
agent, although, provided gross undercure is avoided, variations in modulus
from one formulation type to another are generally minor. Variations in
cure temperature are also shown to provide significant but relatively minor
changes in modulus.
The introduction in the late 1960s (McGarry and Willner, 1968) of rubber-
modification as a means of enhancing toughness has done more than
anything else to promote the view of epoxies as viable structural materials.
Numerous studies (Kinloch and Shaw, 1981a; Kinloch et al. 1983) have
-
u 150
L
300
250
2DO
120 2~--~3---4&""---..I.5-~---'6
Cure time (h)

Figure 7.10 Effect of cure conditions (cure temperature and time) on the fracture energy
(l.m -2) of piperidine· cured DGEBA epoxy.
shown that the incorporation of reactive liquid rubbers such as carboxyl-

terminated butadiene acrylonitrile rubbers (CTBN), can substantially
elevate toughness whilst having a relatively minor influence on glass
transition temperature, Tg , and properties (particularly modulus) associated
with it. An example of the magnitude of toughness enhancement is shown in
Figure 7.11.
As indicated, the incorporation of approximately 15 phr of CTBN can
result in an approximately forty-fold improvement in G Ic , thus clearly
transforming an essentially brittle polymer into one exhibiting toughness
characteristics similar, if not identical, to various thermoplastic polymers.
Increases in CTBN concentration beyond approximately 20 phr have been
shown to result in a decline in GIc and, particularly modulus, this being due
to a process known as phase inversion which can result in flexibilisation of
the system (Bascom et al., 1975). For this reason, structural epoxy adhesive
formulations of this type generally contain CTBN in concentrations of
10-15 phr.
Variation in cure conditions, as with unmodified epoxy formulations, has
also been shown to have a strong influence on the toughness of rubber-
modified epoxies. In particular the work of Shaw and Tod (1989) has shown,
as indicated in Figure 7.12, that by simply varying cure temperature between
120 and 160°C over cure times ranging from 2 to 6 hours, GIc changes
between 1.8 and 12 kJ. m -2 are possible; as with unmodified systems, the
higher cure temperatures producing by far the highest Gic values. Although
users of epoxy adhesives would clearly prefer to operate at much reduced
~
u
E
U.l
5 10 15 20
(TBN Cone (phr)
Figure 7.11 Influence of elastomer incorporation (CTBN) on the fracture energy of a
piperidine cured DGEBA epoxy.
temperatures, preferably room temperature, studies such as these have

clearly demonstrated the remarkable influence that cure temperature can
have on the toughness of rubber-modified epoxy adhesives and, in turn, on
adhesive joint strength parameters which would be expected to be
dependent on toughness, e.g. peel and lap-shear strength. Indeed,
experiments conducted by the author have shown that lap-shear strength
values, obtained from joints produced from mild steel adherends using the
epoxy adhesive formulation mentioned above, increased significantly with
an increased temperature of cure.
Other epoxy resin characteristics likely to be of importance to the efficient
and safe utilisation of adhesives based on them would include Tg and water
absorption behaviour. To a first approximation, Tg will dictate the high
temperature limit of an adhesive formulation and various factors such as the
type of resin and curing agent employed in the formulation and the curing
regime employed will dictate the level of Tg obtained. Generally,
formulations employing multi-functional resins and curing agents, together
with hot-cure as opposed to room temperature cure systems would result in
relatively high Tg adhesives. For epoxy formulations generally, Tg values up
to approximately 250-300°C have been obtained (Morgan, 1980). However,
for the more common toughened adhesive systems based upon DGEBA
resins cured with DICY and employing CTBN-based toughness modifiers,
Tg values in the region of 100°C are more typical.
-
~ 150
~
.-
:J
a
L.. 1
~
E
~
~130
:J
t...J
170
2 " 3 fo. 5
Cure time (h)
Figure 7.12 Effect of cure conditions (cure temperature and time) on a piperidine-cured
CTBN-modified DGEBA epoxy.
Finally, it is important to recognise that a moisture laden atmosphere can

exert a severely debilitating effect on adhesive joints, as is discussed in
section 7.7. Epoxy adhesive materials, being predominantly organic in
nature, will be prone to water absorption, both the amount absorbed and the
rate of absorption being dependent upon formulation variables such as the
epoxy resin and curing agent types employed, together with environmental
variables such as temperature and relative humidity as well as cure
conditions. A wide range of equilibrium water concentration values and
diffusion coefficients have been quoted in the literature for an equally wide
range of formulations and absorption conditions (Wright, 1981; Ellis and
Rashid, 1984). Equilibrium concentrations offrom 0.25% toinexcessof8%
have been quoted. Although the absorption of water has, in a highly limited
sense, been shown to be beneficial in terms of both improved toughness and
static fatigue resistance (Gledhill et al. 1979), water absorption is generally
considered harmful. As far as the bulk epoxy polymer is concerned, water
can exert strong debilitating influences on mechanical properties such as
modulus, as well as reducing Tg. A reduction in Tg of 200 e per 1% of water
absorbed is a rule of thumb frequently applied for assessing the likely
influence of moisture on the Tg of an adhesive and, in particular, properties
associated with it (Wright, 1981). Since, as mentioned above, Tg can be
regarded as a measure of a polymer's high temperature capability, the
unexpected absorption of water into an adhesive layer, particularly one
designed for use at elevated temperatures, could have particularly severe
implications. The remarkably tough and peel-resistant epoxy-nylon
adhesives which, as explained previously, entered into being under a wave
of near euphoria, only to be brought down to earth with a bump following
realisation as to their poor moisture resistance, serve to emphasise the
importance of, at the very least, attempting to engineer into an adhesive
system, an element of hydrophobicity.
7.6.2 Adhesive joint mechanical properties

In this section a brief account will be given of the adhesive joint mechanical
properties which could be expected of a range of the more common types of
epoxy-based adhesives. Although the examples given cannot be regarded as
exhaustive, it is hoped that they will provide the reader with a feel for the
strength properties available with the various epoxy adhesive types.
Figure 7.13 shows some typical lap-shear strength data for three hot-cure
epoxy adhesive variants, in this specific case as a function of temperature.
Several features are worthy of mention. First, the lap-shear strength
attainable with epoxy adhesives toughened with elastomeric modifiers such
Test ~mperarure (to
Figure 7.13 Typical relationship between lap-shear strength (aluminium substrates) and test
temperature for three types of epoxy adhesive. (a) Rubber-modified; (b) one-part, general
purpose; (c) epoxy phenolic.
as the CTBNs can be regarded as impressive, with lap-shear values in excess

of 40 MPa at and below room temperature. Note how these values are in
excess of these which could be expected of typical general-purpose
unmodified epoxies, particularly at moderate test temperatures; this is a
direct effect of the improvements of toughness and hence peel resistance
brought about by rubber modification. The ability of epoxy-phenolics to
retain significant strength at temperatures up to and in excess of 250°C is
apparent. Although it is clear that significantly higher strengths can be
obtained at lower temperatures for the 'general-purpose' and 'toughened'
systems, in both cases temperatures in excess of 75-100°C cause a
precipitous strength decline. Since the epoxy-phenolic is relatively
unaffected by such temperatures, it clearly offers significant strength
advantages at temperatures in excess of about 100°C. Note in particular how
the epoxy-phenolic retains in excess of 50% of its room temperature
strength at about 250-300°C (Wake, 1982). With all three systems the short-
term high temperature capability will be related to Tg; epoxy-phenolics
clearly exhibiting substantially higher Tg values than the other two systems.
Water absorption, as mentioned previously, will generally result in a
reduction in Tg , the magnitude of the decline being largely dependent upon
the quantity of water absorbed. Thus, water absorption will generally lead to
a deterioration in high temperature load-bearing capacity. It should also be
stated that high temperature capability in this context refers simply to the
ability of adhesive joints to resist short-term high temperature excursions.
For longer-term applications the epoxy-phenolics are generally limited to a
high temperature ceiling of about 250°C, since oxidative degradation results
in a severe decline in mechanical properties at higher temperatures. For the
more common DGEBA-based adhesives, long-term use at temperatures in
excess of about 175°C is generally considered inadvisable.
Room temperature cure epoxy adhesives usually provide properties
somewhat inferior to those indicated in Figure 7.13, particularly regarding
the ability to operate at elevated temperatures. Film adhesives often provide
superior strength characteristics in comparison to their liquid/paste counter-
parts since they generally contain higher molecular weight components
which result in improvements in toughness and peel resistance.
In addition to the use of fracture mechanics as a means of studying the
fracture characteristics of bulk epoxy systems, it has also been employed
extensively to study the toughness and crack propagation characteristics of
adhesive joints, as outlined in section 7.4. Much of this work has been
concerned with assessing the influence of joint geometry on the toughness
and crack growth behaviour of both unmodified and rubber-modified epoxy
adhesives together with attempting to relate bulk to adhesive joint fracture.
Joint geometry within this context refers primarily to the effects of bond
thickness, i.e. thickness of the adhesive layer, and mode of loading.
The effect of bond thickness on the toughness and crack growth
characteristics of unmodified epoxy resin adhesive Jomts has been
considered by various workers. Mostovoy and Ripling (1971) for example,
employed aluminium tapered double cantilever beam specimens bonded
with amine- and anhydride-cured epoxies to study bond thickness effects.
They found that, in the majority of cases, adhesive Gic was virtually
independent of bond thickness from approximately 0.05 to 0.5 mm, i.e. the
common range of thickness traditionally employed in practice. In addition
the values of Gic obtained were usually similar in magnitude to those found
for their 'bulk' counterparts. The same has not been found to be true for
rubber-modified epoxies where considerably more complex behaviour has
been observed in a number of studies (Bascom et at., 1975; Kinloch and
Shaw, 1981). This complexity is demonstrated in Figure 7.14 which shows
the influence of bond thickness on the adhesive fracture energy of a rubber-
modified epoxy adhesive formulation and its unmodified counterpart. The
relative simplicity exhibited by the latter is clear with, as previously
mentioned, no significant effect of bond thickness on toughness. With
the rubber-modified system the dramatic improvement in toughness is
immediately apparent. However, perhaps of greater interest is the
pronounced influence that bond thickness has on fracture energy. As
indicated, this parameter is shown to pass through a maximum value, which
we can call GIcm at a certain bond thickness, tm . At thickness beyond tm , Gic
undergoes a reduction until a value is reached which remains essentially
constant with increased thickness. This constant Gic has been shown to be
similar, if not identical to the Gic values obtained from bulk adhesive
specimens. Reasons for this bond thickness effect have been discussed in
c;< _ ~IC,: o~
:1\
4
E
8 3
I
M: \
l')
~ : 0---0- (a)
o tm
/
o
L__ ~~~~~==~I~~~e~=-~~~==~====~.~=-~(b~)
1 2 3
Bond thickness, t ,(mm)
Figure 7.14 Influence of bond thickness on the adhesive fracture energy of a piperidine cured
DGEBA epoxy. (a) Rubber-modified system containing 15 phr CTBN; (b) unmodified.
(--to
Volume of plastic deformation
Maximum
Decreasing,j, Decreasin~
More festriction I Less constraint
Bond Thid<ness, t ~
t -Value Degree of constraint Plastic-zone G1c(Joint)

due to bond thickness shape
Corrpcrative!y
t.«2Fy) High ?'('I)I!/I/1jj Low
tm( = 2ry) Moderate ~!Q At Maximum
tb(>2ry) Low ~ Below GICM
Approx Equal
tc(»2ry) Almost Nil =@ To GIG Bulk
Figure 7.15 Crack tip plastic zone restriction/constraint mechanism.
terms of what can be described as a crack tip plastic zone restriction/

constraint mechanism (Kinloch and Shaw, 1981). The theoretical basis of
this mechanism is demonstrated in Figure 7.15. As shown, the volume of the
crack tip plastic zone can be regarded as a main factor responsible for the
bond thickness-adhesive GIc relationship. Due to the relatively large crack
tip plastic zones which can develop with rubber-modified epoxi~s, a
situation can exist whereby the dimensions of this zone will be of a similar
magnitude to the bond thicknesses shown in Figure 7.14. Thus, at bond
thicknesses below a critical value (equivalent to the plastic zone diameter),
restriction of the development of this zone in the bond thickness direction is
likely to result in a reduction in plastic zone volume. In addition, both finite
element analysis (Wang et at., 1978) and more recent experimental
observations have shown that the degree of constraint imposed upon an
adhesive layer can dictate to a large degree the distance over which the
principal stresses responsible for the plastic zone exist. For example,
increased constraint brought about by a reduction in bond thickness, has
been shown to produce an extension in the 'length' of the crack tip plastic
zone in a rubber-modified epoxy adhesive, which in turn will result in
increased plastic zone volume.
Because of these two competing effects, the maximum volume of the
plastic zone, and hence a maximum G Jcm , will occur when the maximum
degree of constraint exists, at a given bond thickness, commensurate with
the condition that no restriction on the development of the plastic zone from
the metallic substrate exists. Studies conducted with a rubber-modified
epoxy over a range oftest rates and temperatures (Kinloch and Shaw, 1981)
have shown that this condition generally occurs when bond thickness is
equivalent to the plane stress plastic zone diameter, depicted in Figure 7.15
as tm = 2ryo At bond thickness below tm, although a high degree of constraint
will be imposed upon the adhesive layer thus tending to increase plastic zone
volume by extension along the adhesive layer, the competing influence of
plastic zone restriction in the bond thickness direction will also occur,
resulting in an overall reduction in plastic zone volume and thus GJc. At bond
thickness in the vicinity of tm' the GJc of the adhesive layer will be greater
than that obtained from its bulk counterpart due simply to an enhanced
plastic zone volume brought about by the constraint effect. As bond thick-
ness is increased beyond tm' restriction in the bond thickness direction will
no longer apply since t > 2ryo At the same time constraint on the adhesive
layer will continually decrease resulting in a reduction in plastic zone length
and therefore volume, which in turn will cause GJc to fall until a bond
thickness is reached whereupon GJc of the adhesive layer is similar to that
obtained from bulk adhesive.
Adhesive joint fracture under mode II and mode IIII combinations has
also been studied but to a far more limited extent (Bascom and Oroshnik,
1978; Mai and Vippond, 1978). Table 7.1 shows typical fracture energy
values for similar unmodified and rubber-modified epoxy formulations
obtained under mode I, mode II and combined mode IIII conditions. As
expected, the unmodified adhesive provides the lowest fracture energy
under all the loading conditions employed, particularly mode I. Although
the introduction of a mode II component increases fracture energy by a
Table 7.1 Effect of loading mode on adhesive joint fracture energy
Adhesive Fracture energy (kJ . m-2)

a
GIc G"IIc
30° 45° 60°
DGEBA-amine 0.07 0.15 0.13 0.15 1.45

CTBN-modified DGEBA 3.40 0.31 0.50 0.31 3.55
aObtained from scarf-joint specimen; b obtained from independently loaded

mixed-mode joint.
factor of approximately 2, it is not until virtually all of the mode I

component is removed that a substantial improvement in fracture energy
occurs. The rubber modified adhesive, as shown, once again exhibits more
complex behaviour. The mixed mode 1111 values are substantially lower than
either mode I or II values with the two extreme loading conditions providing
approximately equal fracture energy values. Of particular interest is an
apparent maximum in the 1111 fracture energy value at a taper angle of 45°;
an observation which has been addressed by Bascom and Orshnik (1978).
7.7 Environmental effects
7.7.1 Introduction
A major problem with the use of structural adhesives concerns the adverse
effects which certain environments can have on the mechanical properties of
the joint. Notable amongst these include temperature extremes, stress and
radiation. However, the most severe problem has been associated with the
influence of atmospheric moisture. Experience over many years, gained
from both accelerated tests and long-term exposure to warm/moist
environments, has shown conclusively the potentially disastrous results
which can result from the use of adhesively-bonded components in moist
environments (Hockney, 1970, 1972, 1973; Butt and Cotter, 1976). To
demonstrate the potential magnitude of this effect, Figure 7.16 shows the
influence of water immersion at a temperature of 60°C on the strength of
mild steel butt-joints. The adhesive formulation employed was prepared
from a diglycidylether of bisphenol A (DGEBA) cured using a tertiary
amine curing agent. Clearly this environment has a detrimental effect on the
strength of this particular joint, 1500 hours immersion resulting in only a
15% retention of strength. Additionally, the immersion conditions
indicated were also found to influence locus of failure with a pronounced
change from cohesive within the adhesive layer to apparent interfacial
occurring on increasing immersion time. Thermodynamic considerations
can, in fact, be used to predict such interfacial behaviour as will be described
shortly.
In this necessarily limited account an attempt will be made first to
highlight the major parameters which, from experience, have been shown to
influence joint durability, second to discuss the mechanisms which have
been deemed responsible for the adverse effects of moisture and third to
consider the various approaches which have been considered in an attempt
to improve the environmental resistance of epoxy adhesive bonded joints.
Finally, a brief discussion concerning other environmental factors such as
elevated temperature performance will be given.
40
-
CJ
~
~
CI
c:
QI
L..
+-
VI
+-
c:
·0 0
.-'
I
+-
+-
::J
CO
10
O~------~5~OO~------1~OO~O------~~~
Immers;OI1 ti~, "•.o.6o·c(hl

Figure 7.16 Influence of water immersion at a temperature of 60°C on the strength of epoxy-
bonded mild steel butt-joints.
7.7.2 Moisture-related effects

As previously discussed, experience has shown that water generally present
as atmospheric moisture, can exert a serious debilitating effect on epoxy-
bonded joints. In addition, whilst the locus of failure of well prepared joints
is usually cohesive within the adhesive layer, environmental exposure
frequently focuses failure towards the interfacial zone between the adhesive
and substrate.
Numerous factors have been shown to influence joint durability,
induding:
(i) Environment. As mentioned, water is by far the most hostile environ-
mental agent which a structural adhesive joint can experience.
Generally factors such as temperature and applied stress together
with the amount of water available (e.g. relative humidity), usually
exert a significant influence; increases in the magnitude of each
generally resulting in an increased degree of environmental attack.
Interestingly, a number of studies have indicated that, although
exposure of adhesive joints to high humidity usually results in
strength degradation, many joints appear able to withstand exposure
to lower humidities for extended periods of time without any signifi-

cant strength reductions. Although some researchers have argued
that this signifies the existence of a critical concentration of water
below which environmental attack will not occur (Gledhill et al.,
1980) others have suggested that this is merely a time-related effect.
(ii) Influence of adhesive type. It is now clear that the chemical nature
of a structural adhesive can influence quite dramatically both the
extent and rate of environmental damage. For example both vinyl-
phenolic and nitrile-phenolic adhesives have been shown to exhibit
long-term durability superior to that shown by epoxy-based adhesives
(Bolger, 1973; Buck and Hockney, 1973). In fact one particular
class of epoxy adhesive, the nylon-modified epoxies, have been shown
to suffer serious degradation when subjected to warmmoist environ-
ments (Cotter, 1977) and are now generally considered inappropriate
for humid service conditions. Althought relatively inferior in terms
of durability, epoxy-based adhesives are nowadays preferred due to
a combination of improved bonding characteristics (lower cure
temperature/pressure) and superior peel strengths.
(iii) Substrates. Adhesive joints constructed from most substrate types,
be they metallic or polymeric in nature, can and frequently do exhibit
susceptibility to environmental attack. The mechanisms by which this
degradation occurs can vary significantly. As will be discussed
(section 7.3), these differences in behaviour influence the methods
by which improvements in environmental stability can be imparted.
(iv) Surface treatment. Choice of surface treatment for a particular sub-
strate has, in many cases, been shown to be of critical importance in
imparting acceptable levels of durability with epoxy-based adhesives.
In addition, the use of primers, of various types, has been found
conducive to good moisture resistance as will be discussed.
7.7.3 Failure mechanisms

Water can influence the load-bearing capacity of an epoxy-bonded joint in
several ways. These can include the following:
(i) It can affect the substrate, either chemically (observed with certain
metallic alloys such as aluminium or titanium) or by physical modi-
fication. The latter has been shown to be applicable to various
composite substrates such as carbon fibre-reinforced epoxies (CFRP).
(ii) It can exert an effect on the adhesive layer by chemical modification
(e.g. hydrolysis) or by physical damage (e.g. microcracking), both
of which can be regarded as irreversible. In addition a reversible
plasticisation can occur, influencing in particular the glass transition
temperature, Tg , and properties associated with it. As mentioned in
section 7.1, the absorption of moisture into an adhesive layer can

result in a not inconsiderable reduction in Tg which can itself result
in a substantial reduction in load-bearing capacity at elevated
temperatures.
(iii) Water can influence the interfacial region of the substrate and
adhesive, resulting in an adhesive de-bond which in turn can cause a
virtually complete loss of load-bearing capacity.
Although the effects outlined in (ii), namely bulk modification of the
adhesive layer, can and indeed have been observed with epoxy-bonded
joints, they are not usually responsible for incidents of environmentally-
induced failure. Rather, it is the interfacial regions which are generally
affected by atmospheric moisture, and upon which our attention must be
focused. Two main mechanisms, which can be regarded as complementary,
have been proposed to account for observations of environmentally-induced
interfacial failure. The first, based upon an assessment of adhesive/metal
oxide interface stability from a thermodynamic standpoint, has been shown
to be applicable to steel-epoxy bonded joints. The second, concerned with
elements of metal oxide stability, has been shown to be particularly
pertinent to epoxy-bonded aluminium alloys.
7.7.3.1 Interface stability. The stability of an interface between, for

example, a metal oxide and an epoxy adhesive, can be predicted on
thermodynamic grounds. In a dry environment, the work needed to
separate two phases of unit area in contact can be related to relevant surface
free energy parameters by the Dupre equation.
WA = Yx + Yy - Yxy (7.6)
where WA is the thermodynamic work of adhesion, Yx and Yy are the surface
free energies of the substrate and adhesive phases repectively and Yxy the
interfacial free energy. In the presence of a liquid environment (denoted by
the suffix L), the work of adhesion is
WAL = YxL + YyL - Yxy (7.7)
For adhesive metal oxide interfaces existing in an inert atmosphere, e.g. dry
air WA usually exhibits large positive values indicating thermodynamic
stability of that interface. However, in water WAL is generally negative
indicating interface instability. Some typical values of both WA and WAL are
shown in Table 7.2. For both the metal oxides indicated the change in work
of adhesion from a positive to a negative value on going from a dry to a wet
environment essentially predicts that water would act to displace the epoxy
from the oxide surface thus promoting both a reduction in load-bearing
capacity and an increase in the degree of apparent interfacial failure
observed on failed surfaces. These locus of failure predictions have in fact
Table 7.2 Values of work of adhesion W A (inert environment)

and WAL (in water) for various interfaces
Interface
Epoxy/ferric oxide 291 -255

Epoxy/aluminium oxide 232 -137
Epoxy/CFRP 90 44
been confirmed, at least in the case of epoxy/mild steel joints, using modern
surface specific analytical techniques such as Auger and X-ray
photoelectron spectroscopy (Gettings et al., 1977). The epoxy/CFRP
interface is of particular interest in that both W A and W AL exhibit positive
values indicating interface stability in both dry and wet environments.
Indeed, as mentioned above, this stability is reflected in the locus of failure
generally observed when CFRP joints are aged in hot/moist environments;
the CFRP substrate and/or the bulk adhesive generally being influenced by
water rather than the interface.
7.7.3.2 Metal oxide stability. Studies have shown that, with the surface
pretreatments that have been and are currently employed with aluminium
alloys, the surface oxide layer is generally regarded as the 'weak-link' in the
presence of moisture, as opposed to the interface and adhesive. As
described in section 7.3, the three usual methods for pretreating aluminium
are chromic-sulphuric acid etching, chromic acid anodising and phosphoric
acid anodising. Work conducted by Noland (1975) using X-ray photo-
electron spectroscopy, showed that the oxide produced on aluminium alloys
by the chromic-sulphuric acid etch was unstable in the presence of moisture
changing to a weaker hydrated form. Furthermore, Noland observed that
epoxy/aluminium joints pretreated by this method exhibited an apparent
interfacial failure when the joints were exposed to hot/humid conditions.
However, XPS showed that failure had occurred in fact in the weak hydrated
oxide layer. Although using the thermodynamic arguments previously
described, interfacial instability would be expected between aluminium
oxide and an epoxy adhesive, it has been suggested that the apparent
discrepancy between this thermodynamic prediction and the observed locus
of failure can be reconciled by the existence of a mechanically interlocking
component (Schmidt and Bell, 1986). The three surface treatments
mentioned, particularly the anodising techniques, have been shown to
produce surface oxide layers exhibiting topographical characteristics
intuitively consistent with the ability to promote mechanical interlocking.
7.7.4 Approaches to improved durability
Two main approaches can be considered as a means of alleviating the
undesirable effects of atmospheric moisture with epoxy bonded metal
joints. These are:
(i) Improve the moisture resistance of the most vulnerable part of the
joint, i.e. the interfacial zone.
(ii) Employ epoxy adhesive formulations which exhibit hydrophobic
characteristics which would restrict the transport of large concen-
trations of water to the vulnerable areas of the joint (Shaw et al.,
1988).
Both of these approaches will now be considered.
7.7.4.1 Improving the moisture resistance of the interface. As will be

apparent from the previous discussions, the approach to interfacial stability
will clearly depend upon the precise location of the failure path. Experience
has shown that, with epoxy-bonded mild steel joints, with the substrates
being pretreated using a simple degrease/grit blast procedure, failure in the
presence of moisture is usually interfacial in nature (Gledhill and Kinloch,
1974) the water literally causing a de-bond between the epoxy adhesive and
the mild steel substrate. Thus, in this case it is necessary to focus attention
upon the epoxy-oxide interface. Many of the studies conducted within this
context have been concerned with the use of coupling agents employed as
pre-bond primers.
7.7.4.2 Coupling agents. The majority of coupling agent studies have

been devoted to one class of material, namely organosilanes. These
materials were first recognised as having potential for use in adhesively-
bonded structures following successful exploitation in glass-reinforced
plastic (GRP) structures, where their incorporation between the glass
reinforcement and polyester matrix was shown to provide substantially
enhanced composite strength retention in moisture laden environments
(Plueddemann, 1982). Following this recognition a number of investigations
have confirmed this potential (Kinloch et al., 1975; Gledhill et al., 1990).
Most commercially available organosilanes are based upon the following
generalised structure
R-Si-(X)3
Where X is a hydrolysable group (considered necessary for interaction with

