Characterization of Novel Co-

Anhydride cured Epoxy Resins

by

Jens Rocks

A dissertation submitted in partial fulfillment of the

requirements for the degree of Doctor of Philosophy in the
Queensland University of Technology (QUT)

2004

Supervisors:
Professor Graeme G. George (Principal)
Associate Professor Peter Fredericks
Dr. sc. nat. Felix Greuter
 Abstract i

Abstract

Epoxy resins are widely used as coatings, encapsulations, structural composites,

castings, and adhesives in a number of electrical applications. Recently, novel
uncatalyzed co-anhydride cured epoxy formulations that exhibit a high performance
property profile, have been introduced. The objective of this thesis was to perform a
comprehensive material characterization of these new resin/hardener combinations,
which are potentially used as electrical insulation material in medium and high
voltage engineering.
The thermal cure behaviour of commercial tetraglycidyl-diamino-diphenylmethane
(TGDDM) and a co-anhydride mixture consisting of maleic anhydride (MA) and
hexahydrophthalic anhydride (HHPA) was extensively studied. Different analytical
real-time methods, such as FT-Raman spectroscopy, differential scanning
calorimetry, and chemo-rheological methods were applied to investigate the principal
polymerization mechanism and the related curing kinetics. It was demonstrated that
kinetic parameters that were based on isothermal measurements provided consistent
and reliable results. On the other hand the general limitations of different dynamic
methods for kinetic parameter calculations were outlined and discussed.
Temperature modulated differential scanning calorimetry provided a powerful
technique to confirm a TGDDM/MA and TGDDM/HHPA sub-network structure of the
co-anhydride cured epoxy. A generalized time-temperature-transformation-diagram
was developed in order to predict the complex material transformations (e.g. gelation
and vitrification) occurring during the entire isothermal curing process.
In the selected system, the mechanical deformation and fracture behaviour as a
function of temperature, strain rate, inorganic filler fraction, particle size, and
filler/matrix-adhesion were thoroughly studied by using compression-, tension- and
double torsion fracture-tests. The potential of hyperbranched polymers (HBPs) as low
viscosity toughening modifiers for highly crosslinked anhydride-cured epoxy networks
was experimentally evaluated. The effects of the HBP molecular structure, in
particular the specific shell chemistry, on thermo-mechanical properties, final
morphology, and blend concentration, were assessed. For the neat investigated
epoxy-system the most efficient toughening modifier was obtained for a molecular
HBP-design that provided a pseudo-homogeneous blend morphology. Thus, by using
suitable HBPs in a concentration of 20% w/w, the fracture toughness, expressed by
the critical stress intensity factor (KIc) of 0.58 MPam0.5, was increased by over 50% to
0.88 MPam0.5. The corresponding Young’s modulus and glass transition temperature
were only affected to a limited extent by the addition of the HBP-additive. The
ii Abstract

toughest epoxy blend (critical stress intensity factor of about 1.6 MPam0.5) was
achieved by the incorporation of 60% w/w inorganic silica particles. The application of
hybrid concepts by utilizing synergistic toughening mechanisms (HBP and silica),
revealed only moderate benefits within the investigated highly crosslinked materials.
As the examined epoxy/anhydride formulations are generally considered for high
temperature applications, it was essential to determine their long-term thermo-
oxidative ageing performance. The long-term thermo-oxidative ageing behaviour has
been investigated by means of thermo-gravimetric analysis (TGA), dynamic
mechanical analysis (DMA) and vibrational infrared spectroscopy (FT-IR) methods
with special emphasis on fundamental understanding of the ageing mechanism.
Effects of the thermal ageing on the characteristic viscoelastic and flexural behaviour,
weight-loss and oxidation susceptibility of the examined epoxy networks were
assessed and discussed, thus providing an understanding of the principal material
endurance properties.
The thermo-mechanical behaviour and the related structural changes with thermal
ageing were examined by Cole-Cole plots in combination with a molecular theory
previously developed by Perez. It was demonstrated that this new methodology
provides a connection between conversion, glass transition temperature, and
mechanical relaxation data and allows a fundamental molecular interpretation with
respect to physical and chemical ageing phenomena.
A variety of thermo-gravimetric experiments were carried out in order to determine
and model the specific weight-loss profile as a function of anhydride nature and
ageing temperature. The influence of different inorganic fillers on the thermo-
oxidative response was systematically studied. Different models were applied to
extract meaningful kinetic parameters in order to describe the thermo-oxidative
degradation and to facilitate an extrapolation of weight-loss data outside the
experimental time- and temperature-scale.
FT-IR micro ATR-spectroscopy was used to identify, localize and quantify the
complex oxidation behaviour. A thermo-oxidative degradation mechanism, that
involved predominantly radical oxidation processes and C-N as well as C-O chain
scissions, was proposed to account for the experimental observations. Specific
oxidation-front profiles were constructed to describe the heterogeneous oxidation
processes.
Finally, the comprehensive material characterization of these novel co-anhydride-
cured amino-glycidyl resins in terms of curing mechanism and kinetic, deformation
and fracture behaviour, and thermo-oxidative ageing performance, allows the
Abstract iii

assessment of the potential of these materials for demanding applications in

electrical and electronic industries.
iv Contents

Contents

1 Introduction & scope of work 1

1.1 Classification and history of epoxy thermosets....................................................1

1.2 Scope and survey of this thesis...........................................................................5
1.3 References..........................................................................................................6

2 Epoxy-anhydride crosslinking reaction 8

2.1 Introduction .........................................................................................................8

2.1.1 The uncatalyzed epoxy/anhydride curing reaction ...