a substrate) and R is an organo-functional group capable of some form of
interaction with the intended polymeric component (Plueddemann, 1982).
Experience in the use of organosilanes has suggested that they perform
best when applied as very dilute solutions. Thus areas which have received
specific attention have included the influence of solvent type, solution age,
film drying conditions as well as, of course, the chemical nature of the
coupling agent. As far as the last is concerned, by far the majority of studies
and indeed eventual use in service, has been devoted to one specific
organosilane, namely y-glycidoxypropyltrimethoxysilane, the structure of
which is shown below.
/0"
CH2-CH-CH 2-O-(CH2h-Si-(OCH3h
Before briefly describing some of these studies it is perhaps worth while to

mention that this particular organosilane has been chosen for use with
epoxies due primarily to the presence of the terminal glycidoxy group; the
idea here being that reaction between this group and epoxide groups
contained within the adhesive would result in a covalent bond linking the
adhesive with the organosilane, which in turn could undergo reaction with
the oxide surface via the methoxy groups. Such a 'covalent bridge' linking
adhesive to substrate has been viewed by many adhesion scientists, the
author included, as providing greater environmental stability to the
interfacial region than one consisting purely of secondary force interactions.
Whether such a mechanism exists in practice has been a subject of con-
siderable debate for many years. We will return to this subject later in this
section.
Returning to the experimental variables mentioned above, Figure 7.17
non-silane joint
Solution age (m;~)
Figure 7.17 Effect of organosilane solution age a'nd solvent type on the strength of epoxy
bonded mild steel butt-joints following immersion in water at 60°C for 1500 hours (aqueous
solution) and 1000 hours (ethanol-based solution).
shows the effect of two of the variables mentioned, namely solvent type and
solution age on the durability characteristics of mild steel butt-joints bonded
using a simple tertiary amine cured epoxy adhesive. Although rather
complex, this figure demonstrates several interesting features.
First, for joints pretreated prior to bonding with a 1% v/v aqueous
solution, the substantial improvement in moisture resistance over joints not
treated with organosilane is clear; the treated joints exhibiting joint strength
values of between 17.5 and 34.5 MPa following 1500 hours water immersion
at 60°C, in comparison with 5.8 MPa for the untreated joints. Second, a
maximum in joint strength is apparent over aqueous solution ages of
between 30 and 90 minutes. Third, changing from an aqueous to an ethanol-
based solvent system results in striking differences in behaviour. These can
be itemised as follows:
(a) The ethanol based solutions exhibit no apparent dependence upon
solution age.
(b) Joint strength values obtained from the ethanol system are sub-
stantially lower than their aqueous solution based counterparts.
(c) Joint strength values obtained from the ethanolic organosilane
solutions are virtually identical to those obtained from the untreated
joint.
Although such studies have provided clear evidence of the substantial
improvements in moisture resistance which can be conveyed to epoxy-
bonded joints by organosilanes it is clear that various experimental
parameters are of vital importance so as to avoid misleading and
pessimistically false impressions of their durability-enhancing charac-
teristics. Our inability to predict precisely the appropriate optimum
molecular and experimental parameters can be attributed, partly at least, to
the fact that at present no universally acceptable mechanism has been
proposed to describe their action. However, one particular theory, the
chemical bonding theory, has generally received the most widespread
support amongst adhesion specialists and thus a brief description here would
not be out of place.
The chemical bonding theory is the oldest, best known and possibly the
most easily understood of all the theories that have been proposed to
account for coupling agent action (Plueddemann, 1982). The concept is
strikingly simple, namely that in place of the predominantly secondary force
interactions that would exist between a high energy substrate, such as a
metal oxide, and an epoxy adhesive, a suitably employed coupling agent
would act as a 'covalent bridge' linking together the organic and inorganic
components of the joint. This covalent linking would be expected to provide
far greater moisture resistance than mere secondary force interactions
alone. The sequence of reactions which could be expected in the formation
of an interfacial 'covalent bridge', starting from the initial organosilane, can
be viewed in the following manner:

(a) Hydrolysis of the alkoxy groups on the organosilane molecule
resulting in an organofunctional silanol.
R R
X-~i-X
I
~ HO-Si-OH
I
+ 3XOH
X OH
(b) Condensation of hydrolysed silane, i.e. reaction of organofunctional

silanol hydroxyls with hydroxyl groups on the substrate surface (e.g. metal
oxide) and interaction of silanol hydroxyls to form polymeric networks.
R
I metal oxide
H 0 - Si-O H + ""'("""'"
()H OH OH
metal oxide
~
oI OH I
0
I
OH
I
R-Si-O-Si-O-Si-O-Si-R
I I I I
ORR 0
1. I.
-SI- -SI-
I I
(c) Reaction/interaction of the organofunctional group, R, with the

epoxy adhesive could then be envisaged as the next and final stage in the
formation of the 'covalent bridge'.
A substantial amount of evidence has accumulated over the years in
support of this theory. Gettings and Kinloch (1977) employed XPS and
secondary ion mass spectroscopy (SIMS) to study the interactions
between an organosilane film (derived from an aqueous-based solution of
y-glycidoxypropyl-trimethoxy silane) and a mild steel surface. Results
obtained from SIMS analysis showed the existence of species such as FeSiO +
radicals, which Gettings and Kinloch justifiably attributed to the presence of
Fe-O-Si bonds within the interfacial zone; this clearly providing evidence
for the existence of a chemical bond between the metal oxide and the
organosilane layer. In addition, infrared spectroscopy has indicated
evidence of silanol condensation between various organosilanes and high
surface area particulate silica compounds (Plueddemann and Collins, 1975).
Furthermore, Fourier transform infrared spectroscopy has made possible a
direct study of silanol reactions with lower surface area glass fibres (Ishida
and Koenig, 1978). Such studies have shown clear evidence of silanol
condensation reactions between organosilanes and surface hydroxyls.
Covalent bonding between the organosilane and the polymer component
has been more difficult to demonstrate, although considerable indirect
evidence has indicated that such reactions can occur. For example in studies
conducted as early as 1962, Plueddemann, whilst evaluating a wide range of
coupling agents in glass-reinforced polyester and epoxy composites, found

that the effectiveness of the organosilane paralleled the reactivity of its
organofunctional group with the resin (Plueddemann et al., 1962).
Although such evidence has indicated the existence of interfacial
reactions and possibly 'covalent bridge' formation, a number of important
observations have provided cause for doubt. Although organosilanes would
appear capable of forming oxane bonds (M-O-Si) with various oxide
surfaces where M=Si, Ti, AI, Fe, etc., it is not obvious that such bonds
should contribute outstanding water resistance to the interface since oxane
bonds between silicon and iron or aluminium, for example, are not
resistant to hydrolysis. Yet, as demonstrated in this section, substantial
evidence is now available indicating the benefits to be gained from
organosilanes. In addition, the direct vulcanisation of rubber against
organosilane treated glass has been shown to produce covalent bonds at the
interface with, however, poor adhesion characteristics (Plueddemann and
Collins, 1975). Also certain organofunctional silanes have shown good
coupling abilities with inert polymers such as polyalkanes, where no
chemical reaction could be demonstrated nor would be expected
(Plueddemann, 1982). These discrepancies have resulted in other theories
such as those based upon wetting and surface energetics, deformable and
restrained interfacial layers. However, a variant of the chemical bond
theory, the reversable hydrolysable bond theory has been proposed by
Plueddemann (1982) which explains many of the discrepancies outlined
above. With this theory, chemical bond formation is once again the main
feature. However, the inherent vulnerability of oxane bonds to the effects of
moisture is accounted for by the proposed ability of the hydrolysis reaction
products to reform and reproduce oxane bonds, as shown in the following
simple hydrolysis equilibrium reaction.
M-O-Si + H 20 ~ M-OH + HO-Si-
Clearly for an equilibrium condition such as this to be maintained, the

adhesive contributing to the interfacial zone must have the ability to hold
silanols at the interface. Equilibrium conditions would be lost if silanol
groups resulting from hydrolysis were to be physically removed from the
interface. Consequently the ability of an organosilane to promote increased
moisture resistance would depend as much on the morphology it helps to
control in the interfacial zone as its ability to develop primary covalent
bonds across the adhesive-substrate interface. The development of a
rubbery zone would result in its retraction from the interface as silanol bonds
are released through hydrolysis. Water diffusing through the polymeric
phase would then force the adhesive further from the substrate surface until,
eventually, complete loss of adhesion occurred. Conversely, silanols formed
at a rigid interface would not be capable of moving beyond the dimensions of
molecular segments and would, therefore, be in a position to reform oxane
bonds, with the original or adjacent active sites on the substrate surface.
Although the reversible hydrolysable bond mechanism still fails to explain

all of the effects observed with organosilanes, it is the theory currently
favoured by most adhesion scientists, the present author included.
Although much of the research described has been concerned with
organosilanes employed as pre-bond primers, i.e. applied to substrate
surfaces prior to epoxy adhesive application and bonding, it is fairly
common commercial practice, at least with some manufacturers, to
incorporate organosilanes into epoxy adhesives. Since, as expressed in this
account, organosilanes are essentially 'interface active' materials, this
formulatory approach would clearly rely on the ability of the organosilane
molecules to diffuse to the interfacial zone and undergo the necessary
interfacial reactions prior to cure of the adhesive. Although a number of
researchers have harboured doubts as to the likelihood of such processes
occurring in practice, it would seem unlikely that manufacturers would
include coupling agents in adhesive formulations without evidence of a
beneficial contribution to adhesive joint durability.
In addition to organosilanes, other types of coupling agent have been
considered for use with epoxy adhesives. Most notable amongst these have
been organotitanes and zirconates (Monte and Sugarman, 1980), although
to date they have not achieved the prominence enjoyed by organosilanes.
7.7.4.3 Metal oxide stability. As previously mentioned (section 7.3.3.2),

surface specific analytical techniques have shown that the relatively poor
environmental (hot/wet) performance of aluminium joints pretreated prior
to bonding using the chromic sulphuric acid etch, is due primarily to the poor
hydration resistance of the surface oxide layer. Although this surface
pretreatment has been shown to produce generally acceptable results when
employed in conjunction with phenolic-based adhesives, the use of epoxy
adhesives has resulted in relatively poor performance.
Two anodising surface treatments are now generally employed for
pretreating aluminium alloys; namely phosphoric and chromic acid
anodising (section 7.3). Both have been shown to provide superior hot/wet
performance to the etching procedure. These differences in behaviour have
been attributed to two main factors. Firstly, Noland (1975) using XPS and
Ahearn et al. (1983) using ellipsometry, demonstrated a pronounced
improvement in hydration resistance of the oxide produced by phosphoric
acid anodising compared with the chromic-sulphuric acid etch. It has been
proposed that this improved stability is associated with the presence of an
aluminium phosphate layer on the anodised oxide (Venables, 1985). As
mentioned in section 7.3.2, a mechanical interlocking contribution to the
adhesion of epoxy aluminium joints has been proposed. Experiments
described by Poole and Watts (1985) have, in a sense, confirmed this view as
indicated in Table 7.3. As shown, a grit blast prior to chromic-sulphuric acid
etching and chromic acid anodising results in a substantial reduction in
apparent interfacial crack propagation.
Table 7.3 Wedge test crack propagation data obtained from 120°C
curing epoxy-bonded aluminium alloy joints
Surface treatment Crack growth (mm) during exposure to

SO°C/96%RH
1h Sh 24h 96h
CAE/CAA 3.9 13.1 29.3 3S.S

GB/CAE/CAA 2.3 3.7 6.2 6.9
CAE, chromic acid etch; CAA, chromic acid anodise; GB, grit
blast.
The use of primers in addition to the aluminium anodising surface

treatments is now regarded as providing optimum environmental resistance.
Corrosion inhibiting primers comprising strontium chromate-epoxy
combinations are possibly the most common.
However, there has been a desire in recent years to replace chromium-
containing primers with more environmentally acceptable formulations.
Recent work by Pike and co-workers (1988) has focused on inorganic primer
systems based on amorphous inorganic aluminium oxides such as aluminium
alkoxide. Promising results have been obtained allowing potential for
eliminating the use of chromium-containing organic primers for corrosion
protection.
The correct choice of surface pretreatment has also been shown to be of
vital importance for achieving moisture resistance with epoxy-bonded
titanium alloys. Figure 7.18 shows crack growth data obtained from epoxy-
bonded wedge test specimens where the titanium substrates were subjected
to the surface treatments outlined in section 7.3; i.e. alkaline/hydrogen
peroxide etch, chromic acid/fluoride anodise and sodium hydroxide
anodise. A simple grit blast treatment is also shown for comparison. The
superior performance of all three chemical treatments in comparison to the
grit blast is clearly apparent, with the sodium hydroxide anodise providing
marginally superior behaviour to the chromic acid/fluoride treatment.
7. 7.4.4 Reduction in moisture absorption. Since the previous discussions

have focused attention on the debilitating effects of water on adhesive joint
behaviour, particularly at the interfacial zone between substrate and
adhesive, it seems intuitively reasonable that a means of preventing water
reaching these critical regions in large concentrations would enhance
moisture resistance. Indeed studies conducted in particular by Comyn and
co-workers (1981) have demonstrated a linear relationship between joint
strength and the total quantity of water absorbed by the adhesive.
One approach adopted by the current author and colleagues (Shaw et al.,
1988) has been to consider the concept of fluorination as a means of
(Il)
(b)
c
(d)
o 100 1511
Exposure time (hI
Figure 7.18 Influence of hot/humid atmosphere (50°C/96% Rh) on wedge test crack growth
characteristics of epoxy bonded titanium joints. (a) Grit blast; (b) alkaline/hydrogen peroxide
etch; (c) chromic acid/fluoride anodise; (d) sodium hydroxide anodise.
enhancing the hydrophobic characteristics of epoxy resins. In these studies

the following type of epoxy structure was employed,
which when cured with suitable curing agents, resulted in equilibrium water
concentrations as low as 0.2%. With regard to non-halogenated epoxies, the
amount of water the cured adhesive will absorb will depend upon many
factors including the type of resin and curing agent employed. In an
excellent review, Wright (1981) has shown that water uptake may differ by a
factor of ten between resin types and by a factor of three with a similar resin
but different curative. In fact the amount of water absorbed by a cured
epoxy adhesive can vary enormously from about 1% to 10% with 2-6%
being fairly typical for a wide range of formulations. Fluoroepoxies of the
type demonstrated above can therefore be considered as substantial
improvements over the majority of 'conventional' epoxy systems with
regard to hydrophobic character.
Further analysis of these systems, so as to determine whether the
hydrophobic characteristics exhibited by bulk polymer are translated into

highly moisture-resistant adhesive joints, has not been conducted. It would
be of major interest to determine the existence or otherwise of such a
correlation.
7.7.5 Other hostile environments

In addition to atmospheric moisture, other potentially harsh environmental
conditions can include stress, low and high temperatures, radiation, and
various liquid contaminants, e.g. fuels, de-icing fluids, hydraulic oils (i.e.
those encountered in various aircraft). Although space does not permit
discussions about each of these interesting areas, it is perhaps pertinent to
discuss, very briefly, just one, namely epoxies as high temperature
adhesives.
The majority of epoxy adhesives are formulated to yield high temperature
limits only marginally greater than lOO°C. This is indicated in Figure 7.19
which shows the retention of room temperature strength as a function of test
temperature for a proprietary epoxy adhesive system (DICY-cured, rubber-
modified formulation). As indicated, this adhesives load-bearing capacity is
0
~ \0
\
!;.
£
@'
III
~70
IQI
0-
0
\0
\
f
~0
L.
'610
IS 0
1~
10 \ 0,--
0-
0 100 1 200 250
Test temperature (0 [ )
Figure 7.19 High temperature strength retention of a dicyandiamine cured, rubber-modified
epoxy adhesive.
greatly diminished with increasing temperature, with less than 10% room
temperature strength being retained at 140°C. This type of strength decline
can usually be related to the onset of Tg. Indeed Tg can, to a first
approximation, be considered as the high temperature limit for any
structural adhesive.
Improvements in high temperature capability can be obtained by
increasing the degree of cross-linking by a suitable choice of epoxy and/or
curing agent. The use of polyfunctional epoxies and curing agents can
produce cured products having Tg values in excess of 200°C, resulting in
substantial retention of properties at temperatures in excess of 150°C.
However, the propensity of epoxies to absorb, in some cases, considerable
quantities of water, which in turn can reduce Tg , together with the somewhat
brittle characteristics bestowed by the high cross-link density, has greatly
diminished this approach to high temperature capability. In addition, due to
their molecular structure, even the most highly cross-linked formulated
epoxy adhesives are unable to withstand long-term use at temperatures in
excess of approximately 175°C. For such applications other adhesive types
must be considered, including for example, the epoxy-phenolic-based
adhesives discussed in chapter 6 (section 6.2).
7.8 Applications
From humble beginnings in the 1940s and 1950s, epoxy adhesives are
nowadays employed in a diverse range of applications associated with a
number of important industries including aerospace, automobile, civil
engineering and electrical/electronics including of course the domestic home
market. Possibly the main impetus for their current popularity has been the
aircraft industry which has spearheaded both the development and
application of structural adhesives generally. Indeed, structural adhesives
were first used in aircraft construction in the early 1940s with a vinyl-
phenolic adhesive system being used to bond aluminium components on the
de Havilland Hornet aircraft. Since this time various adhesive types have
been used for a wide range of applications in both civil and military aircraft
(Alberici, 1983), with epoxy adhesives now enjoying prominence at the
expense of the older phenolic types.
Aircraft applications to date have included the bonding of reinforcing
doublers and stiffeners to both fuselage and wing panels; the former being
employed to reinforce holes for mechanical fasteners and window/door
openings. In addition, epoxy film adhesives are frequently employed for the
production of honeycomb sandwich structures where aluminium or Nomex
honeycomb is bonded to aluminium or composite skins. Such structures
have been employed to produce aircraft structures such as rudders, flaps,
elevators, ailerons, doors, floor panels as well as engine structures such as
nacelles and thrust reversers since they provide attractive combinations of

strength, stiffness and weight in addition to offering excellent fatigue and
sound insulating characteristics.
In addition to manufacture, epoxy adhesives, particularly cold cure
systems, are also employed in the repair of aircraft components which have
suffered damage due to accidents (ground impacts for example), lightning
strike, hailstones and bird impact.
Epoxy adhesives have also been used in the construction of helicopter
blades where their ability to confer enhanced fatigue resistance over the
more traditional mechanical attachments has been amply demonstrated.
Epoxies have also been employed in slightly more esoteric aerospace
applications such as spacecraft, where again they have been utilised for the
bonding of stiffeners and doublers to skins and also for honeycomb
structures. Additionally they have also been utilised in the manufacture of
rocket motor components for guided weapon applications.
In contrast to aerospace, the car industry has approached the use of
structural adhesives with a degree of caution. Although adhesive bonding is
employed in a number of manufacturing processes, the all bonded car,
where traditional joining techniques such as spot welding are totally
replaced by adhesive bonding, is not yet a reality. Where it has been
employed, the one-part heat-cured toughened epoxy pastes have
predominated with applications including use either in cosmetically critical
areas such as doors, boot lid or bonnet or in a combination with spot welds so
as to reduce their number or to help provide an increase in stiffness of the
vehicle.
Epoxy adhesives have also been employed in a range of civil engineering
applications where their use over the past thirty years has increased
considerably. Notable examples include the repair of cracks in concrete
structures such as dams and bridges where the adhesive is poured, brushed
or pumped under pressure into cracks. Epoxies have also been employed as
gap filling materials for incorporation between precast segmental concrete
units, as well as the bonding of external steel reinforcement to various
concrete structures.
References
Anderson, G.P., Bennett, S.J. and De Vries, K.L. (1977) Analysis and Testing of Adhesive
Bonds, Academic Press, New York.
Adams, R.D. and Wake, W.e. (1986) Structural Adhesive Joints in Engineering, Applied
Science, London.
Ahearn, J.S., Davies, G.D., Sun, T.S. and Venables, J.D. (1983) Adhesion Aspects of
Polymeric Coatings, ed. Mittal, K.L. Plenum, New York.
Albericci, P. (1983) Durability of Structural Adhesives, ed. Kinloch, A.J. Applied Science,
London.
Allen, K.W. and Alsalim, H.S. (1977) J. Adhesion 8,183.
Allen, K.W. (1979) Brit. Polym. J. 11,50.

Arnold, D.B. (1981) Developments in Adhesives - 2, ed. Kinloch, A.J. Applied Science,
London.
Bascom, W.D., Cottington, R.L., Jones, R.L. and Peyser, P. (1975) J. Appl. Polym. Sci., 19,
2545.
Bascom, W.D., Cottington, R.L. and Timmons, e.o. (1977) Appl. Polym. Symp. 32, 165.
Bascom, W.D. and Orshnik, J. (1978)J. Mats. Sci., 13,1411.
Bolger, J.C. (1973) Adhesion and Adhesives Vol. 3, ed. Patrick, R.L. Marcel Dekker, New
York.
Buck, B.I. and Hockney, M.G.D. (1973) Aspects of Adhesion-7, eds. Allen, K.W. and Alner,
D.J. Transcription Books, London.
Butt, R.I. and Cotter, J.L. (1976) J. Adhesion 8,11.
Comyn, J. (1981) Developments in Adhesives - 2, ed. Kinloch, A.J. Applied Science, London.
Cotter, J.L. (1977) Developments in Adhesives -1 , ed. Wake, W. e. Applied Science, London.
Cotter, J.L. and Mahoon, A. (1982) Int. J. Adhesion and Adhesives 2, 47.
Deryaguin, B.V. (1955) Research 8, 70.
Deryaguin, B.V., Drotova, N.A., Karasser, V.V., Kirillova, Y.M. and Aleinikova, LN.
(1957) Proc. 2nd Int. Congress on Surface Activity -Ill. Butterworths, London.
Deryaguin, B.V. and Smilga, V.P. (1969) Adhesion, Fundamentals and Practice, McLaren,
London.
Ellis, B. and Rashid, H.U. (1984) J. Appl. Polym. Sci. 29,2021-2038.
Gettings, M., Baker, F.S. and Kinloch, A.J. (1977) J. Appl. Polym. Sci., 21, 2371.
Gettings, M. and Kinloch, A.J. (1977) J. Mats. Sci., 12,2511.
Gledhill, R.A. and Kinloch, A.J. (1974) J. Adhesion 6,315.
Gledhill, R.A., Kinloch, A.J. and Shaw, S.J. (1977) J. Adhesion 9, 81.
Gledhill, R.A., Kinloch, A.J. and Shaw, S.J. (1979) J. Mats. Sci. 14,1769.
Gledhill, R.A., Kinloch, A.J. and Shaw, S.J. (1980) J. Adhesion 11, 3.
Gledhill, R.A., Shaw, S.J. and Tod, D.A. (1990) Int. J. Adhesion and Adhesives 10,192.
Griffith, A.A. (1920) Phil. Trans. Roy. Soc. A221, 163.
Hockney, M.G.D. (1970) RAE Tech. Rep. 70081; (1972) RAE Tech. Rep. 72100; (1973) RAE
Tech. Rep. 73016.
Huntsberger, J .R. (1967) Treatise on Adhesion and Adhesives, Vol. 1 ed. Patrick, R.L. Marcell
Dekker, New York.
Irwin, G .R. (1964) Appl. Mats. Res. 2,65.
Ishida, H. and Koenig, J.L. (1978) J. Coli. Int. Sci. 64,565.
Kaelble, D.H. (1971) Physical Chemistry of Adhesion, Wiley, New York.
Kinloch, A.J., Gledhill, R.A. and Dukes, W.A. (1975) Adhesion Science and Technology, ed.
Lee, L.H. Plenum, New York.
Kinloch, A.J. (1980) J. Mats. Sci. 15,2141.
Kinloch, A.J. and Shaw, S.J. (1981a) Developments in Adhesives - 2, ed. Kinloch, A.J.
Applied Science, London.
Kinloch, A.J. and Shaw, S.J. (1981b) J. Adhesion 12, 59.
Kinloch, A.J. and Young, R.J. (1983) Fracture Behaviour of Polymers. Applied Science,
London.
Kinloch, A.J., Shaw, S.l. and Hunston, D.L. (1983) Polymer 24,1341.
Lee, H. and Neville, K. (1967) Handbook of Epoxy Resins, McGraw-Hill, New York.
Mai, Y.W. and Vipond, A.S. (1978) J. Mats. Sci. 13,2280.
Marceau, J.A. and Scardino, W. (1975). Durability ofAdhesive BondedJoints, ADML TR-75-
3, Dayton, Ohio.
Marceau, J.A., Moji, Y. and McMillan, J.C. (1977) Adhesives Age 20,28.
May, C.A. (1988) Epoxy Resins, Chemistry and Technology, Marcel Dekker, New York.
McGarry, F.J. and Willner, A.M. (1968) Toughening of an epoxy resin by an elastomeric second
phase. Res. Rept. R68-8, Massachusetts Institute of Technology.
McMillan, J.e. (1979) Bonded Joints and Preparation for Bonding, NATO/AGARD Lecture
Series 102.
McMillan, J.e. (1981) Development in Adhesives - 2, ed. Kinloch, A.J. Applied Science,
London.
Moji, Y. and Marceau, J.A. (1976) US Patent 3,959,091.
Monte, S.l. and Sugarman, G. (1980) Proc. 35th Ann. Tech. Conf SPI, Paper 23-F.
Morgan, R.I. (1980) Developments in Reinforced Plastics -1, Applied Science, London.
Mostovoy, S. and Ripling, E.l. (1966) f. Appl. Polym. Sci. 10, 135l.
Noland, 1.S. (1975) Adhesion Science and Technology, ed. Lee, L.H. Plenum, New York.
Parker, B.M. (1983) Composites, 14,226.
Partridge, I.K. and Bucknall, e.B. (1983) Polymer, 24,639.
Pike, R.A. and Lamm, F.P. (1988) Adhesives, Sealants and Coatings for Space and Harsh
Environments, ed. Lee, L.H. Plenum, New York.
Plueddemann, E.P. (1982) Silane Coupling Agents. Plenum, New York.
Plueddemann, E.P., Clark, H.A., Nelson, L.E. and Hoffmann, K.R. (1962) Mod. Plast. 39,
136.
Plueddemann, E.P. and Collins, W.T. (1975) Adhesion Science and Technology, ed. Lee, L.H.
Plenum, New York.
Poole, P. and Watts, 1.F. (1985) Int. f. Adhesion and Adhesives 5,33.
Poole, P. (1985) Industrial Adhesion Problems, ed. Brewis, D.M. Orbital Press, Oxford.
Potter, W.G. (1970) Epoxide Resins, Iliffe, London.
Sage, G.N. and Tin, W.P. (1982) Composites, 13,228.
Schliekelmann, R.I. (1979) Bonded foints and Preparation for Bonding, NATO/AGARD
Lecture Series 102.
Schmidt, R.G. and Bell, 1.P. (1986) Advances in Polymer Science 75, ed. Dusek, K. Springer-
Verlag, Berlin, pp. 33-7l.
Sharpe, L.H. and Schonhorn, H. (1964) Advances in Chemistry Series 43, Gould, R.F. ed.
American Chemical Series, Washington.
Shaw, S.l. (1984) PhD Thesis, City University, London.
Shaw, S.l., Tod, D.A. and Griffith, 1.R. (1988) Adhesives, Sealants and Coatings for Space and
Harsh Environments, ed. Lee, L.H. Plenum, New York.
Shaw, S.l. and Tod, D.A. (1989) f. Adhesion 28, 23l.
Sherriff, M. (1976) f. Adhesion, 7,257.
Smith, T. and Haak, R. (1979) Treatment ofAm 355 Steelfor Adhesive Bonding. SC5169.16FR.
Ting, R.Y. (1988) Epoxy Resins, Chemistry and Technology, Ch. 5 ed. May, C.A. Marcel
Dekker, New York, pp. 551-60l.
Titow, W.V. (1978) Adhesion -2, ed. Allen, K.W. Applied Science, London.
Trantina, G.G. (1972) f. Composite Mat. 6, 192.
Venables, 1.D. (1985) f. Mats. Sci. 19, 243l.
Voyutskii, S.S. (1963) Autohesion and Adhesion of High Polymers, Wiley, New York.
Wake, W.e. (1982) Adhesion and the Formulation of Adhesives, 2nd Edn, Applied Science,
London.
Wang, S.S., Mandel, 1.F. and McGarry, F.l. (1978) Int. f. Fract. 14,39.
Wenzel, R.N. (1963) Ind. Eng. Chern. 28,988.
Wright, W.W. (1981) Composites 12, 20l.
Young, T. (1805) Trans. Roy. Soc. 95,65.
Young, R.I. (1980) Developments in Reinforced Polymers - 1, ed. Pritchard, G. Applied
Science, London.
8 Composite materials
F.R. JONES
8.1 Introduction
Composite materials technology exploits the high specific moduli and

strength of reinforcing fibres to produce low density high performance
structures. For engineering applications, the anisotropy of unidirectionally
reinforced materials is of crucial importance. It is therefore necessary to
develop techniques to align the fibres in the direction of the principal
stresses. For many applications of epoxy resins, this involves stacking
preimpregnated fibrous sheets (generally known as prep reg) to form
laminates. The fibres can have any number of geometrical arrangements but
a quasi-isotropic material might be obtained with the laminae arranged with
the fibres in the different plies at these orientations: O2°/ ± 45°/90 4°/
± 45% 2°. The subscripts refer to the number of individual plies. The same
laminate could be referred to as [0 2/ ± 45/902]s where s denotes a
symmetrical lay-up.
The other way of providing quasi-isotropy is to use woven reinforcements
or to combine differing fibre placement techniques. The most important
fibres are those based on E-glass, carbon and aramid polymer and are
available in various forms and with a wide range of properties. Over recent
times, the use of textile technology to weave or knit preforms for resin
impregnation has been developed. Short or discontinuous fibres are not
generally employed in epoxy resins because of their poorer reinforcing
efficiency compared with their continuous counterparts.
8.2 Fibre reinforcements
By comparing the mechanical properties of potential reinforcing fibres with

matrix and bulk materials (Tables 8.1 and 8.2), the benefits of fibre
composite technology become clear. Apart from steel, the fibres all have
significantly enhanced strength and stiffness. This point is made even more
pertinent by comparing their specific properties, in which strength and
modulus are normalised to their respective densities. In Figure 8.1, the
specific values of modulus and strength for each are plotted. Efficient
structures can be fabricated from materials with the highest values and in
Figure 8.1, it is seen that 'carbon fibre' and the intermediate modulus fibres
Table 8.1 Typical mechanical properties of reinforcing fibres
Relative density Young's modulus Tensile strength Failure strain Fibre diameter
(GPa) (GPa) (%) (urn)
Eglass 2.55 72 1.5-3.0 1.8-3.2 10-16

Sglass 2.5 87 3.5 4.0 12
S-2glass 2.49 86 4.0 5.4 10
Carbon-mesophase pitch 2.02 380 2.0-2.4 0.5 10
Carbon-PAN
Type A high strength 1.8 220-240 3.0-3.3 1.3--1.4 7
Type A high performance 1.8 220-240 3.3--3.6 1.4--1.5 7
Type A high strain 1.8 220-240 3.7 1.5-1.7 7
Type II high strength 1.8 250 2.7 1.0 7
Type I high modulus 2.0 330-350 2.3--2.6 0.7 7
Intermediate modulus 1.9 280-300 2.9-3.2 1.0 7
SiCO-continuous 2.5 200 3.0 1.5 10-15
SiC-whisker 3.2 480 7.0 1-50
Si C Ti C 0 continuous 2.35 200 2.8 1.4 8-10
Boron 2.6 410 3.4 0.8 100
a-alumina (FP) 3.9 380 1.7 0.4 20
{:I-alumina (Saffil) 3.3 300 2.0 0.5 3
d/{;l-alumina (Safimax) 3.3 250 2.0 3.1
'Y/ alumina (Safimax LD) 2.0 200 2.0 3.2
FesoB2o metallic glass 100 3.6
H.C. steel 7.8 210 2.8 250
Polyphenylene terephthalamide
Aramid - high modulus 1.47 180 3.45 1.9 12
Aramid - intermediate modulus 1.46 128 2.65 2.4 12
Aramid -low modules 1.44 60 2.65 4.0 12
Aramid - staple fibre (polymetaphenylene-
isophthalamide) 1.4 17.3 0.7 22.0 12
Polybenzothiazole (PBT) 1.5 250 2.4 1.5 20
Polyamide 66 1.44 5.0 0.9 13.5 =10
Polyethene (theory) 200
Solution spun 1.0 100--150
Drawn =1 6--80
Table 8.2 Typical mechanical properties of isotropic bulk materials
Young's Tensile Failure

Relative modulus strength strain
density (GPa) (GPa) (%)
Steel 7.8 210 0.34-2.1

Aluminium alloys 2.7 70 0.14-0.62
Glass 2.6 60
Resins:
Phenolic 1.4 7 "'0.5
Epoxy 1.2 2-3.5 0.05-0.09 1.5-6
Polyester 1.4 2-3.0 0.04-0.085 1.25
Thermoplastics:
Nylon 66 1.4 2.0 0.07 60
High density polyethene 0.96 1.3
Low density polyethene 0.91 0.25
in particular, together with the high performance polymeric fibres offer the
best solution.
Where light weight is more important than the cost of materials, such as in
aerospace and sports goods applications, these reinforcements dominate.
However, where economic considerations are more important, in industrial
applications, glass fibres still have much potential. For the polymeric fibres,
• Kevlar 149
200
.
Asbestos (crocidolile)
~
.c 150-
Kevlar 49.
~:' '"
~ /
[ll"
Carbon fibre
5>
c
i"
• S glass Boron (filament
on tungsten). \
;;;
• Si·Ti·C·O
~HM
~c
·S,·C·O
2 100
u
"
0.
(f)
• Nylon
• E glass
50
•
.Polyester .
• Steel (wire)
Acrylic
.
~b.~.~ Alumina
• Aluminium (bulk metal)
o 5 10
Specific modulus (Mm)
Figure 8.1 Specific properties of fibre reinforcements. XA, 1M, HM, refer to differing grades
of carbon fibre. For other definitions see Table 8.1.
COMPOSITE MATERIALS 259
PBT is still only available in sample quantities for strategic applications and
so aramid fibres such as Kevlar, Nomex and Twaron are the main
contenders. However, differing fibre and composite performance are more
important criteria for the selection of the grade of aramid or carbon
reinforcement.
It is important to recognise that the strength of a material is determined
by the presence of Griffith flaws which reduces the theoretical strength
significantly. The fibre form reduces the preponderance of flaws so that
filament strength is larger than that of the bulk material. However, the
strength of a filament is still determined by the presence of defects and as a
consequence the reported values are gauge length-dependent and are
usually average values determined from either unimpregnated or resin-
impregnated bundles. Individual filaments within the bundle will have a
different strength so that the statistics of fibre strength contributes
significantly to damage mechanisms of the composite. A typical distribution
of filament strengths is given in Figure 8.2. A further consequence of the
distribution of flaws on the fibre surfaces is the need for protection by a resin
coating or so-called size during handling immediately after manufacture.
For this application emulsified epoxy resins are commonly employed for
carbon and glass fibres.
8.2.1 Manufacture of carbon fibres from polyacrylonitrile (PAN)

precursors
The process consists of five major stages: (1) precursor spinning; (2)
oxidation; (3) carbonisation; (4) graphitisation; and (5) surface treatment.
1.0 ------- ------- - -",.,,...

0.'
0.8
0.7
5 0.11
I 0.11
OA ·
0.3
Q.2
0.1
0.0 +----I~-r_____,r__--,--,----r---,.--_.__-._-_r_-_r___;
~ 1~ 1~ 1~ 1~ ~ ~ ~ ~ ~ ~ ~ ~
SlREHGnt /0'"
Figure 8.2 The probability of failure of single E-glass filaments under tensile stress showing the
statistical aspects of strength.
The special acrylic precursor fibre is obtained by wet-spinning a carefully

filtered polyacrylonitrile copolymer solution. The degree of orientation of
the macromolecules during steam stretching has a profound effect on the
properties of the final carbon fibre.
Oxidation of the precursor fibre is carried out under restraint in oxygen
atmosphere at 200 to 220°C to maintain the molecular orientation induced
during stretching. Poly acrylonitrile is the favoured precursor because of the
self-induced thermal polymerisation of the pendent nitrile groups, to
produce a ladder polymer. Because of the limited number of adjacent
isotactic nitrile groups, the cyclised sequences involve approximately 5-15
groups. In addition, in areas where the molecules are not so well orientated,
cyclisation reactions between syndiotactic nitrile groups will cause the
ladder polymer to take on a curved rather than planar ribbon-like form. The
cyclised rings of the thermally-modified PAN are also puckered. It is now
known that these reactions precede the oxidation reaction and that oxygen
aids the aromatisation of the cyclised polymer into a planar aromatic
structure which facilitates condensation of the graphitic nuclei in the
carbonisation and graphitisation stages, which are carried out in an inert
atmosphere over the temperature range of 400 to 2500°C. In the original
work at the Royal Aircraft Establishment (now DRA-RAE), Farnborough,
UK, the tensile strength of the fibre reached a maximum at 1500°C, the
so-called type-II or high strength (HS) fibre (Watt, 1985). Graphitisation
at a temperature near 2500°C produced the type-lor high-modulus (HM)
fibre. Improvements in precursor and spinning technology have led to
improvements in fibre strength. This is illustrated in Figure 8.3 where clean
moo,,I
1500
C1l
D-
c:l
r:
c;,
:1 100
I
I
C1l
D-
I
filtered-clean c:l
c
---- --1300
t:
e! I
V)
::J
1ii strength -S
.!!i "0
0
'iii
c ~
<l>
l- normal
/
/
I I
1000 2000 3000
Heat treatment temperature °C
Figure 8.3 The effect of heat treatment temperature and clean room conditions on the
strength and modulus of PAN-based carbon fibres. (Redrawn from data in Watt, 1985).
room conditions have reduced the number of strength-reducing flaws. A
wide range of commercial fibres are now available of which the type-A fibre,
which has been heated to only 1000°C is often preferred. More recently,
intermediate modulus (1M) fibres have become available which combine
high strength with an improved modulus. A modulus approaching 800 GPa
(i.e. 80% theoretical) can be produced by hot-stretching techniques.
The microstructure of the carbon fibres is determined by the relatively
small size of the graphite nuclei and the curved ribbon-like molecular
structure of the oxidized acrylic fibre which yields a turbostratic graphite. A
skin-core structure is most pronounced for the high modulus fibres, as a
result of graphitisation at the high temperatures involved and the subtle
chemical differences in oxidised PAN fibre which arise from the diffusion-
controlled oxidation of the thermally cyclised polymer.
Surface oxidation is an essential post-carbonisation stage. This process can
be carried out chemically, thermally, or electrolytically, with the last being
the favoured method. The treatment is essential to the optimisation of the
adhesion of the fibre to the resin matrix and the performance of the
composite. If the surface treatment is too severe the composite material will
behave like a brittle monolith. Figure 8.4 demonstrates the need for a
carefully controlled surface oxidation to optimise the performance of the
1.5 ~--------------......,
(c)
. -' ~.-. _. _._--
0.5 I
100 300
DFT/%
Figure 8.4 A schematic representation of the effect of electrochemical oxidative surface

treatment on the properties of carbon fibre epoxy composites. (a) Interlaminar shear strength
of a 0° laminate; (b) impact strength; (c) notched tensile strength of a (0°/ ± 45%°), laminate.
(Redrawn from data given by Dunsford et al., 1981.)
composite. The interlamina shear strengths (normalised to the normal

industrial treatment and denoted 100%) at different degrees of treatment
are compared to the notched tensile strengths (also normalised to the value
for 100% treatment) for varying surface treatments. The impact strength of
a multidirectional composite reaches a maximum at commercial treatment
levels.
The nature of the adhesive bond between the fibres and matrix is still the
subject of controversy. This is largely because of the complicated surface
microstructure and microchemistry which originates from the treatment and
the techniques employed to assess the chemical differences and the true
value of the interfacial shear strength of the matrix. Most of the controversy
arises from attempts to deactivate an oxidised surface. Apparent removal of
oxygen moieties has not always led to a reduction in the fibre-matrix
interfacial shear strength. As a consequence some workers (Drzal et al. ,
1982) have maintained that mechanical interlocking is more important.
Significant increases in surface area are not achieved but the surface micro-
porosity is increased (Denison et aI., 1988). Recent work has confirmed
tha.t the surface oxygen concentration of oxidised fibres increases with time
after oxidation, probably by continued water adsorption. Furthermore,
de treated fibres (Denison et al., 1989) were shown to exhibit the same
acidity as present before thermal desorption of oxygen species. This was
interpreted as the presence of 'active' sites located in the micropores of the
high modulus fibres, which could be at 'unsaturated' carbon atoms or other
functional groups. Either the latter reform on a desorbed fibre surface or the
former can react with the matrix resins, during composite fabrication. The
micropores arise from erosion at intercrystallite boundaries on high
modulus fibre surfaces to provide a location for these active sites. For Type-
A fibres the edges of the basal planes probably run directly to the fibre
surface to provide reactive sites (Oberlin, 1985).
On balance, the weight of evidence appears to favour the formation of
chemical bonds to the resin matrix at the fibre surface. Most telling is the
number of observations which have demonstrated strong adsorption of
epoxy molecules. The functional groups most likely involved are amines,
carboxylic and hydroxyl groups. For the surface-treated fibres the last two
appear to be the most abundant species which account for = 50% of the
surface oxygen, with react ability to epoxy resins. At least 50% of the
remaining surface oxygen appears to come from adsorbed water.
8.2.1.2 Sizing resins for carbon fibres. To protect the fibres from damage
during handling and subsequent fabrication the fibres are dressed with an
epoxy resin size. This is usually applied from an emulsion, immediately after
the surface treatment. The role of these resins in the formation of the
interfacial bond has not been studied in detail but it is now recognised that
these may not have the same chemical structure as the resin matrix and the
presence of an interphase region needs to be considered.
8.2.1.3 Carbon fibres from pitch. High-strength carbon-fibre is also

manufactured commercially from petroleum pitch (Bacon, 1980; Rand,
1985). This involves the polymerisation of the condensed aromatic
molecules to form a mesophase or liquid crystal solution at temperatures
above 350°C for extrusion melt spinning into fibres for thermosetting in air
and carbonisation at temperatures around 2000°C. The main difference
between the two types of fibre lies in differing microstructures. The pitch
fibres tend to have a radiating crystal structure. The differences arise
because the crystal nuclei are formed prior to spinning and orientation of
these nuclei occurs during spinning. This results in fibres with a lower
transverse strength than for PAN-based fibres.
Commercial limitations have arisen from a variability in the pitch
precursor and the high cost of producing the mesophase.
8.2.1.4 Polymerisation induced at carbon fibre surfaces. Walterson

(1985) has observed complications during attempts to react epichlorohydrin
(ECH) with C-fibre surfaces, to probe the reactivity to epoxy groups.
Alexander et al. (1991) chose to adsorb ECH from the bulk liquid at 120°C to
more closely emulate the reactions which could occur at the fibre surface in
neat epoxy resin. The chlorine atom provides a highly sensitive element for
surface analysis by X-ray photoelectron spectroscopy (XPS) and for imaging
by time of flight secondary ion mass spectrometry (ToF SIMS). They
showed that the ECH was adsorbed non-stoichiometrically but uniformly
across the fibre surface. The ToF SIMS spectrum produced a series of
secondary ions up to a mass of 1400 D. The fragmentation pattern was
indicative of a polymer based on (CHzCHzOCHz)w This demonstrated that
polymerisation of ECH had been induced. It was also possible to show that
hydroxyl groups were formed. The latter can result from the reaction of
active hydrogen species (e.g. COOH) at the fibre surface. Because chlorine
levels were lower than anticipated, the loss of hydrogen chloride was
postulated with the induction of a polymerisation, which results in
macromolecules growing from the surface. This would appear to be in
contrast to the work of Denison et al. (1988) where the difunctional epoxy
appeared to be bound at both ends.
8.2.2 Aramid fibres

These are based on poly 1,4-phenylene terephthalamide (1) (Kevlar,
Twaron) and poly 1,3-phenylene isophthalamide (2) (Nomex) (Dobb,
1985). The fibres are produced by an analogous technology to pitch-based
foc-@-LH-@-NHt foSQ>-tNH-@NHf
(1) (2)
carbon fibres in that liquid crystal or nematic spinning dopes are employed.
The fibres are wet spun from 99.8% H 2S0 4 at 90°C, into an aqueous
coagulant bath and washed with a weak alkali to remove acidic impurities.
Residual sodium sulphate is considered to be responsible for the observed
variations in fibre-moisture absorption (3 ± 0.5%). The final morphology is
achieved by a subsequent but brief heat treatment for a few seconds in an
inert atmosphere at temperatures between 150 and 550°C when
recrystallization occurs. For high-performance, heat treatments at
temperatures of the order of 450°C lead to a perfection in the lateral order of
the polymer.
The spinning conditions result in a radial crystallite structure similar to
pitch-based carbon fibres. Because of the conformational arrangement of
the polymer chains, a pleated structure is formed.
The lower-modulus, 1,4-aramid with less lateral order is primarily used
for tyre cord and ropes. The higher modulus fibres which have a more
perfect crystal structure are used for composites. The irregular chain
conformation of the 1,3-aramid is not suitable for high stiffness, and has a
lower modulus. These are generally used as staple fibres for honeycomb
applications.
The highly aromatic molecular structure gives rise to high thermal stability
to the polymer. However, ultraviolet (UV) light can sensitise the oxidative
degradation process. Although the degradation products are self-screening,
it is essential, in applications where UV environments are encountered, to
employ protective films.
The polar nature of the molecule means that the equilibrium moisture
content is significant and steps are required to dry the fibres and to prevent
moisture ingress during use.
The aligned crystalline morphology is responsible for the poor fibre
transverse strength and the tendency to fibrillate. Consequently, the fibres
have poor compressive strengths. Composites also have relatively poor
transverse properties because of the limited adhesion chemistry between
fibre and matrix and between fibre crystallite planes. These properties can
be utilised in applications where impact resistance is critical, such as ballistic
protection.
Attempts have been made to improve the fibre-resin adhesion by using
plasma-type surface treatments and increase fibre integrity by y-irradiation
cross-linking, but commercially-treated fibres are not generally available
(Verpoest, 1987).
8.2.3 Glass fibres (Loewenstein, 1983; Jones, 1989)
A number of glass compositions are commonly used for the production of
glass fibres. The most important is E-glass, whereas for specialist
applications others can be used: for example, S-glass where a higher
modulus is required, ECR glass for acid resistance, and AR glass for alkali
resistance (i.e. cement matrices). C-glass is usually produced only in the
form of veil for chemically-resistant GRP linings.
The fibres (other than C-glass veil) are prepared by drawing from the
bushing below a molten glass reservoir. Since the strength of glass fibres is
determined by the distribution and size of surface defects the most
important and commercially sensitive stage in the process is the surface
treatment or size which is applied by rubber roller prior to assembly into
rovings. The rovings consist of strands, each containing 2000 filaments. The
rovings can be either used directly, rewound onto a bobbin (a small twist
being put in to improve handling) for further textile processing, or chopped
into short fibres for the production of chopped strand mat.
The surface 'finish' is a critical stage in the process, ensuring compatibility
with the resin or matrix and high-speed conversion into composite materials
with minimum catenation or fluffing of the strands, fibre breaks, and
damage. The 'finish' consists of an adhesion promoter called a coupling
agent, a protective polymeric coating or size, binder, and other additives
(e. g. fatty acid ester or amide) which ensure good handle ability . It is applied
in an aqueous medium and for epoxy resin matrices this will invariably be an
emulsified epoxy resin. Some manufacturers use polyester/epoxy resin film
formers for compatibility with a range of polyester and epoxy resins. Fibrous
mats or reinforcing tissues will have an additional polymeric binder, which
can be selected to control the wet-out of the fibres in the chosen fabrication
process.
The adhesion promoter which is generally a silane coupling agent, is
expected to generate a strong interfacial bond between the fibre and matrix.
They are generally of the form
R-(CHzkSi(OR'h
where R is a resin reacting or compatible group. As shown in Table 8.3 these

are predominantly amines or epoxides. However, since some rovings may
be of general use in epoxy and polyester resins, mixtures with vinyl,
methacrylate or other type of silane may be employed. Typically n varies
from 0-3. The -Si(OR')3, with R'=CH3 or CzHs , is hydrolysed in the
aqueous medium (often in the presence of ethanoic acid) to -Si(OH)3 for
condensation with surface silanol groups to form a covalent bond. The
statistics of silanol group condensation intra- and intermolecularly leads
to varying degrees of oligomerisation/polymerisation and interfacial
condensation. This leads to a complex interfacial deposit of = 100-200
molecular layers which is reported to involve an oligomeric or physisorbed
component overlying loosely and chemisorbed layers. It is now accepted

that the last two are really a graded polymeric network of variable hydrolytic
stability. At the fibre surface the deposit has the highest network density and
is strongly bound to the glass surface. As a consequence the mechanisms of
adhesion are complex and may involve chemical covalent bonds to the
matrix resin. Alternatively a semi-interpenetrating network may form to
give a 'mechanical' bond at the molecular scale. Furthermore, a number of
reports have highlighted that the presence of the oligomeric component can
be detrimental to the interfacial bond strength. More recently, it has been
reported that the glass substrate ions can be incorporated into the silane
(Wang, 1992). These effects can also contribute to the hydrothermal
stability of the interphase region. This has led to a number of theories on the
structure of the silane interface (Rosen, 1978; Jones, 1989). For example,
Plueddemann (1990) has proposed a reversible hydrolytic bonding mech-
anism which allows stress relaxation through reversible hydrolysis and
condensation to account for the wet performance of the composites. It is
therefore clear that an interphase will form between the fibre and the matrix.
However, this is complicated by the presence of the polymeric size and
binders which have to dissolve into the resin to form an interfacial region or
modified interphase. Thomason and Morsink (1988) make the point that the
coating constitutes up to 1% of the fibre weight of which = 80% is acetone-
extractable, representing = 5% of the resin matrix in the composite.
They showed clearly that this can reduce the glass transition of the matrix
significantly at a rate of 2 K per %. Since this will probably be unevenly
distributed, the thermomechanical properties of the interphase will be
complicated by the glass transitions of the coating and the variably
plasticised matrix.
Table 8.3 Typical silane coupling agents for resin-based composites
Vinyl CH 2=CHSi(OCH 3 h
/0"
*Epoxy CH2CHCH20CH2CH2CH2Si( OCH 3)3
CH 3
I
Methacrylate CH 2= C-COOCH 2CH 2CH 2Si(OCH 3 h
*Primary amine H 2NCH 2CH 2CH 2Si( OCH 3 h
*Diamine H2NCH2CH2NHCH2CH2CH2Si( OCH 3 h
*Mercapto HSCH 2CH 2CH 2Si(OCH3)3
Cationic styryl
Cationic methacrylate
Cycloaliphatic epoxide
• Epoxy resin compatible.

8.3 Fabrication of composites
8.3.1 The reinforcement form

The form of the fibre reinforcement is dictated by the fabrication process,
which in turn is an inherent aspect of the engineering design philosophy.
Discontinuous fibres allow complex mouldings to be manufactured rapidly.
With thermoplastics and thermoset moulding powders, the fibre length is
generally less than 5 mm. As discussed elsewhere a critical fibre length exists
below which the reinforcing efficiency is seriously undermined. Generally
the critical length of fibres in thermosets is larger than for thermoplastics.
Therefore for injection moulding, dough or bulk moulding compounds
(DMC or BMC) containing longer 10-20 mm fibres are preferred where
length retention in the artefact is improved. For longer discontinuous fibres
(> 20 mm) they are bound together with a polymeric binder to form a
random or chopped strand mat (CSM). Fibres of similar length are also
incorporated into a preimpregnated sheet known as sheet moulding
compound (SMC). The fibres can be present in various geometries in SMC.
SMC employs thickened resins but with the wide range of epoxy resins,
prepreg technology is employed.
For epoxy resins, however, the predominant fabrication processes
incorporate continuous carbon, aramid or glass fibres and the techniques
involve rovings directly, as woven cloth or as preforms. The woven cloth can
take on several forms: unidirectional, plain, satin, triaxial, hybrid fibre
fabric or woven tape. The more complex fabrics are generally employed
where drape onto the former is crucial. The convenience of a complex
reinforcement is often off-set by a reduction in average composite
properties. These can be used directly or preimpregnated with uncured resin
or as a 'prepreg'.
The fabrication of fibre composites with epoxy resins is dominated by
those involving continuous fibres: lamination in an autoclave, pultrusion,
filament winding, hot compression moulding. These techniques are
reviewed in Table 8.4.
RRIM and Structural RIM generally employ polyurethane or polyamide
matrices but for the latter, attempts to generate fast curing epoxy resins
suitable for this application are under development (see section 8.3.3.2).
8.3.2 Prepreg mouldings

High-performance carbon-fibre and glass-fibre laminates are fabricated by
prepreg moulding using autoclave or pressclave techniques. A 'prepreg'
consists of collimated fibres or woven cloth pre impregnated with an epoxy
resin (or other thermoset or high performance thermoplastic) modified with
a thermoplastic or elastomeric toughening agent and containing a latent
Table 8.4 Fabrication processes for composites from epoxy resins
Processes Reinforcement Vol. Method Typical application

(%)
1. Spray-up Chopped rovings (G) 15-30 Direct application of chopped Pipe linings
rovings and resin
2. Autoclave Continuous (G , C, A) 60 Forming of stacked prepreg in Aerospace components
vacuum bag
3. Filament winding Continuous rovings, -50 Wet resin carried by fibres onto Pipes and containers, geodesic
occasionally particulate mandrel at predetermined angles. shapes
fillers included (e.g. sand) Variants: prepeg tape, woven tape,
(G,C) dry preform manufacture
4. Pultrusion Continuous fibres, occasional 60--80 Wet resin-impregnated roving or I -beam or rods
CSM and fillers (G, C) cloth formed through a heated die
5. Compression moulding Chopped or continuous fibres 10--60 Pre peg or wet resin/preform pressed Medium-sized complex mouldings
or preforms and fillers to shape in a closed mould between
(G, A, C) matched dies
6. Resin injection, resin Continuous or discontinuous 25-30 Wet resin injected into woven, knitted Panels
(RTM) transfer fibre plus preform polymer or CSM preform within a closed
foam inserts (G, C, A) mould
7. Reinforced reaction Short-milled glass fibre or 5 Fibre-resin dispersion and hardener Large complex parts (e.g. auto
injection moulding fibre preform (G) 20 mixed at rapid mixing head and parts for impact resistance but
(RRIM) injected into mould. with low stiffness)
Chemistry controlled to give rapid
reaction
8. Structural RIM Continuous fibre preform 30--50 Similar to 6 with the more rapid RIM Large structural parts.
(G,A,C) chemistry Rapid production possible
Fibres: G = glass; C = carbon; A = aramid.

hardener. The latent hardener such as the insoluble dicyandiamide (DICY)
is micronised in the resin solution prior to wet-impregnation into the
collimated fibres and woven cloth and drying. The curing temperature and
the kinetics of cross-linking can also be adjusted by incorporating a
promoter or accelerator (e.g. Monuron or Diuron). The resin viscosity is
adjusted to give the prepreg the appropriate tack by adding thermoplastic
polymers or 'advancing' the resin with partial reaction with the hardener.
The prepreg is stored between plastic release films in a sealed container in
a freezer. After thawing and temperature equilibration to prevent moisture
condensation and ingress, the unidirectional sheet is cut to appropriate
dimensions and stacked at predetermined angles to produce a balanced
laminate (e.g. O2°/ ± 45°/902°/ ± 45%2°; the subscripts define the number
of prep reg lamina in each ply; the choice of stacking sequence is determined
by the required laminate properties). The prepreg stack is laid up between
release cloths and 'blotting paper or similar' bleed pack to soak up the
excess resin in the prepreg (Figure 8.5). The bleeding of excess resin
facilitates the removal of entrapped air and residual volatile impurities
during consolidation of the prep reg into a void-free laminate. The curing
schedule is also carefully controlled since the exothermic nature of the
chemical reactions can lead to a temperature build-up in thick sections.
Some of these processing difficulties can be overcome by using
thermoplastic matrices but these have their own processing problems.
Prepreg moulding techniques have a complicated technology, but offer the
mechanical engineer the opportunity to orientate fibres in the direction of
the principal stresses within complex structures.
8.3.3 Matrices for fibre composites

Plastic matrices for fibre composites generally can be divided into
thermoplastics, chemosets, and thermosetting polymers. The last two
broadly fall into the same category of thermosetting resins, but for the
purposes of this discussion we differentiate between those requiring
Bag pressure line
Bag Porous membrane Rele layer

Thermocouple
Prep reg laminae
Dam
Figure 8.5 Autoclave moulding arrangement for laminates based on epoxy prepreg. (After
Purslow and Childs, 1986).
additional hardeners, curing agents, or catalysts (the chemosets) and those

that are quenched at a partially reacted state and continue to cure on
heating. In both cases, curing involves the generation of a three-dimensional
cross-linked molecular structure. Although it is generally assumed that the
cross-linking process is homogeneous, there are notable examples where an
abundance of indirect evidence points to a heterogeneous amorphous
molecular structure. For the epoxy resins, it is more likely to result from the
heterogeneous nature of some of the curing agents. The main types are
epoxy or unsaturated resins such as polyester or vinyl esters. The latter are
low polymers with Mn < 2000 g . mol- 1 whereas the epoxides are generally
of relatively low molecular weight. These two systems satisfy the
requirements of an appropriate matrix.
1. Low viscosity for fibre impregnation
2. High reactivity on curing
3. Chemical control of cure without volatile formation
4. Good mechanical properties of the cured resin.
Criterion 3 is not readily achieved by the thermosets (e.g. phenolics,
polyimides) since these involve condensation rather than step-growth or
addition polymerisation. To prevent damage accumulation as a result of
volatile formation, higher pressures and more complicated cure schedules
are required. The bismaleimides (e.g. PMR-1S) contain thermally
polymerisable unsaturated groups and provide significantly higher
temperature performance (up to 400°C) than the advanced epoxy matrices.
Epoxies generally satisfy the above criteria because the curing occurs
without evolution of volatiles, in a controllable manner with low shrinkage.
8.3.3.1 Epoxy matrix resins. The difunctional diglycidyl ether of

Bisphenol A (e.g. Ciba-Geigy MY7S0) and related oligomers of degrees of
polymerisation up to four, form the basis of the general-purpose resins for
both adhesives and composites. The glycidyl amines have a higher
functionality and potentially higher cross-link densities and glass transition
temperatures and form the basis of the resins for aerospace composites. The
latter are used as prepreg resins and are often mixtures based on the
tetrafunctional MynO. Typical resins for laminates are given in Table 8.5.
Judicial choice of resin blend and hardener is required to provide controlled
cure and viscosity profiles during cure. Furthermore, thermoplastic
modifiers such as polyethersulphone and phenoxy resins are also used to
control process variables.
The Ciba-Geigy 913/914 systems are based on this technology, as are
some Fiberite systems. Both the molecular weight of the thermoplastic and
the end-groups contribute to processing and performance. For example
amine end-group terminated polyether sulphones ensure chemical bonding
with the epoxy resin. As will be discussed later, moisture absorption leads to
Table 8.S Typical epoxy resins used for composite fabrication
Difunctional:
0 ~:©-O-CH2-tH-CH2tO@+:&-O-CH2-CH~CH2
= 0), MY 750
f ©-
Diglycidyl ether of bisphenol A (n
CH2~CH-CH2
o 0
CH
CH 3
3 n OH
n
c
CH3
0
Trifunctional:
MY 0510
a significant reduction in glass transition temperature of the cured resin

which seriously limits the maximum service temperature. To mitigate this
effect highly cross-linked resins are preferred. However, to ensure that the
microcrack resistance of the composite is maintained, toughening modifiers
are required. The thermoplastic modifiers can also act in this way. However,
in contrast to adhesives, carboxyl-terminated rubber (CTBN)-toughened
epoxies are not generally employed for composites. This probably demon-
strates that the advantage of modification with thermoplastics may lie
predominantly with processing improvements.
Other important systems employ thermosetting modifiers such as
bismaleimide or cyanurate resins to achieve high temperature performance
with durability.
It is important to emphasise the difference between hardeners and
catalytic curing agents. The latter are used in small quantities and primarily
yield a cured resin with polyether network chains. The former are molecules
with active 'hydrogen' atoms which can add to the epoxy group. There are
two main types: the acid anhydrides and the amines and polyamides. They
have a powerful influence on the cured resin properties as a result of
the molecular structure of the network chains. More details are given in
chapter 2.
(3)
Amine hardeners react with epoxies to form hydroxy amines (3). At first
sight, this might be expected to promote adhesion to the fibres through the
hydroxyl groups, but the polarity promotes moisture sensitivity and
hardening or catalytic systems which provide a less polar molecular structure
are preferred. For industrial lamination, tertiary amine-promoted
anhydride cure is often employed. Nadic methylenetetrahydrophthalic
anhydride (NMA) is commonly employed with the diglycidyl ether
(MY750) and tetrafunctional glycidyl amine (MY720) with benzyl dimethyl
aniline (BDMA) or the similar Epikure K61b. This system provides a matrix
with lower water absorption, which can be attributed to the 'polyester'
copolymer molecular structure (4).
A polyether structure originates from the use of catalytic systems of which
the cationic boron trifluoride amine complex is favoured (5). For advanced
composites, BF3 complexes, often in combination with diamino diphenyl-
sulphone (DDS), act as latent hardeners becoming active at :::: 160°C.
Therefore, the resin viscosity profile with temperature can be adjusted by a
precure (sometimes referred to erroneously as B-staging) to optimise the
o o 0
x>
II + II +
C -NR3 0 C -NR 3
+ NR3 ~
II II II I
o o 0 0-
X X>
° 0
L"R'
/0,
8
>-
C CH 2-CH-CH 2"'"
II I
° ° °
"C/ /rF
C~
~o
(4)
consolidation of the prepreg stack and removal of entrapped air in the

fabrication of a laminate by vacuum bag/autoclave techniques, prior to the
onset of cure.
Micronised dicyandiamide (DICY) also provides latent curing ability to
prepreg with improved shelf-life. Both high and low temperature curing
systems are available. Without the presence of 'Diuron' or 'Monuron'
amine accelerator, the rate of curing is determined by the melting of DICY
at 165°C at which temperature it dissolves into the epoxy resin. With
'Monuron' the curing temperature can be reduced to 120°C. While low
OH
~ + I
0- CH 2 -CH """'"
I~
(5)
H 2N-C-NHCN
II
NH
H,N-C-NH-CN
+ . II
NH
OH
I
HO-CH
I
JVVVV\I\I\I\/'CH-CH, CH?
.'-....,. I .
N-C-N-CN
/ II
JVVVV\I\I\I\/'CH-CH, NH
I . +
OH
Scheme 8.1 Cure of epoxy resins with DICY. Alternative reaction schemes are discussed by
Amdouni et al. (1990).
temperature systems have clear advantages, not least the control of the
exotherm during cure, careful control of the curing schedule is required for
the production of good quality void-free laminates. Thus a dwell time at
'minimum' viscosity has to be incorporated, usually = 90°C for these
systems. The other disadvantage of the low temperature curing systems
based on DICY is the increased probability of residual DICY hardener in
the resin. Another observation is the inconsistent appearance of a micro-
structure in these cured systems. Much of the variability can be reduced by
ensuring that the hardener is micronised to a uniform fine particle size prior
to prepreg manufacture. However, microstructural variations can still
occasionally appear but these could also be attributed to phase separation of
the thermoplastic modifier.
DICY is used in catalytic quantities 4--6 phr but the curing chemistry is
complex producing a structure expected of a polyamine as indicated in
Scheme 8.1.
It should be pointed out that the complete reaction of the epoxy groups is
not normally achieved except after long times at high temperatures so that
higher-he at-distortion temperatures and chemical resistance can only be
achieved by appropriate postcuring. The mechanical properties of epoxy
resins are therefore determined by the hardener or curing systems and time-
temperature profile. The time temperature transformation curing diagrams
of Gillham (1986) are particularly valuable in this context. This is discussed
in chapter 1. The effect of post-curing on mechanical properties can be
generalised by appreciating that the heat distortion temperature and
Young's modulus increase with cross-link density and the concentration of
aromatic groups within the network. However, there are exceptions to this
where a higher glass transition can arise alongside a lower modulus, since the
glassy state is frozen in at higher molecular free volume.
8.3.3.2 Rapid curing resins for structural RIM. Much progress has been
made recently in the development of rapid curing epoxy systems for
alternative composites processing.
Stanford (Mortimer et al., 1992) has reported gel times of < 25 s at 40°C
using BF3 diol complexes (6) of Dianol 240 and 320 to cure DGEBA (Dow
Chemical DER332).
HO[CH'-iH-ol!Q\-VO\Lo-yH-cu,lOH
L
BF"
l R L~ CHP l R
(6)
where with R = H (6) the average value of x and y = 2.15. With

R = CH3 , X = Y = 1.
To provide control over resin viscosity for RIM, processing co-monomers
for the cationic polymerisation of the epoxy groups are included. Tetra-
hydrofuran yielded a resin of low Tg but with butyrolactone (7) a higher Tg of

103°C could be achieved. Gel times of 64 and 80 s were observed.
(7)
The shortest gel time of 35 s was observed with the BF3 complex with diol
(8). The maximum cure temperature was 70°C.
HO-CH2-CH2 -O-@-O-CH2 -CH2 -OH
(8)
This appears to be a significant advance over the literature citations in which

demould times were> 2 mins at the higher temperature of > 100°C.
8.4 Mechanical properties of unidirectional laminates
A unidirectional fibre composite is highly anisotropic so that in real

laminates the fibres are arranged at angles to one another. Since transverse
to the fibres, matrix and interfacial properties dominate and the statistical
variability in fibre strength dominates in the fibre direction, the
characteristic failure mode is one of damage accumulation, through
debonding, microcracking and fibre fracture. However, under stress these
events are comparatively innocuous under most service conditions but can
act as nuclei for more severe failure processes such as stress corrosion
cracking of GRP in acidic environments.
8.4.1 Longitudinal modulus, E1

When a tensile load is applied in the longitudinal direction to the perfectly
bonded continuous fibres the strain in the matrix, Em will be equal to the
strain in the fibre Ef. Since Ef > Em the stress carried by the fibres, af, is
greater than that in the matrix, am' Ef and Em are the Young's moduli ofthe
fibre and matrix respectively. The stress in the fibres and matrix respectively
is given by
(8.1)
(8.2)
where fc is the strain on the composite. The stress on the composite Gc is
given by equation 8.3.
(8.3)
By equating the volume fraction of fibres V f with the fractional fibre area and
the volume fraction of matrix Vm with the fractional matrix area, the
Young's modulus of the composite in the fibre direction, E" is given by,
(8.4)
and since Vm = (1 - Vf)
(8.5)
This is known as the law of mixtures and assumes that both the fibres and
matrix are perfectly elastic. At high V f any deviations due to viscoelastic
behaviour of the matrix are relatively insignificant. Similarly the additional
stresses due to differences in the Poisson's ratios of the fibres and matrix
only lead to an error of less than 2%. Experimental verification of equation
8.5 has been established in several studies.
8.4.2 Longitudinal tensile strength, Glu
Under a rising tensile stress parallel with the fibres, either the fibres or the
matrix will fail depending upon the relative failure strains of the fibres (ffu)
and matrix (fmu)' Two cases need to be considered (Figure 8.6). Case 1 when
fibre °fu -----------

fibre
Of _._._._"-
'"'"
Q)
bi
Strain Strain
(a) (b)
Figure 8.6 Individual fibre and resin stress/strain curves defining the failure processes
described in Figures 8.7 and 8.8. (a) Case 1, f mu > flu; (b) Case 2, f mu < flu'
the failure strain of the matrix is greater than that of the fibres (a typical
fibre-reinforced epoxy resin) and Case 2 when matrix failure strain is lower
(a typical fibre-reinforced ceramic).
Case 1
With uniformly strong fibres at high volume fractions, when the fibres fail all
the load is thrown upon the weaker matrix. The effective cross-sectional
area is reduced and the composite will immediately fail. The strength of the
composite alu is given by
(8.6)
where a~ is the stress in the matrix when the fibre breaks as shown in Figure
8.6.
At low volume fractions (typically less than 0.006--0.03 for GRP and
CFRP) the matrix can carry the load, leading to multiple fracture of the
fibres. The strength is given by
(8.7)
Equations 8.6 and 8.7 are plotted in Figure 8.7 where it is seen that a critical
fibre volume fraction (Verit ) for reinforcement exists. The matrix contributes
insignificantly to the failure strength of the unidirectional laminate in the
fibre direction. The strength is given approximately by equation 8.8.
Single Fracture
~
~
~
u
~
'"'
'"
~
~
.n
~
!
a.
.~
"
~
~
>:
a
mu
o 1.0
Vcri tical
Vf
Figure 8.7 Fibre volume fraction dependence of fracture strength for a unidirectional
composite, Olu, manufactured from a resin of failure strain higher than that of the fibres
(emu> efu)'
(8.8)
However, as shown in Figure 8.2 the fibres have a statistical failure
strength and strain, and in practice a cumulative weakening process occurs
with fibre breaks at random. The first fibre break may lead to immediate
failure if the flaw has sufficient stress intensity to cause the crack to
propagate. This situation only arises when the fibres are very strongly bound
to a matrix of low fracture toughness. Generally fibre-resin deb on ding
occurs to a limited degree at the fibre ends so that stress transfer through
shear enables the fibre damage to accumulate until the multiple fractured
fibres reach their ineffective lengths. The probability of fracture of adjacent
fibres is increased. This leads to a sequential failure and the production of a
Griffith flaw which can result in composite failure. The generation of a
critical number of adjacent fibre breaks can be inhibited by dispersing a
higher failure strain fibre into the bundle to increase the ultimate strength in
what has been termed the hybrid effect.
The tensile fracture surface has a brush-like appearance with varying
degrees of fibre fracture and pull-out making meaningful strength
predictions difficult. For further discussion of this problem the reader is
referred to the texts by Hull (1981) and Kelly (1973).
ease 2
Under these circumstances the matrix fractures first, throwing all additional
load onto the fibres (Figure 8.6). At low Vf as illustrated in Figure 8.8, the
Single Multiple Matrix

Fracture Fracture
v' 1.0
f
Vf
Figure 8.8 Fibre volume fraction dependence of fracture strength of a unidirectional composite
manufactured from a matrix of failure strain lower than that of the fibres (cmu < Cfu)'
fibres are unable to support the additional load and the composite fractures.
The strength is given by
(8.9)
where of is the stress on the fibres when the matrix fractures (see Figure 8.6).
At higher V f when V f > V f (= Omu/Ofu - Of + omu) the fibres are able to
support the extra load when the matrix fails. Additional loads cause multiple
cracking of the matrix. Further details are given by Aveston and Kelly
(1973). The strength of the laminate is clearly determined by the average
strength of the fibres O"fu and
(8.10)
In practice the distribution of values of 0fu means that some fibres may fail
before the onset of matrix failure and continual damage accumulation of
matrix cracks and fibre fracture occurs.
8.4.3 Transverse modulus, E t

A simple model for predicting the transverse Young's modulus of a uni-
directional composite assumes a continuity of stress across the fibre-resin
interface such that
(8.11)
thus
(8.12)
or
(8.13)
The rule of mixtures equation 8.13 is a poor estimate of the actual transverse
modulus and a number of other attempts have been made. The most
rigorous approach is that of Hashin and Rosen (1964) in which upper and
lower bounds may be calculated. The Halpin and Tsai (1969) equations are
generally applicable and allow for variations in packing geometry and
regularity with the factor;. These rules are compared in Figure 8.9 where it
is clearly seen that transverse fibres are inefficient reinforcing elements.
(8.14)
where
q- 20
E
z
~
ur
vi 15
:J
:;
"0
0
E
~
.0; 10
c
$
'i!?"
'c>"
<f)
5
~
f-
0 0.8
VI
Figure 8.9 The effect of fibre volume fraction, V" on the transverse tensile modulus of a
unidirectional composite. To accommodate the range of V, both epoxy resin (V, = 0.25-D.6)
and an unsaturated polyester (V, = 0.08-0.3) resins have been used. (e), Glass-polyester;
(0), glass-epoxy; (-), Law of mixtures (equation 8.12); (---), Halpin-Tsai equation (equation
8.14) which gives the best fit to the data. (Redrawn from Bailey and Parvizi, 1981.)
8.4.4 Transverse strength, 0tu
The transverse strength of a unidirectional composite will be determined by

(a) the transverse strength of the fibres, (b) the interfacial bond, or (c) the
matrix, whichever is weakest. This simple model fails to account for stress
concentrations in the matrix.
The composite will therefore fail at a stress 0tu which is significantly less
than 0mu such that
(8.15)
The failure strain of the matrix in a composite, f mu , is significantly reduced
by strain magnification as a result of the stiff fibre. Thus for a stress
continuity, the strain in the matrix resin will be proportionally higher than
the strain in the composite by the ratio of the moduli Em/E f •
The strain magnification theory of Kies (1962) modified by Garrett and
Bailey (1977) gives ftu = f mu {1 + 2[(.n/Vf)1/2 - 2rl}, where
(8.16)
However, as shown in Figure 8.10 this is still only an approximate solution
since it does not take into account additional stress intensities at debonded
fibres or other defects.
c:
.~
1il \,
~ \
::> '\
~ 2 '\
...
'\
CD
li? "-
"-
CD
>
!Il
"- ,
C "
'"
~ 1 "
......... ' .................
~-
0. 0..2 0.4 0..6 0..8

Fibre volume fracture
Figure 8.10 The effect of fibre volume fraction, VI, on the transverse failure strain of a
unidirectional E-glass-epoxy composite. (-), predictions of the Kies theory.
The discrepancy between theory and experimental therefore reflects the

presence of strength-reducing flaws. In this case, they can occur at specimen
surfaces but also at partially debonded fibres as a result of poor wet-out or
other mechanism. The former aspect can be reduced by measuring the first
transverse cracking strain of a crossply laminate but compensation for
residual stresses has to be carried out.
Table 8.6 gives typical properties of 90° and 0° composites from carbon,
glass and aramid fibre epoxy where the poor composites transverse
properties are confirmed. The transverse strength is a complex function of
the fibre, matrix and interfacial strength. In the case of aramid fibres, the
aligned crystalline structure and poor inter-crystallite bonding also
contribute to the low transverse strength. Fibre surface treatments are
common for glass and carbon fibres but as discussed previously they have
proved less effective for aramid fibres. However, these fibres impart high
impact strength to their composites because of the energy absorption
Table 8.6 Mechanical properties of unidirectional epoxy resin composites
Fibre VI EI E, alu a,u flu f,u VI' V,

(GPa) (GPa) (GPa) (GPa) (%) (%)
Carbon 0.62 127 8.3 1.7 0.039 1.16 0.48 0.29 0.019
Glass 0.55 42 14 0.92 0.056 2.2 0.5 0.27 0.09
Aramid 0.62 83 5.6 1.31 0.039 0.35
, V is the Poisson ratio in the longitudinal (I) and transverse (t) directions.
associated with transverse fibre failure and fibre-matrix debonding. Design
should employ a careful selection of fibres to provide the required
performance.
Fibre surface treatments and coatings are currently receiving more
attention as a means of improving and optimising interlaminar shear and
transverse properties.
8.4.5 Off-axis properties

The modulus of a unidirectional fibre composite with fibres at an angle (J to
the applied stress is given by:
lIEo = cos4 (JIE} + (l/G} - 2vJE}) cos2 (J sin2 (J + sin4 (JIEt (8.17)
where (J is the angle of the fibres to the applied stress and Vc is the Poisson
ratio for the specimen. However in an off-axis test failure can occur by
transverse, longitudinal or intralaminar shear. In the maximum stress theory
(Tsai and Hahn, 1980; Tsai, 1985), the stresses reach a critical value for
fracture in the principal material directions. Thus the criterion for failure is
a, = a,u
at = atu (8.18)
T = Tu
600
\
\
\
500 \
E 400
z 0
~
1u
iii 300 sin e cos e

'"~
ii)
:;'"
U 200
u:'" ~
\
\
100 /"0-..
Tsai·Hill '"
/0, = s~n~ e
criterion ~
o
Off-axis angle, e (degrees)
Figure 8.11 Effect of angle of test (8) on the strength of a unidirectional carbon fibre (high
strength) epoxy composite (Vf =0.5) showing the predictions of the Tsai-Hill model.
(Redrawn from the original given by Hull, 1981.)
For angles of (J to fibres between 0° and 90° the applied stress (at angle (J), G()
will have the following components and the failure will be determined by the
weakest:
G() = Glulcos2 (J
G() = Gtu /sin2 (J (8.19)
G() = iu/sin (J • cos (J
This is illustrated in Figure 8.11, where these equations are compared for
epoxy-based CFRP. At angles of 0-5° the failure is dominated by
longitudinal fracture, from 5-18° by intralaminar shear failure and from
45-90° by transverse fracture. In the region from 18-45° a mixed mode
failure occurs. The actual angles at which these transitions take place will
clearly be dependent upon the values of Gtu , Glu, iu. The Tsai and Hill
criterion which is based on maximum work theory however gives a better
prediction of the off-axis strengths in this region (Halpin and Tsai, 1969; Tsai
and Hahn, 1980; Hull, 1981; Tsai, 1985).
8.5 Failure processes in laminates
Fibre composites are constructed with fibres at angles to one another to

maximise resistance to multiaxial stresses. For advanced composites,
stacking of unidirectional plies (e.g. prepreg) at appropriate angles is
commonly employed. Balanced constructions are required to prevent
warpage. The nomenclature describes the number of plies of each type: for
example a simple 8-ply crossply laminate, 02°/904%2° or more complex angle
ply laminate, O2°/902°/ ± 45%2°/ ± 45°/902%2°. From theforegoing it will be
appreciated that the failure processes are highly complex. In particular the
free edge stresses in angle ply laminates can significantly modify the
sequence of failure of individual laminate. The detailed discussion of these
failure mechanisms by Hull (1981), Tsai (1985) and Tsai and Hahn (1980)
are recommended. However, the failure of model crossply laminates
illustrates the micromechanics of these types of composites.
B.5.1 Crossply laminates

A model 0°/90%° crossply laminate is illustrated in Figure 8.12. A typical
stress-strain curve for an epoxy-glass laminate is given in Figure 8.13 where
the first knee in the response of this laminate coincides with the observation
of stress whitening (Figure 8.14). This coincides with fibre-matrix
debonding and a cure-dependent viscoelastic response in the epoxy resin
matrix. But these de bonds initiate transverse cracking at an applied strain of
0.35%. Because the transverse cracks are pinned by the 0° plies (i.e.
delamination at the ply interface does not occur) the stress builds up again in
t I direction
t direction
If
'" '"
Splitting in
longitudinal (0°)
lamina
Transverse
crack in
transverse (90°)
lamina
Figure 8.12 A 0°/90%° crossply laminate showing transverse cracks which progressively form
under a rising tensile stress. Longitudinal splits in the 0° laminate also occur at high composite
strain.
125
100 Ultimate Failure

at 2.4%
~ 75
~
VI
VI
~
iJj 50
Stress Whitening
25
Strain (%)
Figure 8.13 Low strain part of a stress/strain curve for a 0°/90%° laminate showing the onset of
cumulative damage beginning with fibre-matrix debonding (stress whitening) and transverse
cracking (see Figures 8.12 and 8.14) (after Manders et aI., 1983).
(nOO (k)03L (1)036 (m) 0 51. (n)On (0)090 (p)1.08 (q) ISO (,)190
.20mm.
Figure 8.14 Progressive multiple cracking in the transverse ply of a glass·epoxy crossply
laminate. Note the presence of longitudinal splits at higher strain. The numbers represent the
strain at which the photograph was taken (after Manders et al., 1983).
the 90° ply at a transfer distance of = 2 mm from the crack. This results in a
black band either side of the transverse cracks as shown in Figure 8.14.
Multiple cracking of the inner ply continues as the load on the specimen is
increased until saturation has occurred when the shear-lag stress transfer can
no longer exceed the transverse strength of the remaining elements.
Under continued loading Poisson strains are induced which cause the
longitudinal ply splitting shown schematically in Figure 8.12. The onset of
ply splitting is also seen in the photographic record of the cracking sequence
in Figure 8.14.
Parvizi et al. (1978) used a modified shear lag analysis to calculate the
additional stress transferred back into the transverse ply at a distance from
the crack, y. These authors assumed that the cracks would form midway
between the previous ones. If iloo is the additional stress required to load the
transverse ply to its cracking strain then
iloo = ctluEtd/b[1 + exp (- q,1I2 t) - 2 exp (- q,1I2t)/2)rl (8.20)
where q, = EcGt(b + d)/EIEtbd2 and d is the semi-90° ply thickness, b is
the 0° ply thickness, tis the crack spacing and Oa is related to iloo according to
(8.21)
The nature of the accumulation of transverse cracks dictates that the crack
spacings predicted by these equations are a stepped curve. However, it has
been suggested that since the transfer distance y is small by comparison with
the crack spacing that the first cracks form statistically at the weakest points
in the 90° ply (see for example Manders et al. (1983) and Peters (1984».
Kimber and Keer (1982) showed that the average crack spacing is related to
the lower bound of the shear lag analysis.
The statistical aspects of crossply cracking can arise from variations in
fibre dispersion within the laminae, and could be worse when a memory of
the individual tows remains. Peters and Chou (1987) have analysed the
process in terms of an increasing element strength of a weak link theory and
have argued that the statistics can be extrapolated to obtain the true
transverse strength of the 'defect-free' ply.
8.5.2 Constraint cracking

Practical laminates make use of the fact that the transverse cracking strain
increases with decreasing 90° ply thickness as a result of the suppression of
transverse crack propagation by energy constraint. Parvizi and Bailey (1978)
used equation 8.22 to predict the minimum transverse ply failure strain
.f~i~ = [2bEIYlp1l2/(b + d)EIEcJ (8.22)
where Yt is the fracture surface energy of the transverse ply. This equation
provides a good correlation with experimental values at ply thickness less
than 0.5 mm. Complete suppression occurred at 2d < 0.1 mm. At large ply
thicknesses factors which have been discussed above control the transverse
ply failure strain. Therefore laminates should ideally be constructed from
finely dispersed plies, to control damage accumulation. However, it should
be recognised that debonding may still occur and that environmental factors
might dictate a different laminate construction. Furthermore the growth of
transverse cracks is a prime mechanism of stiffness reduction during fatigue
and energy constraint may not be a good mechanism of crack suppression
under these circumstances. Therefore prepreg technology utilising fine, well
dispersed tows provides the most appropriate fabrication route for advanced
composites. In many ways this effect has discouraged the development of
cheaper, larger tow carbon fibres. The developments of interphases and
higher failure strain matrices are required to encourage these developments.
8.5.3 Epoxy resin matrix failure strain

A study of equation 8.15 shows that the transverse cracking strain is directly
proportional to the matrix failure strain. The increased toughness generally
associated with the matrix failure strain can also restrict propagation and the
transverse cracks take on a diffuse character.
A matrix of increased failure strain could have a larger expansion
coefficient but since this could also result in an increased residual thermal
strain a lower transverse cracking strain may still result.
8.5.4 Thermal strains in crossply composites

The expansion coefficients of a unidirectional lamina in the transverse and
longitudinal directions at and al which are dominated by those of the matrix
and fibres am, af respectively, can lead to the induction of a significant
thermal strain in the laminate. A large difference exists between am (60-

100 X 10- 6 K- l ) and af (glass = 5, HMS carbon (axial) -0.54, Kevlar 49
(axial) -2 x 10- 6 K- l ). The equations of Schapery (1968) have been shown
to predict the expansion coefficients of the composite reasonably accurately.
al(T) = (EmamVm + EfafVf)/(EmVm + EfVf) (8.23)
at(T) = am Vm(1 + vm) + afVf(l + Vf) - alvJ (8.24)

For the crossply laminate, the higher longitudinal stiffness constrains the
large transverse contraction on cooling from the post-curing temperature
putting the plies into tension in the transverse direction and into
compression in the longitudinal direction. The thermal strain in the
longitudinal direction of the transverse ply, c~7, is given by
th El b(at - al)(Tl - T2)
C - (8.25)
(Elb + Etd)
~~~--~~--~
tl -
where Tl is the stress-free temperature. (at - al) (Tl - T2) can be obtained
directly from the curvature of an unbalanced 0°/90° coupon in a manner
analagous to a bimetal strip.
= (b + d)/2Q + E 1 b + E td . [lIEl b + lIEt d)

3 3
(at - al){Tl - T2) (8.26)
6Q(b + d)
The radius of curvature Q is related to the displacement of the coupon from
flat 0 by
(8.27)
where L is the length of the beam and Tl can be experimentally determined

by recording the temperature at which the beam becomes flat (i.e. when
0=0).
The transverse ply cracking strain, Ctlu, is consequently reduced by the
magnitude of ctt~ and transverse ply failure strain Ct~ is given by equation
8.28.
(8.28)
Typical values of c~~ can range from 0.2% to > 1.0% depending on the stress
free temperature and the difference in the longitudinal and transverse
expansion coefficients (~a). It is generally larger for carbon fibres in
equivalent resins since af along fibre is slightly negative. Since Ctu can be
typically 0.6--0.8%, at worst the propensity to crack during cooling can be
high for crossply laminates manufactured from resins with high Tg and cured
at temperatures in excess of Tg while at best the first cracking strain, Ctlu, can
be reduced to a low level.
8.5.5 Poisson-generated stresses and longitudinal splitting
In GFRP, the Poisson-induced strains can cause longitudinal splitting of the
0° plies. The effect of thermal strains is similar to that described above.
However, longitudinal splitting has not been observed in CFRP prior to
failure even though the Poisson strains are greater because of the lower
overall failure strain of the composite.
8.5.6 Angle ply laminates

For laminates in which the fibres are arranged at angles to one another the
micromechanics are complicated by the presence of interlaminar shear
stresses at the free edges. Delamination of off-axis plies can be initiated at
the free-edge or at damage in an infinitely wide specimen. From a
consideration of the stresses occurring in symmetrical angle ply laminates at
± () to the uniaxial tensile stress it is clear that the shear stresses will be
maximised at approximately ± 45°. At larger angles, transverse cracking
will dominate the failure process although the actual value will largely
depend upon the relative magnitude of the transverse, longitudinal and
shear strengths. At smaller angles, shear dominated failure occurs.
The interactions between the stresses in the individual plies and the
viscoelastic nature of the matrix resin makes the prediction of the stress-
strain curves complex. However, this type of laminate is often subject to
biaxial stresses and under these conditions, transverse cracking dominates
the failure of each laminae at all angles.
In practice, complex laminates with balanced symmetrical configuration,
e.g. [0°/ ± 45°/902°]5 are used to provide load-bearing ability. The stacking
sequence is selected to minimise the free edge stresses but the damage
mechanisms will follow the general phenomena described in the previous
sections, with fibre debonding developing into transverse cracks in the 90°
plies, which in turn can initiate damage in the ± 45° plies. Debonding in
± 45° plies, where there is no adjacent 90° ply can also occur at higher strains
leading to an extensive pattern of cracks. On further loading, delamination
at ply interfaces, in-plane longitudinal cracking of the 90° plies and
longitudinal splitting of 0° plies occurs. To optimise the performance, the 90°
plies are normally finely dispersed and those with a large Poisson's ratio (i.e.
± 45° plies) should be located in the laminate mid-plane. A further variant
on ply geometry is the use of woven fibrous prepreg whose drape properties
enable complex curved surfaces to be fabricated more easily, but with the
introduction of a more complex micromechanics involving fibre cross-over
and resin-rich areas. In service, microcracking is likely to occur so that
design practice needs to take account of the presence of benign cracks in
composite structures.
8.5.7 Discontinuous fibre composites

A variety of discontinuous fibre composites is available ranging from the
relatively long aligned fibre sheet moulding compound SMC through to the
in-plane random dough moulding compound DMC and the short, random,
reinforced fibre reinforced thermoplastics FRTP and thermosets FRTS.
The performance of these materials differs from continuous fibre
composites in several respects:
1. Lower volume fraction of reinforcing fibres (Vf )
2. The reinforcing efficiency (i.e. the aspect ratio of the fibres)
3. The orientation distribution of the fibres
4. The effects of flow-induced orientation leading to a heterogeneous
fibre distribution.
The major advantage of short fibre materials comes mainly from
fabrication economies and they do not playa major role in epoxy resin-
based composites. However, the modulus of discontinuous in-plane random
fibre composites can be given by a modified law of mixtures equation which
includes orientation ('YJo) and length ('YJl) factors:
Ec = 'YJO'YJI Ef V f + Em(1 - Vt) (8.29)
or for 'YJl = 1, by integrating over range of orientation angles
Ec = 2n
5 0JCI2 Eod() (8.30)
where Eo is defined by equation 8.17.

In practice the following simplified expression is more useful
Ec = 3El/8 + 5EI/8 (8.31)
The length efficiency factor is complicated by the nature of the transfer of
load from the matrix by shear at the interface of the fibre. The strength of
short fibre composites is dominated by these mechanisms which have been
discussed in detail elsewhere (Cox, 1952; Kelly and Tyson, 1965; Folkes,
1982; Bader, 1985). The length/diameter ratio of the fibre has a critical value
but for 99% efficiency, the fibres need to be 10 times this value. The
minimum efficient reinforcing length of glass fibre in an epoxy resin is
therefore in excess of 10 mm and nearer 100 mm for the higher modulus
resins. As a consequence, short fibre reinforced epoxy resins do not find
extensive applications.
8.6 Effect of moisture on the performance of epoxy resins
Most epoxy resin composites are employed in applications where humid

environments are encountered. Since these are often in aerospace, the effect
of moisture absorption and desorption on the mechanical properties of the

laminates is crucial to the design procedures. The selection of a resin system
is often determined by these factors.
8.6.1 Moisture absorption kinetics

In comparison to heat transfer, mass transfer of the water molecule is slower
by a factor of = 106 . Thus moisture equilibrium of a 12.5 mm thick carbon
fibre epoxy laminate at 350 K can take = 13 years whereas thermal
equilibrium only requires 15 s. Further, thermal equilibration can be ignored
when defining the moisture problem. Moisture absorption is generally also
considered to be independent of concentration. Therefore the diffusion of
moisture into cured epoxy resin is generally considered to obey Fick's law:
2
ac = D ~ (8.32)
at x ax2
where Dx is the diffusion constant in the direction normal to the surface, cis
the moisture concentration, x is the distance from the edge of the plate
normal to the surface, t is the time.
For fibre composites, diffusivities perpendicular (t) and parallel (1) to the
fibres can differ and
(8.33)
where a is the orientation angle of the fibre to the x direction (for in-plane
fibres a = 90°). Dx is related to the diffusivity of the matrix D m ,
Dx = Dm [(1 - Vf) cos 2 a + (1 - 2VVfln) sin 2 a] (8.84)
and
Dr = (1 - Vf)Dm (8.35)
Dt = (1 - 2VVfln)Dm (8.36)
D can be determined from the moisture absorption kinetics since it can be
shown that:
M = 4Moo "\ / i D (8.37)

h Vn
where
D = Dx ( 1+
h
- yDy- + -h yDz)2
- (8.38)
1 Dx n Dx
where h, I, n are thickness, length and width respectively of the coupon.
M is the moisture concentration at time t and Moo is the equilibrium (i.e.
maximum) moisture content at infinite time.
For a homogenous resin casting Dx = Dy = Dz equation 8.38 reduces to
D = Dx ( 1 + I +
h h)2
~ (8.39)
The distribution of moisture through the thickness of a composite can be

computed from these equations. Figure 8.15 illustrates that the moisture is
unevenly distributed throughout the laminate for a significant proportion of
its service life. The effect of thermal excursions on the moisture distribution
profile can be appreciated from the effect of temperature on D. Since D
increases with temperature, the moisture profile tends to even out as a
result, as shown in Figure 8.15.
(a) Thickness (mm)
(b) Thickness (mm)
Figure 8.15 Theoretical moisture distribution within a unidirectional bismaleimide epoxy

(Narmco 5245C) laminate, (a) subjected to 96% RH at 50°C for 3 days; (b) above following a
rapid thermal excursion (thermal spike) to 150°C.
8.6.2 Effect of resin structure
The value of M", is strongly dependent on the polarity of the cured resin
structure. Generally the more polar, the higher the equilibrium moisture
concentration. As a result, hardeners which yield 'polyester' or 'polyether'
molecular structures rather than hydroxy amines are preferred for matrices.
Absorbed water acts as an efficient plasticiser for cured epoxy resins and
as a consequence the glass transition temperature of the resin is reduced. As
shown in Figure 8.16 and as a general rule for most resin systems, the Tg is
reduced by::::: 20 Kl1 % moisture (Wright, 1981).
0
'\
\
10 - \
\
\
20 L \
\
6' I
30 r
\
e.... \151
\
0
- ~
~
::J
....Cll
ctS
a.
40 'J \
i;.
III \
\
- ~
E 50 \
Cll \
c I
(;
\ 0
0 60 r \
+= \
151
'iii
-
c \
~ 70 (;
\ 0
(;\ 151
/Jl
/Jl \
ctS 80 \
0> III \
.!: \
Cll 90 \ III
/Jl
ctS \
....
Cll \
100 \
(.)
Cll 'J ~x
0 \
\\
151
110 (;
\
120 \
(; (; \c:,
\
130 \
\
(; (; \
140
0 2 3 4 5 6 7 8
Moisture content (%)
Figure 8.16 The effect of moisture on Tg of a series of differing epoxy resins showing generality
of the effect of moisture plasticisation. Each symbol represents a different resin system.
(Redrawn from Wright, 1981).
250
200
x
~ 150
OJ
"""" 100
50
o ~~--~--~--~--~--~--~--~
o 2 3 4 5 6 7
M (wt%)
Figure 8.17 Change in Tg of Narmco 5208 epoxy resin as a function of equilibrium moisture
content. Each symbol represents the results of different authors where variations in cure give
differing 'dry' values. The wet Tg however tends to the same value in all work.
Figure 8.17 indicates the consistency of this observation with the variation
of Tg with differing moisture concentrations for nominally the same resin
system, as given by different workers (Mulheron, 1984).
The major consequence of this result is illustrated in Figure 8.18 where it
is seen that not only the glass transition temperature but also the modulus is
reduced. Since for most composite matrices, the average moisture
absorption at equilibrium leads to a reduction in Tg by about 50°C, 150-
\
"- '-
,
W
::J
::;
'-- ---
wet/- -, ......
"C
o '-
E "-
OJ
>
~
"-
Qi
\
II: \
\
\
'l
o 150 300"C
Temperature
Figure 8.18 The effect of moisture on the temperature dependence of the relative matrix
modulus of Narmco 5208 epoxy resin.
200°C resins can only be used where service temperatures are limited to 100-
150°C. This has led to the development of more highly functional resins
which have a higher cross-linked density, Tg and service temperature in
humid environments.
While simple resin systems appear to obey Fickian diffusion kinetics
under absorption and desorption conditions, the compounded resins
employed for advanced composites may show irreversible effects. For
example, the presence of residual DICY can increase the value of Moo
significantly. Moreover, on immersion in water, blister formation may occur
as a result of osmotic effects accompanied by leaching (Jones et al., 1987).
As shown in Figure 8.19 the presence of a second polymeric component may
lead to moisture-induced changes in the matrix-dominated expansion
coefficients and this can have severe effects on micro crack development in
laminates during thermal excursions (section 8.5.4) (Jacobs and Jones,
1991).
150
C
(!)
'(3
~a
u
c:: 100
.2
<n
c::
:.::
'0.x"
OJ
'f
~
Cil ~
E
OJ
.c:: 50
f-
o
o 50 100 150 200
Temperature (oC)
Figure 8.19 The temperature dependence of the transverse thermal expansion coefficients of
the bismaleimide-modified epoxy resin based carbon fibre composite (Narmco S24SC). Dry
(lower curve) and wet (upper curve). The cross hatching indicates the relative thermal strain
induced on cooling from Terit (dry) and Terit (wet) (lower and upper curves respectively). Tcrit is
the temperature to which a crossply laminate needs to be heated and then cooled to induce a
f
value of f:~ which exceeds tu and cause thermal cracking. Since Tg (dry) is below Terit (dry)
thermally-induced cracking does not occur but the reverse situation operates for the wet
laminate and thermally-induced cracking does occur (Jacobs and Jones, 1991).
8.6.3 Effect of moisture on thermal residual strains

Since absorbed moisture is accommodated in the molecular structure of the
resin through dipole-dipole and hydrogen bond interactions, this leads to
swelling. Much debate has centred around the extent of this effect and the
accommodation of water in the free or unoccupied volume of the dry glass.
However, matrix swelling negates the restrained contractions which had
already occurred on cooling through the strain-free temperature from the
post-cure temperature after fabrication. As a result, the thermal strain is
reduced by moisture ingress as shown in Figure 8.20.
For other systems the effect is bigger and, for example with Fibredux
913G, the residual stresses are reduced to a very low level as a result of
moisture absorption.
8.6.4 The combined effect of humidity and thermal excursions

In service, composites often undergo thermal excursions and since under
most conditions moisture absorption continues at the same time, the
combined effects are important. For a moisture-equilibrated laminate, in
which the thermal strains have been reduced, excursion to a higher
temperature could lead to an enhanced level of thermal strain than present
in the as-fabricated material. This occurs when am(wet) > am(dry) while the
reduction in strain free temperature Tl (equation 8.25) which is related to
Tg , is insufficient to offset the corresponding increase in (at - al). The effect
is illustrated in Figure 8.19 for a bismaleimide-modified epoxy. In this case
since the magnitude of al is such that it can be ignored, then Etl is directly
proportional to the area under the curve. It can be seen that the enhance-
1.1
1.0
•
:>..
•
L
== 0.9
"U
(,)
"-:;:::
:5
(,)
0.8
O. 7 '----'----'----L_"'--....L..---'----'_"'---'---L..---'~
o 0.1 0.2 0.3 0.4 0.5 0.6

Moo (%)
Figure 8.20 The reduction in relative thermal strain induced into the transverse ply (€:~) of an
epoxy-based carbon fibre crossply laminate with differing equilibrium moisture levels. A
typical value for a dry 02/90 2/0 2 laminate is 0.63%.
ment in (at - a,) more than offsets the decrease in (Tl - T2)' Furthermore,
the transverse cracking strain is also reduced as a result of moisture ingress,
so that whereas a thermal cycle to Tcrit(dry) for the dry laminate does not
lead to microcrack formation, a thermal cycle to Tcrit(wet) does. Thus a wet
laminate subjected to a series of thermal cycles can lead to significant levels
of damage. A corollary is that moisture absorption by the prepreg can also
lead to significantly enhanced levels of thermal strain in the as-fabricated
laminate which may cause process difficulties.
8.6.5 Thermal spiking

Since the timescale for moisture equilibration is long, the effect of thermal
excursions on the kinetics of diffusion is also of importance. This is
especially relevant in aerospace applications where for example a supersonic
dash by an aircraft can raise the laminate temperature momentarily by
= 100°C. There have been several reports of enhanced moisture absorption
as a result of a thermal spike. A typical example is shown in Figure 8.21
where the isothermal moisture absorption is compared to that which occurs
under non-isothermal conditions.
During the thermal spike, moisture is lost but on reimmersion in the
humid environment, the moisture climbs to a higher value. The exact
0.8
?f- 0.6
8
::'E
0.4
0.2
a 0.8 1.6 2.4

yItlh (x 103 s1l21mm-')
Figure 8.21 Comparison of moisture absorption by a 0° carbon fibre epoxy laminate (Narmco
5245C) under isothermal conditions (continuous line) and subject to intermittent thermal
spikes to 150°C (saw-tooth). The environment was 96% RH at 50°C. The saw-tooth illustrates
partial drying during thermal cycling and the subsequent enhanced water gain.
mechanism of this phenomenon is not fully clear but may involve the re-
equilibration of the network molecular structure, since the critical thermal
spike temperature above which this phenomenon is observed appears to be
related to the glass transition of the wet resin.
8.7 Selection principles
Epoxy resin-based composite materials generally find application in high

performance applications where high specific strength and stiffness
dominate the requirements (aerospace structures or sports goods). In most
of these cases, therefore, carbon fibre reinforcements are preferred. The
grade of fibre employed is determined by a high strength or high stiffness
criterion. For the former, Type A fibres and for the latter HM or 1M fibres
are employed. On a cost per equivalent strength, carbon fibre composites
compare favourably with the cost of aluminium, its natural competitor.
However, cost savings during fabrication can make the composite solution
beneficial. On similar arguments glass fibre-reinforced epoxy is highly
competitive and economic at:::::: 113 the material cost. In comparison to other
plastics the cost of epoxy glass composites is low (except for polyester-based
laminates) but where the fabrication rate and the complexity of the
moulding dominates, other plastics solutions win out.
On a cost for similar stiffness basis, glass fibre laminates still compete
effectively with high performance metals such as titanium but the carbon
fibre solution is expensive and fabrication economies or lower running costs
(in the case of airlines) need to be considered in the design.
It is often advisable to use hybrid systems which combine the benefits of
glass and carbon reinforcements, as in the case of the prop-shaft for a motor
vehicle and the rotor blade of a helicopter.
Aramid fibres offer higher performance than glass fibres but the off-axis
properties are somewhat poorer. Therefore aramids find application where
high stiffness and strength are combined with impact performance. Where
high retained compression strength after impact is required, as for structural
aerospace components, aramids are generally unsuitable.
Toughened epoxy resin textile fibre (nylon) reinforced films (e.g.
Redux 330) are used as adhesives for joining composite laminates. These
represent an additional application of epoxy resins in composites materials
engineering.
8.8 Conclusions
Epoxy resins form the basis of the advanced composites industry as the
preferred matrix for carbon fibres. These systems are designed to give good
process viscosity profiles and involve latent hardeners for use in prepreg
technology. The micro mechanics of laminate systems is strongly dependent
on the properties of the matrix. However, the absorption of moisture and
the maximum achievable glass transition temperature can limit the
maximum service temperature of the composite to < 200°e.
8.9 Glossary of symbols
8.9.1 Laminates
D Interfibre spacing
E(E) Young's modulus (average value)
G Shear modulus
L Length of composite coupon
T2 Measurement temperature in equation 8.25
Tj Strain-free temperature
V Volume fraction
b Outer ply thickness
c Specimen width
d 90-ply semi thickness
df Fibre diameter
rf Radius of fibre
t Crack spacing
y Shear lag transfer length
a Linear expansion coefficient
Yt Fracture surface energy of the 90° ply parallel to fibres
<5 Displacement of 0°/90° beam from flat
c Strain
a Stress
e Radius of curvature of a 0°/90° beam
T Shear stress
£ Tsai-Halpin geometry and dispersion factor
'YjO,'Yjl Orientation (0) or length (1) factors for random discontinuous
composite
v Poisson ratio
Diffusion theory
D Diffusion constant or diffusivity
M Moisture content at time t (in %)
Moo Moisture content at t = 00 (in % )
C Moisture concentration (g . mm -3)
Time
x Distance into the specimen normal to the flux

a,fJ,y Orientations of the fibres with respect to x, y, z dimensions
h Specimen thickness
I Specimen length
n Specimen width
Subscripts
a Applied stress (oa) or strain (e a)
f Fibre (e.g. Of stress in fibre)
m Matrix (e.g. em strain in matrix)
c Composite (or Lc critical length for a short fibre in a composite)
1 Longitudinal or 0° direction of a unidirectional ply
t Transverse or 90° direction of a unidirectional ply
u Ultimate value at failure
fu Ultimate value for failure of a fibre
mu Ultimate value for failure of the matrix
lu Ultimate value for failure of a longitudinal composite
tu Ultimate value for failure of a transverse composite
tIu Ultimate value for transverse cracking of transverse ply
tu Ultimate value for failure of a transverse ply
tI Property of transverse ply in longitudinal direction
It Property of longitudinal ply in transverse direction
11 Property oflongitudinal ply in longitudinal direction
Crit Critical value
Superscripts
th Thermal
Property in the component when other fails
P Poisson
Min Minimum value of property
References
Alexander, M.R. and Jones, F.R. (1991) In Composites, Design Manufacture and Application,
Vol. 1, eds Tsai, S.W. and Springer, G.S. Sampe, Corina, USA. Chapter 11, paper G.
Amdouni, N., Sautereau, H., Gerard, J-F. and Pascault, J.P. (1990) Polymer 31,1245-1253.
Aveston, J. and Kelly, A. (1973) I. Mat. Sci. 8,352.
Bacon, R. (1980) Phil. Trans. R. Soc., London A294, 437.
Bader, M.G. (1985) In Handbook of Composites, Vol. 4, ed. Kelly, A. and Mileiko, S.T.
Elsevier, North Holland, p. 177.
Bailey, J.E. and Parvizi, A.11981) I. Mat. Sci. 16,649.
Cox, H.L. (1952) Br. I. Appl. Phys. 3,72.
Denison, P., Jones, F.R. and Watts, J.F. (1988a) Surface and Interface Analysis 12, 455.
Denison, P., Jones, F.R. and Watts, J.F. (1988b) In Interfaces in Polymer, Ceramic and Metal
Matrix Composites, ed. Ishida, H. Elsevier, New York, pp. 77-86.
Denison, P., Humphreys, P., Watts, J.F. and Jones, F.R. (1989) In Interfacial Phenomena in
Composite Materials, ed. Jones, F.R. Butterworths, Guildford, pp. 105-110.
Dobb, M.G. (1985) In Handbook of Composites, Vol. 1, eds Watt, W. and Perov, B.V.
Elsevier, Amsterdam, pp. 673-7.
Dunsford, D.V., Harvey, J., Hutchings, J. and Judge, C.H. (1981) RAE Technical Report
TRSJ096, HMSO, London.
Drzal, L.Z., Rich, M.J. and Lloyd, P.F. (1982) J. Adhesion 16, 1.
Folkes, M.J. (1982) Short Fibre Reinforced Thermoplastics, Research Studies Press,
Letchworth, UK.
Garrett, K.W. and Bailey, J.E. (1977)1. Mat. Sci. 12,157.
Gillham, J .K. (1986) Encyclopaedia Polym. Sci. Eng. 2nd edn, Wiley, New York, pp. 519-524.
Halpin, J.G. and Tsai, S.W. (1969)AFML-TR67, 423.
Hashin, Z. and Rosen, B.W. (1964) J. App. Mechanics Trans. ASME 31,223.
Hull, D. (1981) An Introduction to Composite Materials, Cambridge University Press, UK.
Jacobs, P.M. and Jones, F.R. (1991) In Composites, Design, Manufacture and Application,
Vol. 2, Ch. 16. eds Tsai, S.W. and Springer, G.S. SAMPE, Corina, USA, paper G.
Jones, F.R. (1989) In Interfacial Phenomena in Composite Materials ed. Jones, F.R.
Butterworths, Guildford, pp. 25-32.
Jones, F.R., Shah, M.A., Bader, M.G. and Boniface, L. (1987) Sixth Int. Conf. on Compo
Materials, 2nd European Conf. Compo Materials ICCM VIlECCM2 Vol. 3, ed. Matthews,
F.1. et al. Elsevier, London, pp. 443-436.
Kelly, A. (1973) Strong Solids. Clarendon Press, Oxford.
Kelly, A. and Tyson, W.R. (1965) J. Mech. Phys. Solids 13, 329.
Kies, J.A. (1962) US Naval Research Report No. 5752.
Kimber, A.C. and Keer, J.G. (1982) J. Mat. Sci. Letters 1, 353.
Loewenstein, K.L. (1983) The Manufacturing Technology of Continuous Glass Fibres.
Elsevier, Amsterdam.
Manders, P.W., Chou, T-W., Jones, F.R. and Rock, J.W. (1983) J. Mat. Sci. 18,2876.
Mortimer, S., Ryan, A.J. and Stanford, J.L. (1992) Proc. 5th Int. Conf. FRC '92, paper 2
Plastics and Rubber Inst, London.
Mulheron, M.J. (1984) PhD Thesis, University of Surrey, UK.
Oberlin, A. (1985) In Developments in the Science and Technology of Composite Materials eds
Gunsell, A.R., Lamicq, P. and Massiah, A. Elsevier Applied Science, London. pp. 15-20.
Parvizi, A. and Bailey, J.E. (1978) J. Mat. Sci. 13,2131.
Parvizi, A., Garrett, K.W. and Bailey, J.E. (1978) J. Mat. Sci. 13,195.
Peters, P.W.M. (1984) J. Compo Mat. 18,845.
Peters, P.W.M. and Chou, T-W. (1987) Composites 18, 40.
Plueddemann, E.P. (1990) Silane Coupling Agents 2nd edn. Plenum, New York.
Purslow, D. and Childs, R. (1986) Composites 17,127.
Rand, B. (1985) In Handbook of Composites, Vol. 1 eds Watt, W. and Perov, B.V. Elsevier,
Amsterdam, pp. 495-575.
Rosen, M.R. (1978) J. Coatings Tech. 50,70.
Schapery, R.A. (1968) J. Compo Mat. 2,380.
Shen, C.H. and Springer, G.S. (1977) J. Compo Materials 11,107.
Thomason, J. and Morsink, H. (1988) In Interfaces in Polymer, Ceramic and Metal Matrix
Composites, ed. Ishida, H. Elsevier, New York.
Tsai, S.W. (1985) Composites Design, USAF Mat. Lab. - Think Composites, New York.
Tsai, S.W. and Hahn, H.T. (1980) Introduction to Composite Materials, Technomic, Westport.
Verpoest, I. (1987) In Proc. Europ. Symp. on Damage Development and Failure Processes in
Composite Materials, eds Verpoest, I. and Wevers, M., Leuven, Belgium, pp. 70-77.
Waltersson, K. (1985) Compo Sci. Tech. 22,223.
Wang, D., Jones, F.R. and Denison, P. (1992) J. Adhesion Sci. Tech. 6,79.
Watt, W. (1985) In Handbook of Composites, Vol. 1. eds Watt, W. and Perov, B.V. Elsevier,
Amsterdam, pp. 327-387.
Wright, W.W. (1981) Composites 12, 201.
Further reading
Chawla, K.K. (1987) Composite Materials - Science and Engineering, Springer-Verlag, New
York.
Hull, D. (1981) An Introduction to Composite Materials, Cambridge University Press,
Cambridge.
Jones, R.M. (1973) Mechanics of Composite Materials, McGraw-Hili, Washington DC.
Kelly, A. (1973) Strong Solids, Clarendon Press, Oxford.
Kelly, A. and Rabotnov, Yu.N. (1985) Handbook of Composites, Vols 1-4, Elsevier Science,
Amsterdam.
Tsai, S.W. (1985) Composites Design 1985, Think Composites, Dayton, Ohio.
Tsai, S. W. and Hahn, H.T. (1980) Introduction to Composite Materials, Technomic, Westport,
Connecticut.
Donnet, J.B. and Bansal, R.C. (1984) Handbook of Fiber Science and Technology, Vol. 3,
Marcel Dekker, New York.
Plueddeman, E.P. (1982) Silane Coupling Agents, Plenum, New York.
Weatherhead, R.G. (1980) FRP Technology, Fibre Reinforced Systems, Applied Science,
London.
9 Coatings and other applications of epoxy resins
X.M. CHEN and B. ELLIS
9.1 Introduction
The current use pattern for epoxy resins is similar to that suggested by Lee
and Neville in 1967 who estimated consumption as coatings 55%, composite
matrices 20%, castings 10%, adhesives 5% and miscellaneous 10%. These
figures may be compared with the 1990 and 1991 estimates given in chapter
1, Table 1.1, even though the categories are not identical. Lee and Neville
(1967a) pointed out that there are difficulties in obtaining exact
consumption estimates. However, the main difference is that coatings
have a slightly lower share of the epoxy market. Epoxy resins are usually
more expensive than their rivals, such as phenolic resins for coatings or
laminates. Therefore, epoxy resins find application because of their superior
properties, which include both processing and those of the cured resin. The
processing is convenient since it is possible to formulate compositions with
the required rheological properties, such as low viscosity, and there is also a
wide choice of hardeners so that it is possible to cure at ambient as well as
elevated temperatures. Because epoxy resins can be cross-linked without
the formation of low molecular weight products, volatiles are not evolved
during cure, and the resins have only a relatively low shrinkage during
curing. Their mechanical and electrical properties are superior to other
resins and they have good heat and chemical resistance.
There is a large range of epoxy resins available commercially and some of
the more important types are listed in Table 9.1. Exact specifications can be
obtained from resin suppliers and many mixtures are offered by
compounders. Selection of a resin-hardener system can be made by a critical
assessment of processing and end-use requirements, e.g. the electrical
properties, or resistance to UV radiation. It is possible to enhance a desired
parameter simply by altering the cure schedule without a change in the initial
composition of the resin-hardener system. Thus, it is possible to increase
the glass transition temperature (Tg) of the cured resin by post-curing.
Such effects can be appreciated by study of the time - temperature-
transformation diagram, TIT, given in Figure 1.2 and the detailed
discussion of cure given in chapter 3. Extensive data compilations of the
properties of epoxy resins are available and Table 9.2 is a selection of the
'tensile' mechanical properties and also an estimate of the heat distortion
temperature. The Young's modulus of epoxy resins is somewhat lower than
Table 9.1 Summary of commercially available epoxy resins a
E.e.w. Viscosity Applications

(250°C) cp
1. Bisphenol A-epichlorohydrin resins

a. Liquid resins
i. Wide ranging characteristics 95-450 5000-30000 Table 9.3
ii. Chain extension 185-200 2000-7000 Contain specific catalysts for conversion of these resins
iii. Lower viscosity resins Contain reactive diluents, e.g. glycidyl ethers of iso-octanol or
butane-l,4-diol
b. Solid resins
i. So. pt. 60-180°C 500-6000 Coatings, especially powder coatings
Manufacture of epoxy resin esters for coatings
With amino- or phenolic-resins; high chemical resistance for
tube coating primers and stoving enamels
Castings
Prepregs
ii. Mixtures and solutions Solvents include butoxy( ethoxy )ethanol, xylene, acetone,
methyl isobutyl ketone and combinations
2. Epoxy-novolacs
Derived from bisphenol A-epichlorohydrin resins
i. Phenol, liquid 170-190 30000-90000 Coatings, encapsulators, laminating, moulding, adhesives
ii. Phenol, solid (So. pt. 90-100) 190-220 Moulding powder
(So. pt. 90-95) 275-330 Adhesives, coatings
iii. Cresol, solid (So. pt. 35-90) 200-240 Adhesives, coatings, mouldings
3. Polyhydric phenols
i. Bisphenol F
Diglycidyl ether of bisphenol F 158-175 5000-7000 Additive for BPA resins in coating and flooring, e.g.
CH, Araldite 07281
- - O-CH,-CH-CH,
"(~rn-cHoJ9 ~ /\
ii. Others include:
Tetraphenol ethane derivatives Powder coating
Polyglycidyl ether of tetraphenol ethane: Shell 1031
"O~ )§l0"
5§5rn-rn~
HO OH
4. Amines
Reaction products of epichlorohydrin with
amines such as:
~
r$) NH2
C$JNH2
~ NH2
DDM Aniline p-amino-phenol
Triglycidyl p-amino phenol Union Carbide ERL-0500

5. Halogen-containing epoxides Flame retardancy applications: general purpose laminating,
moulding
6. Cyclo-aliphatic resins
Range of resins lower viscosities than BPA Good UV light stability, electrical properties: general purpose
resins casting resin, filament winding. Acid scavenger, plasticizer
3 ,4-epoxycyclohexylmethyI3 ,4-epoxy- 131-143 350--450 Union Carbide ERL-4221
cyclohexane carboxylate
o
°D~-O-CH2-o0
So. pt. = softening point, 0c. a For a fuller list see Tanaka (1988).
Table 9.2 Mechanical properties and range of heat distortion temperatures of cured epoxy
resins
Mechanical properties
Young's Tensile Elongation Heat distortion

modulus strength at break temperature
(GN. m- 2 ) (MN .m- 2 ) (%) (0C)
Badge type:
Flexible 0.1-1 10-70 5-10
Rigid 1.5-3.5 25-90 2.5-4 100--200
Epoxy-novolac 1.5-3.5 70 2 150-250
Cycloaliphatic epoxy 3.5 90 2-10' Upto200
Phenol-formaldehyde 2-4 20-60 1.5 150-170
Unsaturated polyesters 1.5-3.5 40-90 <2 Up to 200
, With up to 30 phr of a polyol flexibilizer, elongations at break may be up to 30%.
that of phenol formaldehyde (P/F) and unsaturated polyester resins but the
tensile strengths are comparable and may be higher. This is because the
extensibility, that is the fracture strain, of the epoxy resin is much higher
than that of the 'brittle' resins. It is possible to select from a range of
flexibilizers (see chapter 4) to 'tailor' the mechanical properties to suit the
application requirements. For instance the 'tensile strength' can be
significantly affected by cure treatment and the fracture behaviour of epoxy
resins should always be considered when an application demands optimum
strength. There is a detailed discussion of fracture behaviour in chapter 5
which is also relevant to the adhesive (chapter 7) and composite (chapter 8)
applications of epoxy resins.
With standard BPA-EpiCI resins the heat distortion temperatures are
similar to those of PIF and polyester resins, but with suitable hardeners and
a post-cure at elevated temperature it is possible to attain higher glass
transition temperatures and hence higher use temperatures. For instance a
BADGE-type resin cured with trimellitic anhydride and cured for 1 hour at
120°C followed by 2 hours at lS0°C may have a heat distortion temperature
of up to 2S0°C.
Cyclo-aliphatic resins have very good outdoor ageing properties. This is
because they do not contain aromatic rings which absorb the UV radiation
which leads to degradation processes. Also, the chlorine, and other halogen,
content of the cyclo-aliphatic resins is essentially zero and hence these resins
have very good electrical properties.
Data compilations such as that given in Table 9.2 are indicative of the
properties that may be achieved with epoxy resins. However, by comparison
with the PIF and unsaturated polyester resins it would not be possible to
assess the superior technical properties of the epoxy resins. Their processing
properties offer considerable advantages. The range of hardeners available
is a major asset since it is possible to formulate a curing system to suit almost
Table 9.3 Some typical curing agents and applications for BADGE-type resins
Hardener' Physical state Concentration Typical cure schedule Approximate heat Suggested applications 3
at ambient (phr)2 distortion temperature
COC)
1. Aliphatic amine
Triethylene tetramine Liquid 13 a. 25°C, 7 days a. - General purpose
b. 25°C, 1 day plus post- b. Up to 120 Co, A, Ca, M, F, Sand
cure at elevated T FW
leffamine D-400 Liquid 55 1l0-120°C, 30 min -30 Flexible resin Co, A,
Ca,FandS
2. Aromatic amine
DDM Solid 27 1 hat 150°C Upto 150 Glass cloth laminate
+3hat 180°C
DDS (orDAS) Solid 25 5 h at 125°C Upto175 A, Ca, M, E, Sand FW
+ 1 hat200°C
3. Polyamides Liquid Depends on type a. Ambient temperature Up to 100 a. Co,A,Ca,F,M
7 days andS
b. 1 h at 100°C Up to 100 b. CaandA
4. Anhydrides
Methyltetrahydro- Liquid 80 2hat20°C 130 Liquid hardener
phthalic anhydride +2hat 150°C processing
(MTHPA) A,Ca,M,SandFW
Trimellitic anhydride Solid 35 1 h at 120°C 250 High heat distortion
(TMA) +2hat 150°C temperature
A,Ca,M,SandFW
5. Miscellaneous
Dicyandiamide (DICY) Solid 4 1 h at 180°C 150 A, Ca, M, E, Sand FW
BF3-Monoethylamine Solid 3 1 hat 120°C 170 A, Ca, M, E, Sand FW
(BF3-MEA) +2hat 170°C
, See chapter 2 for fuller details. Many hardeners are mixtures and/or derivatives and are not included in this list. Many other hardeners may be used. Shell
list 30 curing agents for Epon 828. This table is only illustrative.
2 phr = parts per hundred of resin by weight.
3 Code: Co = coatings; A = adhesives; Ca = castings; E = electrical laminates; F = flooring; M = mouldings, S = structural laminates; FW =
filament winding.
every specific requirement. The published tables can often be very

misleading. For instance, in such data compilations the water absorption is
often reported in terms of the w/o absorbed by Va inch thick specimen in 24
hours so that the value will not give either an estimate of the amount or even
an indication of the rate of absorption. Also, the heat resistance and
especially the solvent resistance of the resins are not adequately assessed.
The epoxy resins have very good solvent resistance, which is a major
advantage in many applications.
It is important to note that there is continual improvement and innovation
in the resins, hardeners and other additives available. Resin manufacturers
and converters have a policy of upgrading their products and offering
speciality resins for particularly demanding applications. Laboratory
samples and experimental products are usually available from resin
manufacturers. The range of applications for a BADGE-type resin with
different curing agents is given in Table 9.3.
In the present chapter it is not possible to discuss each application of epoxy
resins in any detail, its purpose is to give an overview of the range of
applications that currently provide a market for epoxy resins. It is important
to note that many of the compositions that are used industrially contain
many components, which are often not disclosed. Thus, full technical
information on a product should be obtained from the resin manufacturer,
compounder or supplier.
9.2 Surface coatings
9.2.1 Introduction
Surface coatings may be primarily applied for protective or decorative
purposes but usually they have dual or more functions, such as exterior
automotive coatings which are both protective and decorative. They may
have specialized functions, such as anti-fouling paints for ships' bottoms or
the protection of food and beverages in metal cans. There are also 'high-
tech' applications in the electronic and related industries such as coatings for
fibre optics, printed circuit panels and many other applications (Burkhart,
1988; Emerson etal., 1990; Bonneau and Burkhart, 1991). Coatings contain
a polymeric film-forming material, pigment, special additives and may
be dissolved in a suitable solvent mixture or dispersed in water. Powder
coatings do not contain a solvent. There is a range of film-forming materials
which may be classified as either non-convertible or convertible (Martin,
1972). The former includes shellac, cellulose nitrate or acetate, cyclized and
chlorinated rubber and vinyl and acrylic resins, which form films when a
solvent evaporates. Convertible materials include the drying oils, varnishes,
thermosetting resins such as alkyds, phenolics, epoxies and polyurethanes in
Table 9.4 Film forming materials for coatings
Epoxy resin mixtures Other thermosetting resinsc
Application Epoxy Acrylic Alkyd Polyester PlFa AIFb Coal tar Alkyd Polyester AIFb P/F a Polyurethane
resins modified
Can and drum coatings V V V V V V V V

Heavy duty plant V V V V V V V V
maintenance coatings
Automotive - primers, V V V V V V V
topcoats and respray
Pipe coatings V V V
Appliance finishes V V V V V V V V
Marine coatings, primers V V V V V V
and top coats
Coil coatings V V V
Metals, general primers V V V V V V
and top coats
aP/F is a phenol-formaldehyde resin; b AIF includes urea and melamine formaldehyde resins; C Non-convertible coatings include drying oils and
cellulose ethers and esters.
which the film former becomes cross-linked either by reaction with oxygen
or suitable curing agents or hardeners. A general text on surface coatings
is that of Paul (1986).
There are many different coating methods which can be adopted for
specific applications, depending on the coating thickness, coating speed and
type of finish required. A review of 25 methods with outline descriptions is
given by Scharenberg (1985), spray coating is discussed by Coeling (1985)
and powder coating by Richart (1982, 1985). A listing of the polymer
systems used for specific industrial coatings is given in Table 9.4. Epoxy
resin coatings are discussed in detail by Somerville and Smith (1973), Potter
(1970), Lee and Neville (1967) and Tess (1988a). Resin manufacturers offer
advice on the use of epoxy resins for the fomulation of coating compositions,
Ciba-Geigy (1988) and Shell (1991) offer a range of epoxy resins for surface
coatings, including Epon 828, with an epoxy equivalent-weight of 185-192.
This resin and also the Ciba-Geigy equivalent GY-250 have been
formulated as a heat-cured coating for strengthening glass (Chen et al.,
1992). Epoxy resins can be formulated to have much shorter cure times (1 to
30 minutes
at 240-135°C) than equivalent polyurethanes or polyesters (10 minutes
at 200°C) and have excellent adhesion to most substrates, provided the
surface has had an appropriate pre-treatment.
9.2.2 Surface preparation and primers

It is a primary requirement that coatings, when cured, should form a
continuous film. This depends not only on the rheological properties of the
epoxy formulation when applied, but also and in many cases more import-
antly, on the interfacial tension of the coating formulation and the surface
that is to be coated. Any surface features which may hinder film formation
must be treated prior to the coating operation. For the surface to be readily
wetted by the epoxy formulation, a critical condition is that the contact angle
between the formulation and the solid surface is low. Simple cleaning of the
surface by appropriate solvents may be sufficient surface preparation when
there is only surface contamination and non-adherent scale is absent. For
example iron or steel should be rust free. On the other hand, if the coating is
designed to do more than just cover the surface, specific treatments will be
required. For example, the application of a strengthening coating to glass
surfaces requires exposure to water vapour prior to coating to optimize
hydrolytic durability (Ellis et al., 1991). Other surface preparation methods
include mechanical abrasion, chemical etching, and flame treatment, to
name but a few. These are treatments used to modify different surfaces in
order to improve the adhesion of coatings. In chapter 8, there is a discussion
of surface preparation to achieve good adhesion and the same principles are
involved here. Misev (1991a) gives details of the surface treatment of
COATINGS AND OTHER APPLICATIONS 311
aluminium, steel and galvanized steel prior to the application of a powder

coating.
A primer may be used to improve the 'wettability' of the surface and also
to protect a cleaned surface so that it will not become contaminated before
application of the coating. The primer has to adhere well to the substrate
surface and offer good wet ability and bonding of the coating. Silane and
other coupling agents can be used to improve bonding, especially to
hydrated surfaces. Typical coupling agents have two types of reactive
groups, one reacts readily with inorganic substrates and the other with
organic groups. Thus, coupling agents are often used in applications where
inorganic surfaces are to be bonded to an epoxy resin, and they may be
incorporated either into a primer or the coating. Coupling agents are
discussed in chapters 8 and 9.
9.2.3 Solution coatings

An important class of epoxy coatings are those in which the rheological
properties of the composition are adapted for the specific applications by the
use of a suitable solvent or solvent mixture. The rheological properties will
depend on the resin, its type and molecular weight, the pigment and its
concentration and the solvents. The apparent viscosities are normally within
the range of 0.05-1.0 Pas (0.5-10 poise), but it should be noted that
viscosities are highly temperature-dependent. Factors which determine the
selection of solvent include its vapour pressure and hence its rate of
evaporation and flash point, its toxicity as well as its effect on the flow
properties of the composition. However, the paramount condition for
selection of solvent is its miscibility with the epoxy resin together with the
hardener and any other additives present. Solutions should be stable for the
required shelf-life of coating composition and during application.
Selection of a solvent or solvent mixture which is miscible with epoxy
resins is aided by matching the Hidebrand's solubility parameters, <5 (see
Barton, 1983a), of the polymer and solvent. This is defined as
112
<5 = C 1I2 = ( - ~) = (ilf U/V)1I2
where C is the cohesive pressure, or cohesive energy density (cohesive

energy per unit volume), U the molar internal energy, that is the molar
potential energy of the material relative to the ideal vapour pressure at the
same temperature, V the volume, the ilf U is molar vaporization energy,
that is the energy required to vaporize a mole of the liquid to its saturated
vapour. For volatile liquids the solubility parameter (Grulke, 1989) is more
readily calculated from the molar enthalpy of vaporization, il TH,
<5 = [(Llf H - RT)/v]112
where R is the gas constant and T the absolute temperature. For the use of
this equation the vapour phase must obey the ideal gas law. For epoxy resins
the value may be determined from swelling measurements or by calculation
useing additive group cohesive energy constants (see Barton, 1983).
The simple matching of solubility parameters is only suitable for an initial
selection of a solvent because other factors such as polar attraction force and
hydrogen bonding complicate solvent-solute interactions. Tess (1988b)
gives useful diagrams for solvent selection and Barton's (1983) solubility
maps for epoxy resins can also be consulted. Solvents for epoxy resins
include xylene, methyl ethyl ketone (MEK) , methyl isobutyl ketone
(MIBK) and proprietary solvents, such as oxitol. Mixed solvents are often
used so that not only solubility but the rate of evaporation and also the
rheological properties of the solution are optimized for the specific
application. Solvent mixtures include toluene with acetone, MEK, MIBK
and also xylene, MIBK and 2-ethoxyethanol (Lee and Neville, 1967).
After the selection of suitable solvents and subsequent formulation of a
coating solution, other factors need to be considered carefully to ensure
successful coating application. These include the condition of the surface
prior to coating, concentration ofthe solution, curing technique, pot-life of
the solution, etc. It should be noted that the coating solution is a dynamic
system within which chemical reaction (cure) progresses with time. There-
fore care should be taken in its storage. Also, the rate of evaporation of the
solvent in the system depends not only on the vapour pressure of the solvent,
but also the curing method used. Evaporation will be faster at higher
temperatures, but then the rate of cure is higher and the film may harden
before all solvent molecules diffuse out of the coating. The trapped solvent
molecules will affect the properties of the film which may be impaired and
may lead to poor adhesion. Thus, it is always important to consider all the
factors in order to establish a set of optimized application conditions.
The presence of pigment will have a considerable effect on the flow
properties of a coating solution. Nielsen (1977) discusses the effect of fillers
on viscosity and suggests a generalized relationship for the viscosity of a
dispersion, related to that of the fluid 'I,
/ _ 1 +AB¢2
'I '10 -1 - B1jJ¢2
where A = kE - 1; kE is the Einstein coefficient, B = 1 for rigid fillers,
¢2 is the volume fraction of pigment and ¢m is the maximum packing

fraction,
¢m = (True volume of the filler)/(apparent volume of the filler).
However, the flow properties of dispersion systems are often very

complicated and require measurements over an appropriate range of rates
of shear (Utracki, 1988). They may be Bingham bodies with a yield stress,
pseudoplastics or thixotropic.
9.2.4 Dip coats

As probably the oldest and easiest method to apply coatings, dipping has
intrinsic advantages over other coating techniques. For example, compared
with spraying, dipping produces minimum material waste and provides, in a
single dip, a coating layer covering all the immersed parts. Furthermore, in
the dipping process, objects to be coated do not need to be spun to ensure
full coverage, which is inevitable for many spraying operations. The minus
side of dipping is also obvious. Firstly, since the object has to be dipped into
the fluid epoxy coating solution or emulsion, the size of the container (dip
tank) has to be large enough to accommodate a reservoir of material plus the
volume of the immersed part. This means that larger objects which can be
sprayed with ease would be more difficult to coat by dipping. Secondly, since
the container has to be constantly open during continuous operation,
evaporation of the carrier (solvent or water) will cause a change in the
concentration of the system in the dipping tank. Some overflow
arrangement may be installed in order to overcome such problem as well as
providing agitation.
In the actual coating process, the following points deserve careful and
combined consideration for successful applications.
• It is important to minimize the evaporation surface of the liquid.
This requires well thought out tank design.
• The material with which the tank is made must not react with the
coating solution or emulsion.
• Anti-contamination measures should be considered, depending on
the type of application.
• Degree of agitation should be determined to suit the application. For
example, continuous agitation may be required for some emulsion
systems which are likely to agglomerate.
• When the formulation is light sensitive, such as in the case of epoxy
acrylate-based UV curable systems, a light shield should be installed
to prevent unnecessary initiation of cure of the system in the tank.
• Dipping and withdrawal speed are the factors which need careful
monitoring for control of the coating thickness. The concentration of
the coating solution and the temperature difference between the
liquid and to-be-coated objects are also factors which affect coating
thickness.
9.2.5 Epoxy emulsions and other water-based coatings

With the ever growing environmental concern over pollution by chemical
wastes it is expected that, in the not too distant future, tougher
environmental legislation will be introduced. This will certainly change
current industrial practice. As a direct consequence, many industrial
operations that are common today will have to be up-graded when chemical
wastes require further treatment before discharge into the environment, and
some may be banned all together. In either case, substantial capital invest-
ment will be inevitable to ensure on-going industrial production. Volatile
solvents are prime candidates for legislation, but at present the rate of
change is uncertain. Installation of solvent extraction and/or recovery
systems will involve additional costs and could mean that the operation loses
its commercial viability. The long-term solution may lie in changing the
carrier from organic solvents to water wherever possible. Research has been
carried out recently to extend the application of such water-based coating
systems.
To use water instead of solvent as carrier, the immediate difficulty is that
most epoxy resins do not form stable solutions in water. Thus, unlike organic
solvents, the system formulated will be heterogeneous at the colloidal level
when the resin is dispersed or emulsified with water as the continuous phase.
However, for many coating operations heterogeneity up to a certain scale
will not affect the properties of coatings. Water-based epoxy emulsion type
coatings are now readily available and also water soluble epoxy resins have
been produced by the introduction of hydrophilic groups.
Epoxy resin emulsions are produced by mechanical shearing to reduce the
particle size of the resin and addition of a suitable surface active agent to
stabilize the emulsion. One example, of the many of the emulsifiers
available that can be used to stabilize epoxy emulsions is
'O""-@-°frn,-rn,+ ]0
The hydrocarbon 'end' is soluble in the resin and the polar part of the
molecule is hydrophilic where the length of the ethylene oxide chain can be
chosen to suit the application. There are very many such emulsifying agents
available commercially, and mixtures may be used to ensure that an
emulsion is stable (Ash and Ash, 1980-1983). It is essential that the particle
or droplet size of the emulsion is less that one micro metre to ensure that the
emulsion does not settle out. Also, the emulsifying agent should be chosen
to ensure that the droplets do not aggregate or even worse, coalesce. The
stability of an emulsion is affected by shearing and hence the surface active
agent should form a protective layer around each droplet. Changes in
temperature also affect the hydrophile-lyophile balance (HLB) of the
surfactant (Shinoda and Friberg, 1986).
To recapitulate from this discussion it can be easily appreciated that two
basic criteria for the selection of emulsifiers must be satisfied.
• Reduce the interfacial tension between water and resin
• Form protective films around the resin droplets.
Although analytical methods can assist in the selection of emulsifier, a
direct screening of the available emulsifiers is essential in practice to ensure
that all requirements are satisifed.
There are three sets of conditions that an epoxy coating emulsion must
comply with:
1. Preparation, stability and shelf-life
2. Application of the coating
3. Film formation.
The essential factors for the first set have already been outlined. For the
second set the rheological properties of the emulsion must be such that the
emulsion can be applied either by dipping, spraying or electrodeposition.
With dipping the film thickness is determined by 'drainage' of the emulsion
as the coated product is removed from the emulsion. The 'quality' of the film
formed depends on even drainage, such quality involves control of the
coating thickness as well as the avoidance of surface blemishes. It is
particularly important that there is even coverage, which is difficult to
achieve by spray application at places adjacent to corners, which should be
avoided if possible by alternative product design. The drip that forms in the
region of the 'neck' as the product leaves the coating bath can be trouble-
some. With spraying it is essential that fine droplets are formed since when
large they splash where they hit which can lead to a 'more or less grainy
surface' which eventually leads to a film with imperfect specular reflection.
The rheology of the emulsion will affect size reduction in the spray
equipment and also the flow of the emulsion on the coated surface which can
lead to the same type of defects that occur with dip coating, except for a
'neck' blemish. Design of equipment for the support of the product during
spraying is required so that the uncoated region adjacent to suspension
hooks or grips is unimportant or can be subsequently treated appropriately.
Film formation involves the evaporation of water which requires more
heat than for the removal of an organic solvent. The rate of diffusion of
solvent or water through a polymer film is also important since it is essential
that all of the low molecular weight species are removed prior to curing.
With evaporation of water, a stage will be reached when the emulsion
inverts and the epoxy resin becomes the continuous phase. Such a phase
inversion will depend on the concentration of water present in the film and
the temperature as well as other factors. The phase inversion temperature
(PIT) is discussed by Shinoda and Friberg (1986). Evenness and film quality
may be improved by the use of an organic solvent to reduce the viscosity of
the resin prior to cure. The flow properties of the uncured film will depend
on the pigment, its type, concentration and particle size distribution as well
as the presence of other additives. The flexibility of the coating can be
improved by the addition of alkyd, polyester or acrylic resins, polymeric
amines and polyamides may be used as curing agents. A can coating has
been formulated from a higher molecular weight epoxy resin (M. wt -15-
20 000) which was initially dissolved in solvent and then dispersed in water
and cured with either a melamine or urea formaldehyde resin. The baked
films were 'flexible' and resistant to the pasteurization conditions required
for beer cans. Water-based epoxy resin coatings can be formulated for
electrodeposition, both anodic and cathodic (Brewer, 1985). These coatings
were developed for the applicance and automobile industries with cathodic
systems being replaced by the anodic coatings because higher corrosion
resistance can be produced. With electrodeposition (ED) the loss of paint
due to overspray is eliminated, so that the process offers advantages
compared with spraying since with ED it is possible to obtain more uniform
coatings even in recessed regions. Very many formulations have been
proposed for water-based coatings for dipping, spraying or electro-
deposition.
It is possible to build hydrophilicity into an epoxy resin and an example is
the reaction product of an epoxy resin with dibasic acids or anhydrides, and
many different combinations have been evaluated. The synthesis of
hydrophilic epoxy resins may be illustrated by the following reactions with
both epoxy and hydroxyl groups,
oII
C-OH
R'"
'C-OH
II
o
oII 0
II
OH
O-C-R'-C-OH
+ *-CH 2-tH-CH 2-*
- *-CH 2-CH-CH
I 2- *
The carboxyl acid groups are, when neutralized, hydrophilic and the higher
the acid content the more readily dispersible these resins are in water
containing amines or inorganic bases. The degree of hydrophilicity IS
controlled by the extent of neutralization of the pendant acid groups.
9.2.6 Powder coatings

Powder coatings with 100% solids content do not contain a solvent or water
carrier. They are blends of liquid and solid resins with curing agents plus
COATlNGSAND OTHER APPLICATIONS 317
other additives. Thus pollution, fire and explosive hazards are avoided in the
coating plant. Also thick coatings can be applied in a single operation, with
high application rates. The technology of powder coating is the subject of a
recent text (Misev, 1991). There are many varied formulations and several
different methods of application may be used. To formulate powder coating
systems, high molecular weight (usually over 1000) epoxy resins are
compounded with hardeners and other additives such as a flow-control
agent, pigments, fillers and catalysts. The exact combination depends on
requirements of the specific application. Fusion and dry blending are
commonly used techniques in producing appropriate powder formulations.
It should be noted that some dry-blending in a Z-blade mixer may require
heating the components to a relatively 'high' temperature, thus formulations
with reactive hardeners must be avoided. Latent hardeners may be used or
two part mixes, with the hardener and catalyst kept separately until the
coating is applied. The compounded resin is ground in a ball-mill or other
comminuting machine so that it is fine enough to pass a 45 mesh, with a
particle size less than 350 {lm. It may be finer for specific applications. The
hardeners that may be used include dicyandiamide, amine adducts,
polyesters and phenolic resins. Very fast cures of 30 seconds with tempera-
tures up to 250°C can be obtained with acid anhydrides catalysed with
stannic octoate. The anhydride and catalyst can be mixed into separate
portions of resin to avoid premature cure. Also, this type of system is
sensitive to water, which will be absorbed during storage unless suitable
precautions are taken. Water absorption reduces the hardener efficiency
which leads to poorly cured coatings.
Methods of application which are frequently used include fluidized-bed,
electrostatic spray, flock spray, and electrostatic fluidized-bed (Richart,
1982, 1985).
Powder coatings have excellent electrical resistance and thermal stability.
These are properties attractive to electrical and electronic industries as
coatings for vital components which are solvent sensitive and where contact
with water must be avoided. Coating underground pipes is another
important application due to the good resistance to corrosive soil and
cathodic disbonding of such coatings. It can also be applied to many general
objects as protective and decorative coatings. The composition of powder
coatings can be varied to suit the requirements of specific applications.
The adoption of powder coating may involve expensive investment
especially when changing from another type of coating line, colour matching
may also be difficult. However, powder coatings have many advantages over
other coating methods which include, little pollution or fire hazard
problems, readily automated process technology, wide range of coating
thickness possible in one application with exceptional film quality.
9.3 Industrial and related applications
9.3.1 Tooling
Epoxy resin tooling was initially used in the aircraft and car industries but
has now been extended to mould and tool making for many different end-
products, such as ship and boat building, domestic equipment and pottery.
A combination of techniques similar to those used for casting and laminating
are used for the production of moulds and foundary patterns. The resins are
usually formulated with a liquid resin cured with a room temperature
hardener. The properties are modified by the use of additives which not only
improve specific properties but also usually reduce materials' costs. Tools
can usually be fabricated in a fraction of the time required to machine metal
and the costs of epoxy tooling are much lower than conventional metal tools.
Epoxy resins are especially advantageous for the construction of prototypes
and tooling for short runs. Lee and Neville (1967) have a chapter on epoxy
resin tooling in which much detailed information is given which is still
relevant. Special resins for tooling applications can be formulated and
details are available from resin manufacturers. Some of the more important
tooling applications of epoxy resins include:
1. Moulds for various purposes such as form master moulds for vacuum
forming and injection moulding.
2. Jigs and fixtures to assist accurate positioning and checking shapes
and dimension of components for such operations as drilling, cutting,
and welding.
3. Metal-forming tools to replace the much more costly steel tools for
prototype and short run production. Examples of such tools are
stretch blocks, press tools, drop-hammer punches and dies. A rubbery
resilient facing on one of the tools of the set may be required, e.g. for
a drop hammer.
4. Foundry patterns ranging from master moulds from the original
pattern, duplicate patterns made from the master mould to core
boxes.
The benefits of using epoxy resins instead of traditional steel or aluminium
alloy as tooling materials are numerous. In addition to facile forming of
complex shapes the resins have low cure shrinkage and close tolerance can
be maintained. Also, the inert nature of epoxy resin to both heat, when
operational temperature is below its glass transition temperature, and
chemical attack, which may corrode steel, ensures shape stability. Other
important advantages include lower costs than for metals, lighter weight,
also the shape of the tool can be readily adjusted, and lead-times are shorter.
Epoxy resin formulations for tooling applications are normally room
temperature curable systems made up with liquid BADGE-type resins with
polyamines (normally aliphatic) as hardeners. Depending upon specific
application, diluents, plasticizers, fillers can also be introduced into the
system for the purpose of reducing cost and improving required physical and
chemical properties. In cases where special toughness is required, laminates
of epoxy resin with various fibres can also be used. Open mesh 'tooling
fabrics' are available and higher stiffness can be attained by the use of carbon
fibre reinforcement (chapter 8).
9.3.2 Civil engineering

In the field of civil engineering the consumption of conventional materials
such as concrete, wood, metal and glass is astronomical. Epoxy resins
exhibit excellent adhesion to these materials and therefore have been used
widely for surfacing, coatings and for repairs (Maslow, 1979). The most
frequent application of epoxy resins in civil engineering include:
1. Flooring
2. Road and bridge coatings
3. Concrete bonding and repair.
9.3.2.1 Flooring Epoxy flooring can be formulated to be 100% non-

volatile and 100% reactive so that one inch thick coatings may be applied in
a single application. Floors of industrial premises such as chemical plants
and warehouses are subject to constant and severe abuse of various forms.
Typical examples are chemical spillage, harsh detergent cleaning, and heavy
frictional wear by vehicles such as forklifts etc. Since floors of such industrial
buildings are usually constructed with materials such as concrete, which
under adverse service conditions are likely to gradually and then rapidly
disintegrate, epoxy resins have been found to be very useful in aiding the
maintenance of such floors, due to their chemical resistance, high
mechanical strength and facile application.
In formulating a flooring system, the epoxy resin is usually mixed with
hardeners, fillers and pigments according to service requirements.
Colouring materials can also be added to suit different purposes. Prior to
application of the resin the surface has to be thoroughly cleaned and pre-
treated with an epoxy or other primer. Depending on the viscosity of the
formulated system, it can be applied either by brush, roller, squeegee or
trowel. For high viscosity formulations such as those primarily for wear
resistance where large quantities of solid filler are added, the material has to
be screeded by trowelling. On the other hand, low viscosity formulations for
covering continuous floors with normal wear conditions can be applied with
a squeegee or plastic comb. Self-levelling epoxy flooring compositions are
formulated so that their viscosity is low. Epoxy flooring may be classified
into three types, seamless flooring, industrial or epoxy concrete and terrazzo
flooring. Seamless flooring has a 'low' viscosity due to the use of a reactive
diluent and can be applied in thickness of 10--60 mm. They can be used to
coat Portland cement floors, wood or metal after a specific surface
preparation before application of the resin. Industrial epoxy flooring or
epoxy concrete contains 75-85% particulate filler, usually sand (10--100
mesh). With high filler loadings the viscosity is high and the composition is
applied by trowelling with thickness of 3-25 mm CI8 to 1 inch). These
compositions have low shrinkage and may be used to repair cement floors.
Epoxy terrazzo floors contain coloured stone or marble chips and a fine
particulate filler, such as calcium carbonate, and are trowelled to give 6-
12 mm (I;cI;2 inch) thick coatings. After setting overnight they are wet-
ground to a polished terrazzo surface which is readily maintained.
The advantages of epoxy flooring may be summarized as follows:
1. Ease of seamless application, room temperature cure with low
shrinkage
2. Excellent adhesion to many substrates
3. Chemical resistant and durable
4. Tough, flexible and abrasion-resistant
5. Dust-free, skid-resistant and readily maintained.
9.3.2.2 Road and bridge coatings. Epoxy resin coatings have been used
to protect and repair road and bridge surfaces. Spalling is a commonly
encountered problem on the surfaces of roads and bridges. It is caused by
the combination and accumulation of many different factors such as freeze-
thaw cycles, spillage by vehicles, road gritting and de-icing operations.
Epoxy resins that are used for the treatment of these conditions are
usually liquids cured by amine hardeners at ambient temperatures. Coal tar
is used as extender and often in high proportions in their formulation. When
coated onto the structure and cured, the resin formulation forms a film
which prevents the spalling of concretes on the road surface and bridge
structure. Furthermore, the use of such formulations to seal cracks can
significantly slow down their growth by bonding the damaged sections into
an integral part. It also limits water penetration into the structure of the
bridge. Ingress of water to steel reinforcement or framework causes
corrosion. Hence epoxy resin coatings can give substantial protection
against such damage. Some other applications of epoxy resin road and
bridge coatings include:
1. Repair of scaling concrete roads with 'thin' coatings
2. Repair of 'slippery' but otherwise satisfactory asphalt or concrete
road surfaces
3. Fuel spillage protection
4. Lightweight surfacing for bridge decks
5. Waterproof membrane between concrete and an asphalt topping

6. Upgrading additive for bituminous surfacing compositions
7. Road marking and coloured surfaces.
9.3.2.3 Concrete bonding and repair. Use of epoxy resin formulations as

concrete bonding and repairing materials finds application over much wider
areas than just bridges and roads (Allen and Edwards, 1987). Virtually, any
structure with concrete as its primary construction material will develop
cracks with age. Epoxy formulations can be coated onto the surface to
prevent the initiation of such cracks and also to repair existing cracks.
Structural defects occurring in other materials such as wood, brick and metal
can also be repaired by epoxy formulations. Special formulations are
available from resin suppliers.
9.3.3 Moulding compounds

Consideration of the application of epoxy resin moulding powders requires
comparison with both alternative thermosetting resins and also with casting
which is an alternative method of processing liquid epoxy resins. Epoxy
resins can be moulded (Lee, 1988) by either compression or transfer
moulding and also liquid-injection moulding. Reaction injection moulding
also offers processing advantages (Becker, 1979; Manzione, 1989).
The superiority of epoxy resins compared with conventional phenol, urea
and melamine formaldehyde resins is that, during cure, volatile products are
not evolved and hence their moulding shrinkage is much lower. Also, they
can be formulated so that at the moulding temperature they have lower
viscosities so that much lower moulding pressures are required. For con-
ventional P/F, U/F or M/F resins the moulding pressure is 1000 to 5000 psi
(7-20 MPa) whilst for epoxy resin moulding powders the pressure can be
very low, always less than 1000 psi. Thus presses and auxilary equipment can
be simpler, and the moulding tools will also have to withstand lower stresses.
However, because of the low viscosity of the epoxy resin at moulding
temperatures there may be a problem due to excessive flash and 'tight'
fitting moulds are essential. Some of the advantages of epoxy moulding
powders compared with other thermosetting resins may be listed as follows.
1. Better dimensional tolerances and stability
2. Better electrical and mechanical properties
3. Higher chemical and heat resistance
4. Lower water absorption (depends on hardener and filler).
The disadvantages as well as higher cost include the need for closer mould
temperature control, mould release and ejection problems. Mould release
can be difficult because of the adhesive properties of the epoxy resin, which
of course is often an advantage since the epoxy resins bond well to metal
inserts. At the moulding temperature the cured product is flexible and
ejector pins can cause damage to the moulding. This can be avoided by
suitable design so that excessive local stresses are avoided during ejection of
the moulded part.
Compared with casting the moulding of epoxy resins offers the advantages
of simpler handling, higher filler loadings and also fibrous reinforcement. Its
disadvantages include higher capital costs and storage problems with the
uncured moulding powder. Their shelf-life may be limited due to incipient
cure if the storage temperature is not low, for example it may be only several
weeks at 22°C but much longer in sealed packages refrigerated at 7°e.
Epoxy moulding powders are formulated with a mixture of solid or liquid
BPA resins which may be blended with epoxy novolacs. The latter improves
the flow properties and has higher functionality and hence decreases gel and
cure time. The curing agents are often diaminodiphenol methane (DDM) or
acid anhydrides (PA or THPA), but others may be used for specific
purposes, for instance chlorine-containing curing agents have been
suggested for improving flame resistance. The moulding powders contain
suitable particulate fillers, such as silica flour, and can be reinforced with
fibres, 1.5-25 mm (116 to 1 inch) long. Zinc stearate is often included as a
release agent, to assist ejection of the cured product from the mould.
Thermal conductivity can be increased by use of metal powder as fillers, but
should be avoided when the electrical properties are important. The density
of the filled resins ranges from about 1.6 to 2.2. Lower densities, 0.75-1.00,
can be obtained by the incorporation of microballons, composed of either
clay, U/F or P/F resins or glass (Lee and Neville, 1967a). A major
application of moulding powders is the encapsulation of electrical
equipment and electronic devices. It is essential that during moulding the
resin flows so that the component is not damaged or displaced. Typical
moulded parts include deposited carbon resistors; chokes; ceramic,
polyester and glass capacitors; metallized thin-film resistors; pulse trans-
formers; toroid, solenoid and relay coils; printed circuits, solid-state circuits
and many more products (Lee, 1986, 1988). Epoxy resin moulding powders
also have applications in chemical plant, such as valves and filter plates
where their chemical resistance offers major advantages. Structural
components with good thermal and chemical resistance coupled with good
mechanical properties and light weight are also moulded from epoxy resins.
9.3.4 Embedding
Embedding is used to protect and/or decorate electrical or electronic
devices. Allied and related terms are casting, potting, impregnation and
encapsulation. Moulding has been discussed in the section on moulding
powders. The electrical properties required for electrical applications have
been dealt with in chapter 6. Protective functions of the embedding resin
include exclusion of moisture, oxygen, salt spray, solvent, other chemicals
and microorganisms. The encapsulation of a device also protects it from
mechanical shock and vibrations and provides suitable locations for
mechanical handling. Alternative materials to epoxy resins include
silicones, polyurethanes, polyesters, polysulphides and allyl resins (Lee,
1988). The epoxy resins have low viscosities and shrinkage since volatile
compounds are not formed during cure. By suitable formulation it is
possible to obtain a range of properties, from the rigid to flexible,
elongations at break ranging from 3 to 40%. BPA resins and also mixtures
with epoxy novo lacs are used for potting. Cycloaliphatic epoxy resins have
good dielectric properties and excellent weatherability, which is due to the
absence of chlorine and aromatic rings. The cycloaliphatic resins require
acid anhydride hardeners and longer high temperature cures. The pro-
perties of the resin can be modified by the inclusion of fillers and flexibilizers
(chapter 4). Typical particulate fillers include silica, 'talc, clay and calcium
carbonate, mica and chopped glass. Metal powders can be used to increase
the thermal conductivity of the product. Lower density products can be
produced by the inclusion of micro spheres of saran, PIF resin or glass.
9.3.5 Miscellaneous
It is almost an impossible task to list all the existing applications of epoxy
resins since there are currently many new developments. However, it is
worth mentioning several areas in which epoxy resins are used, although on
a small scale, but nevertheless for important specialized applications, such
as textile finishes (Tanaka and Shiozaki, 1988).
9.3.5.1 Oil field applications. Epoxy resins are used in drilling operations
to consolidate poor well formations, especially where the geological
structures are primarily sandy. Resin compositions are adsorbed onto the
well walls and subsequently cross-linked. Such treatment is particularly
effective in minimizing sand production, which could eventually result in the
well being discarded.
9.3.5.2 As reactant to form epoxy-based polymers. Epoxy resins are very

reactive polymers. Their reactions with other chemicals as well as polymers
can result in numerous new ranges of materials so far as these chemicals and
polymers have functionalities smaller than three. For example, reacting
epoxy resin with acrylic acids yields products, the epoxy acrylate resin,
which are curable by UV irradiation.
9.3.5.3 Epoxy resin foams. Blown foams can be formed with epoxy resin
formulations where a blowing agent is present. During the initial stages of
cure such an agent either evolves a gas or evaporates to leave voids in the
cured network. For example, hydrazide derivatives can be mixed into the
epoxy formulation; with the heat generated by cure or provided by heating
elements, they decompose to form nitrogen which is the blowing agent.
9.3.5.4 Epoxy resin stabilizers. Epoxy compounds also find application

as stabilizers and plasticizers for other polymeric materials. For instance
poly (vinyl chloride), PVC, readily dehydrochlorinates especially at
elevated temperatures. The dehydrochlorination reactions are complex and
appear to be autocatalytic (Hawkins, 1989). Thus, for PVC to be processed
it is essential to add stabilizer which 'sequesters' the hydrochloric acid that is
formed. The epoxy group reacts readily with the HCI as it is formed and also
there is a synergistic effect when metal salts are present as well (Stephenson,
1989). Not only are BPA type epoxy resins effective stabilizers for PVC but
also- epoxidized oils, esters and fatty acids will both function as stabilizers
and plasticizers.
References
Allen, R.T.L. and Edwards, S.C. (eds) (1987) The Repair of Concrete Structures, Blackie,
Glasgow, UK.
Ash, M. and Ash, I. (1980-1983) Encyclopedia of Surfactants, Vol. 1-3. Chemical Publishing
Co., New York.
Barton, A.F.M. (1983a) CRC Handbook of Solubility Parameters and Other Cohesion
Parameters, Ch. 1,2. CRC Press, Boca Raton, USA, pp. 1-21. (1983b) Many references to
epoxy resins, see Index.
Becker, W.E. (1979) Reaction Injection MOUlding Van Nostrand Reinhold Co., New York.
Bonneau, M. and Burkhart, A (1991) MEPPE Focus (Jan.), pp. 1-6.
Brewer, G.E.F. (1985) Ency. Polym. Sci. Eng. (Suppl.) Wiley-Interscience, New York,
pp.675-fJ87.
Burkhart, A (1988) Hybrid Circuit Technology, pp. 222-228.
Chen, X.M., Ellis, B. and Seddon, A.B. (1992) Paper prepared for publication but submission
delayed until after patent application.
Ciba-Geigy (1988) Araldite Resins and Hardeners for Surface Coating Systems, Ciba-Geigy
Plastics, Cambridge, UK.
Coeling, K.J. (1985) Ency. Polym. Sci. Eng., Vol. 3, Wiley-Interscience, New York, pp. 567-
575.
Ellis, B., Chen, X.M. and Seddon, AB. (1991) British Patent Application.
Emerson, J.A., Sparapany, J.J., Martin, A.R., Bonneau, M.R. and Burkhart, A (1990)
IEEE, 40th Components and Technology Conference Vol. 1, pp. 6()()....605.
Grulke, E.A (1989) Polymer Handbook, VII 3rd edn ed. Brandrup, J. and Immergat, E.H.
Wiley, New York, pp. 519-559.
Hawkins, W.C. (1989) Enc. Polym. Sci Eng. 2ndedn Vol. 15, Wiley, New York, pp. 560-563.
Lee, H. and Neville, K. (1967) Handbook of Epoxy Resins, McGraw-Hili, New York, pp.I-16.
(1967b) Ch. 23, pp. 1-5, Ch. 24, 7-40.
Lee, S.M. (1986) Ency. Polym. Sci. and Eng. Wiley-Interscience, New York, Vol. 5., pp. 792-
828, see p. 818.
Lee, S.M. (1988) Epoxy Resins, Chemistry and Technology 2nd edn, ed. May, C.A, pp. 799-
810.
Manzione, L.T. (1989) Enc. Polym. Sci. Eng. 2nd edn. Wiley-Interscience, New York,
pp.72-100.
Martin, S.R.W. (1970) Paint Technology Manuals, Gen. ed. Taylor, C.I.A. and Marks, S.,
Part 3, collator Keenan, H.W., Chapman and Hall, London, p. 16.
Maslow, P. (1979) In Plastic Mortars, Sealants and Caulking Compounds. ed, Seymour, R.B.
ACS Symposium Series 113. American Chemical Society, Washington, D.C. pp. 39-60.
Misev, T.A. (1991) Powder Coatings, Chemistry and Technology. Wiley, Chichester, UK.
(1991a) pp. 304-324.
Nielsen, L.E. (1977) Polymer Rheology, Ch. 9. Marcel Dekker, New York, pp. 133-150.
Paul, S.l. (1986) Surface Coatings, Science and Technology, Wiley, Chichester, UK.
Potter, W.G. (1970) Epoxide Resins, Ch. 8, Iliffe Books, London, pp. 144-167.
Richart, D.S. (1982) Kirk-Othmer Enc. Chern. Tech., 3rd edn, Vol. 19. Wiley-Interscience,
New York, pp. 1-27.
Richart, D.S. (1985) Ency. Polym. Sci. Eng. Vol. 3, Wiley-Interscience, New York, pp. 575-
601.
Scharenberg, R.T. (1985) Ency. Polym. Sci. and Eng., Vol. 3, Wiley-Interscience, New York,
p.552.
Shell (1991) Technical Bulletins, Shell Chern. Co., Houston, Texas 77027, USA.
Shinoda, K. and Friberg, S. (1986) Emulsions and Solubilization, Wiley, New York.
Somerville, G.R. and Smith, I.T. (1973) Epoxy Resins, Chemistry and Technology, 1st edn,
Ch. 7, eds May, C.A. and Tanaka, Y., Marcel Dekker, New York, pp. 451-484.
Stephenson, R.C. (1989) Enc. Polym. Sci. Eng. 2nd edn (Suppl.) Wiley-Interscience, New
York, pp. 867-870.
Tanaka, Y. (1988) In Epoxy Resins, Chemistry and Technology, 2nd edn, Ch. 2, Table 49,
pp. 212-231.
Tanaka, Y. and Shorzaki, H. (1988) In Epoxy Resins, Chemistry and Technology, 2nd edn,
Ch. 12. ed. May, C.A. Marcel Dekker, New York, pp. 956-1048.
Tess, R.W. (1988a) Epoxy Resins, Chemistry and Technology, 2nd edn Ch. 8. ed. May, C.A.
Marcel Dekker, New York, Ch. 8, pp. 719-782. (1988b) see pp. 726-735.
Utracki, (1988) In Rheological Measurement eds Collyer, A.A. and Clegg, D.W. Elsevier,
Amsterdam.
General index
For reference to specific curing agents and hardeners see the separate index
abnormal addition to epoxy branching tree 5, 76 water-based 314-316

rings 21 Ii-relaxation cohesive energy density
abrasion resistance 56,319 electrical 192,197-199,202 104-106,311
acid anhydrides, cure kinetics mechanical 110 Cole-Cole plots 184
98-99 bridge coatings 319-320 colour 34,41,48,52-53,56
acrolein 28 brominated phenols 19 complex electrical modulus
adhesion, theories of building industry 319 182
adsorption 209 bulk modulus 104, 106 complex shear modulus 5,80,
diffusion 208-209 bulk moulding (BMC) 267 87,88
electrostatic 208 composites 102, 144,216
mechanical interlocking cable jointing 41 epoxy resin matrix 270-273
208,262 can coatings 39 fabrication 267-269
test methods 218-221 carbon fibre composites 240 properties 276-284
adhesive bond-fibre/matrix carbon fibres 256-263,282 concrete repair 253, 319-320
262,264-267,279,287 carbonation, of curing agents cone and plate viscometer 87
adhesive bond thickness 39-40,43 contact angle 44, 53, 68,
234-235,237 carboxyl-terminated butadiene- 270-273
adhesives 34,41-42,44-45, acrylonitrile (CTBN) 129, coupling agents 243-248,
47-48,49,51,59-60,65-68, 134,136,168,169,228,230, 265-266,311
69-70,102,126,206,304,307 231,234,272 crack blunting 138
ageing 177,201-202 castings 33,48,52,58,63,69, crack branching 162
chemical degradation 6, 27 203, 304, 307 crack growth-adhesive bonds
physical ageing 110 catalysis of cure, kinetics 96 229
alkyds 1, 309 catalytic curing agents 272, crack opening displacement
alloys--epoxy 126 275 150
alumina trihydrate 125 chain extension process 21, 24 crack pinning 125, 165
aluminium powder 203,228 chemical analysis 29-32,74, cross-ply laminates 284
amine hardeners 226, 228 84 crack propagation 152-156
kinetics of cure 90--98 chemical resistance 25,40-41, crack tip blunting 151, 154,
amino resins 304, 309 44,51-53,56,63,65,118,130, 157,158
analysis, chemical 29-30 304,320 crack tip opening displacement
aniline 26 chemical surface pretreatments (CTOD) 153,155,160
annealing 105 213,215,265,310-311 crack tip plastic zone 236-237
anodising 248 chemisorption 209, 262 crack tip radius 155
aqueous dispersions 314-316 chlorine content 31 crack velocity 154, 161, 164
aramid fibres 256-257, 259, assay 31-32 cracking-cross-ply laminates
263-264 organically bound 32 286-287
ash content 27 chromatographic methods 32, craze 136-137,150,170
autocatalytic cure reactions 83,85,87 critical fibre length 267
90,95,97 civil engineering applications cross-links 4-5,78,120,124,
52,70,319 145,159,177,200,275
ball and ring softening point coatings 1-2,44,46,48-49, cure
33 52-53, 60, 70, 102, 144, kinetic analysis 90--98
bismaleimide resins 126, 272, 308-317 wet chemical analysis 83-84
295 can 39 parameter, C 181, 188, 192,
bisphenol A 16,19,21,23 dip 313 194,198
bisphenol F 19,24-25 drum 66-67 cyan urate resins 272
blocking agents 141-142 fast bake 67 cycloaliphatic epoxy resins
bloom 39,43 high solids 41,44,53 27
bond thickness 234-235, 237 maintenance 56
bound chlorine 21 marine 56 d.c. conductivity 177-179,
branching coefficient 75, 77 powder 42,50, 61,304, 191,228
branching theory 75-81, 102, 316-317 dehydrohalogenation 15,22,
109 solvent resistant 43 26
328 GENERAL INDEX
diaminodiphenylmethane ring strain 9-10 permeation chromatography

(DDM) 98, 184, 207 structure 9, 32 32,85
dibutylphthalate 118 synthesis 7-8 point 72,73,75,77,78,80,
dielectric ethylenechlorohydrin 7 87,106,182-183
measurements 175--176 ethylene oxide 7, 14 time 276
permittivity 178--179 exotherm 73, 122, 128,203, glass fibres 45, 256-258,
properties 94, 177-182,323 227,269 265-266
relaxation 178, 184, 188, extenders 123 glass microspheres 125, 165,
191-192 extension ratio 160 203
Diels-Alder addition 27 maximum 152 glass transition temperature,
differential scanning Tg
calorimetry (DSC) 81,86, fatigue-static 158 cure parameter, C 99,
90,93,98,102 fibres 102-104, 108--111
diffusion resin adhesion 264 definition 81-82
water 291-292 strength distribution 259, determination 33,81-82
solvents from coatings 315 276 DiMarzio theory. 202
diffusion control--<:ure fibrous fillers 121, 124,216, effect of
reactions 6, 92-93, 99-102 256-259 cure 6,72,99,108--111
diluents 117-120,225,228, filament winding 45,52,59, rubber modification
304,319 63,267-268,304,307 169-170
dipolar polarization 178--190 fillers 117-126,225,227,319 water sorption 293-296
dough moulding (DMC) 267 advantages 122 effect on
drum coatings 66--67 applications 123 d.c. conductivity 190-191
Durran's softening point 33 effect on viscosity 312 fracture energy release
dynamic mechanical thermal particle size distribution rate, G 156-157
analysis (DMTA or DMA) 89 124--125,135 rate of cure 72
particulate 121,125 toughness and flexibility
E-glass fibres 256-257 flame retardants 31,141,203, 125-130,144--145,159,
elastic moduli 105, 165 304 228,240,272
laminates 276-277, 280-281 flaw size 229 high Tg and elastic modulus
electrical applications 52,63, flexibilisers 63,65,128,225, 139-140,252,288
65,69,175,203-204,304,307 228 physical ageing 110-113,
electrical properties 31,74, flooring 45,47,53,56,304, 200
87,124,130 307, 319-320 relationships
electrically conducting resins fluorinated epoxy resin 19, TTT diagram 6
228 127-128,249-250 TUF (VFT) equation 83,
electrophilic attack on epoxy fluorinated phenols 19 175-176
rings 1,14 fluoroanhydride curing agent WLF equation 83
embedding 322-323 127 unaffected by fillers 122
energy release rate, G fractography 161-164,279 glycidylamines 270
146-149,229-230 fracture behaviour 145, 229 o-glycidyloxypropyltri-
emulsions--epoxy coating fracture energy 144,229 methoxysilane 140,244, 246
314--316 fracture energy release rate, G glycol 130
encapsulation 47,63,69,203, 146-149,158,169,171,229 a-glycol units 30-31
304,322 fracture mechanics 145-150 graph theory 79
environmental testing 222, adhesive bonds 218, Griffiths fracture criteria 145,
238--252,287,291-298 221-222,234 229,259
epichlorohydrin 1,3,7,15, fracture test geometry
16-18,21,22,24,25,26,32, double torsion (DT) 148,
130,263 152,154,156-157,161 halogen 127
epoxidation 9,15,26-27 mode II, adhesive joint halohydrin 7-8, 15
epoxy content- 224,237 heat distortion temperature
determination 29 single edge notch (SEN) 130,203,275
epoxy resins 147-148, 161 heat resistance 25,51-52,60,
chemical analysis 29-32, fracture toughness, K, 137, 63,69,296-298
84--86 140,146,149-150,151,171 Hoeppler viscometer 34
commercially available free volume 110 hydrogen bonding 106
304--305 functionality 4,75-77 hydrolysis of epoxy groups 21
composite matrix 270-275 fusion process 23-24 hydrolysable chlorine 22,27
identification 29 hydrophobicity 127-128,
manufacture 14--29 gelation 5,24,72,74,75,77, 249-251
epoxy-phenolic resins 25,226, 78,83,88, 106, 110, 135, 179, hydroxyl groups 43,90-92,
234 190 95,98,201-220,272
epoxy rings gel determination 30-31
reactivity 11-14 fraction 5,74,177 hygroscopy 39, 43-44
GENERAL INDEX 329
impact strength 48, 129, 130, particulate fillers 121, 125 laminate thermal strains
159 peel strength 129, 220 287-288
infrared spectroscopy 85 peracetic acid 7, 26-28 side reactions 21, 24
injection moulding 267, 321 peracids 7-9,15,16,26-27,28 silane coupling agents 140,
inorganic ions 22 perbenzoic acid 9 265-266, 311
ionic conduction 176, 184, percolation theory 80--81 silica 165, 172,203,228
188-l9l phenol(s) 16,19,24 size exclusion chromatography
ionic content 27 phenol-formaldehyde resins 97
ionic impurities 189 (PIF) 25, 126, 226, 304 softening point 33,61
phenoxy resins 24 sol fraction 5-6, 74, 177
Kevlar fibres 259, 263-264 plasticiser 128 sol-gel transition 79-80,176,
Kohlrausch-Williams- Poisson's ratio 105, 112 182,189-191,194
Watts equation (KWW) 101, strains 286, 289 solubility parameter 105,
180,194 polyamides 129 311-312
polyetherimide 139,140,166 solution viscosity 33-34
polyethersulphone 139, 167, solvent resistance 43,63
laminates 4,33,42,44-45,
228,270 solvents 141,304,311
51-53,55,59-60,64,276-289,
polyglycol diepoxides 130 specific moduli, fibres
307,319
polyhydric phenols 19 256-257
delaminating 289
polyphenylene oxide 167 specific properties 256-258
laminating 267-268, 304
polysulphide 130,203,226 spectroscopy 32, 85-86
lap-shear test 219,233
post cure 6, 101, 275 spot tests for epoxy resins 29
Lewis acids, resin synthesis 26
pot life 42,56,59,73 static fatigue 158
potting 47,63,69,203 stick-slip fracture 154
market-epoxy resins 1-2, prepregs 267, 304 strength, laminate 277-278,
4,146,303 primers 67,140,217,240, 281-284
mechanical properties 248,304 stress corrosion cracking 276
laminates 276-284 carbon fibre size 262 stress intensity factor, K.
resins 144,159,305 glass fibre size 265-266 149-150
mechanical relaxation 176 process zone---<:rack 166, 170 stress-strain curves 106-107
metal surface 213-217 propylenediamine 191 structural adhesives 124, 136
methyl epichlorohydrin 26 pultrusion 267-268 sub-Tg transitions 110,
mixing errors 226 192-194,198-200,202
modifiers 126 styrene 118
modulus qualitative identification-
surface coatings see also
laminates 276-277, epoxy resins 29
coatings 1, 4, 26
280--281, 283 quantitative analysis-epoxy
surface free energy 211-212
maps 102 resins 30
surface tension 211
resin 122, 125,229 surface treatments 213, 218,
specific 256-258 rate equation5--(;ure 90--98, 225,228
molecular diffusion 176-178, 195 anodising 242,248
181 reaction injection moulding aramid fibres 264, 282
molecular weight 4, 32, 78 (RIM) 275,321 carbon fibres 261-263,282
molecular weight distribution reactive diluents 118-120 chemical pretreatment 215,
(MWD) 21,32-33 reactivity of epoxy groups 228,248
morphology 72,73,74, 11-14 glass fibres 265, 282
132-136, 160--161 reinforced reaction injection mechanical abrasion 214,
moisture see water absorption moulding (RRIM) 267 248
moulding powders 304, 307, reinforcing materials 256-259 solvent cleaning 214
321-322 tJ-relaxation 110, 192,
moulds 318 197-199
relaxation functions 188 Taffy process 22
network 4-6 reptation theory 100 Tammann-Vogel-Fulcher
network formation 29,74--81 resorcinol 19 equation (TVF or VFT) 83,
nodular structure 160--161 rheology 87-89 175,186
Nomex fibres 259, 263-264 road surface coatings 319-320 tan d (viscoelastic) 111
novolac (PIF) 25 rubber modification 136, 157, tensile strength 277-280,
nuclear magnetic resonance 167-169,229,233,235, 281-283
32-33,85,101 237-238 terazzo flooring 319-320
nucleophilic attack on epoxy tetraglycidyldiamino-
group 11,14 scaling theory 79-80 diphenylmethane (TGDDM)
nylon 226, 232, 240 shear yielding 137-139,151, 80,128
167,171 thermal analysis 86
organosilanes 140--141,218, shrinkage 124,128, 141, 144, thermal conductivity 122,
243-248,265-266.311 227 123,322
330 GENERAL INDEX
thermal expansion 122, 123, viscometers 34, 87 water absorption

227,288 viscosity effect on
thermal spiking 297-298 effect of glass transition
thermal strain~ross-ply cure 72-73, 8~1, temperature, Tg 234,
laminates 287-288, 295-298 87-88,97,175-176 270-272
thixotropy 124, 227 diluents 117-120 properties of
time-temperature transition fillers 124,312 adhesives 141-142,232,
(TIT) diagram 6,81 temperature 83,135, 238-239,242-251
time-temperature super- 175-176,186 composites 264,290-298
position 158 effect on unstable crack
Tg-TP diagram (CTP) d.c. conductivity 190 propagation 154
103-104, 110,275 diffusion controlled cure effect of blocking agents
toluene 118, 141 99-100 141-142
tooling 42,52,318-319 fibre impregnation fillers 122
toughening agents 225,228, 269-270 water spotting 39
272 relationships wetting 210-213,217
toughness 125-126, 144, 156, TUF (VFT) equation 83, fibre 282
165-169,172,228,230,234 175-176,186 Williams, Landel and Ferry
torsional braid analysis (TBA) WLF equation 83 equation (WLF) 83, 110
89,176-177 resin solutions 33, 141
Twaron fibres 259, 263-264 uncured resins 33-34 xylene 118, 141
vitrification 72,80, 88,
unsaturated compounds, 176-177 yield 108, 154, 156
oxidation 7-9 Vogel-Fulcher-Tamman Young's modulus
equation (VFT) 83,175,186 composites 275-277,
vinyl cyclohexene 27 volatile 43 280-281
viscoelastic properties 8~1, volume relaxation 110-112 resins 104--107, 145, 229
87-88, 104, 109-111, 158, 176
Index of curing agents and hardeners
accelerators 38-39,62,269 4,4' -diamino-diphenyl- hydrazides 60

acid anhydrides 60--65 methane (DDM) 54-56,
acrylonitrile adducts 44-45 73-74,82,98,107,109, imidazoles 50,56-59,62,94
alicyclic aliphatic amines 47 185-186,200--202,227,231, imidazolines 42
aliphatic polyamides 40-42 251-252,307,322 isocyanates 60
aliphatic polyamines 38-53 dicyandiamide (DlCY) isophorone diamine (IPDA)
amines 13,37-60,78-79, 49-50,59-60,269,273-275, 52-53
90--98, 134, 154, 158,226, 295,307
236,307,316 diethanolamine 98 jeffamine 307
N-aminoethylpiperazine diethylaminopropyl amine
(AEP) 47-48 (DEAPA) 45,203
amino-formaldehyde resins diethylene triamine (DETA) ketimines 44-45
37,66--fJ7 150,161,203-204,226
anhydrides 37,60--65,78, diethyltoluenene diamines latent hardeners 267,269, 317
98-99, 134, 140, 164, 272, (DETDA) 55 Lewis acids 63
307,316,322 dimethylaminomethylphenol Lewis bases 62
araliphatic amines 47 56-57,203
aromatic amines 54-56 dimethylaminopropylamine
Mannich bases 43-44,47,53
(DMAPA) 45
3,3' -dimethylmethylene- melamine 50--51
benzophenone tetra-carboxylic
melamine formaldehyde resins
acid di-anhydride (BTDA) di(cyclohexylamine) 51-52
6~7
64-65 dodecyl succinic anhydride
1,8-menthanediamine 52
benzyl dimethylamine (DDSA) 64
diaminodiphenyl sulphone metaxylene diamine (MXDA)
(BDMA) 57-58
bemzyl dimethylaniline 272 (DDS) 54-56,80,98,101, 47
141,159-160, 169, 199, 272, methylene-di(cyclo-
benzyl trimethyl ammonium
hexylamine) 51-52
chloride (BTAC) 58 307
diazabicycloundecene (DBU) methyl pentamethylene
1-benzyl-2-methylimidazole
diamine (MPMD) 45
58-59 57-58
meta-phenylenediamine
betaines 62 diethylene glycolbispropyl-
1,3-bisaminomethyl- (MPD) 54
amine 226
cyclohexane (l,3-BAC) mercaptans 37, 67~8
48-49 ethylamines and derivatives
biguanides 49-51 nadic methyl anhydride (NMA)
38-40
bismaleimides 126,270,272, ethylenediamine (EDA) 38, 64,272
295 187-188,191,196-197 norborane diamine (NBDA)
boron trifluoride (BF3) ethylene oxide-amine 48-49
68-69,272-273,275-276,307 adducts 42-45
boron trihalides 68-69 bisparaaminocyclohexyl
fluoroanhydride curing agent methane (PACM) 51-53,94
carboxylic acid anhydrides 127 pentaethylenehexamine
61~5 (PEHA [4067-16-7]) 38
catalysts 38-39,49,56-60, glycidylether adducts 42-43 phenol-formaldehyde resins
62,272,275 guanamines 50--51 37,65~6
cationic curing salts 69-70 phosphonium salts 69
chlorendic anhydride (HET) hexachloro endomethylene phthalate anhydride (PA)
64,203 tetra-hydrophthalic anhydride 1,63,99,203,322
cycloaliphatic polyamines (HET) (chlorendic anhydride) piperidine 136,155,168-169,
51-53 64,203 171, 230--232, 235
cyclohexane diamine 51-52 hexahydrophthalic anhydride polyamides 40-42,129,307,
cyclohexylaminopropylamine (HHPA) 63,99 316
45 hexamethylene diamine polyamidoamines 40-42
(HMD) 45, 183-185, 191 polyaminoimidazolines 41-42
1,2-diaminocyclohexane hydantoin 151 polybasic carboxylic acid 37
(1,2-DACH) 51-52 hydrazine 60 polyfunctional amines 37
332 INDEX OF CURING AGENTS AND HARDENERS
pyromellitic di-anhydride sulphonium salts 69-70 tetrahydrophthalic anhydride

(PMDA) 64 (THPA) 63,322
polysulphides 37, 130, 203 tertiaryamines 41,49,56-58, tri-glycidylisocyanurate
polyetheramines 46-47 62,99,238,273 (TGIC) 61
polymeric amines 47 tetra-alkyl DDM 55 tri-mellitic anhydride (TMA)
polyoxypropylene derivatives tetra-alkylated di(4-amino- 64,307
46-47 phenyl)-di-isopropyl trimethylhexane diamine
propylenamines 45, 188-189, benzene 55 (TMD) 45-46
191,194 tetraethylenepentamine tris(dimethylaminomethyl)
propylene oxide adducts (TEPA) 38 phenol 56-57,203
42-43 tricyclodecane diamine tetrahydrophthalic anhydride
pyromellitic dianhydride (TeD diamine) 48-49 (THPA) 63,99
(PMDA) 64 triethylenetetramine (TETA) thiols 67--fJ8
38,98,106,108,153,203,226,
siliconeamine curing agent 307 urea catalysts 49, 56-60
127 trimethylene glycol di-p-amino- urea-formaldehyde resins
spiro-acetal diamines 48 benzoate (TMAB) 86,108 66--fJ7

Chemistry and Technology of Epoxy Resins PDF

Uploaded by

Copyright:

Available Formats

Chemistry and Technology of Epoxy Resins PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chemistry and Technology of Epoxy Resins PDF

Uploaded by

Copyright:

Available Formats

Chemistry and Technology

Typeset in 1O/12pt Times by EJS Chemical Composition, Bath

ISBN 978-94-010-5302-0 ISBN 978-94-011-2932-9 (eBook)

As editor, I would like to thank all my co-authors for their contributions

1 Introduction to the chemistry, synthesis, manufacture and

2 Curing agents for epoxy resins 37

3 The kinetics of cure and network formation 72

3.2.1 Branching theory 75

4 Additives and modifiers for epoxy resins 117

5 Fracture behaviour of epoxy resins 144

6 Electrical properties of epoxy resins 175

7 Epoxy resin adhesives 206

7.7.3 Failure mechanisms 240

8 Composite materials 256

9 Coatings and other applications of epoxy resins 303

9.3 Industrial and related applications 318

Dr W.R. Ashcroft Anchor Chemical (UK) Ltd, Clayton Lane,

1.1 Epoxy resins

Resin 1000 tonnest Approximate' Approximate 1000 tonnest Approximate

Epoxy 211 3.8 802 195 741

/0, ---rc5'-¥~o /0,

Diglycidyl ether of bisphenol A

Table 1.2 Number average molecular weight, Kin, of 'Ideal' epoxy

n Mn No. of hydroxyl Epoxy equivalent

complex shear modulus. G*

weight increases slowly

Figure 1.1 The cure of epoxy resins.

for gelation, the process is continuous. Initially there is reaction between

Log te (cure time, tel

epoxy resins it is necessary to select a suitable hardener (see chapter 2) and

1.2 The chemistry of the epoxy group

The original discovery ofthe parent compound ethylene oxide, or oxirane, is

MOH could be sodium or potassium hydroxide and has to be used in

Table 1.4 Outline of the synthetic methods for epoxides

a. Direct oxidation 02/catalyst, Ag/lOO-500°C

3. From a-halocarbonyl compounds

b. Reduction: lithium aluminium hydride

c. Addition of alkoxide ion, followed by ring closure

d. Addition of cyanide ion

Somewhat alternative mechanisms have been proposed which Gritter

where Ar is aryl group.

Table 1.5 Some typical reactions of epoxy groups

1. Addition reactions by nucleophilic substitution

p" <j>H <j>RI

Rl = alkyl, alcohol; Rl = aryl, phenol

iii. Ammonia and amines

Products depend on the reducing agent

a secondary alcohol or primary alcohol or a mixture. When HR' is an amine,

In these reactions the attacking group donates a pair of un shared electrons

The kinetics of these reactions is discussed in detail by Frost and Pearson

1.3 The synthesis and manufacture of epoxy resins

1.3.1 Synthesis of epoxy compounds

(i) dehydrohalogenation of halohydrins and (ii) the epoxidation of alkenes

with the formation of a chlorohydrin. The epoxy group is formed by

The allyl chloride is converted to dichlorohydrin by reaction with

Industrially epichlorohydrin is either converted to glycerol by reaction with

The acrolein is chlorinated to yield 2,3 dichloropropionaldehyde which is

Epichlorohydrin may also be produced from allyl chloride by epoxidation

1.3.2 Epoxy resins manufactured from epichlorohydrin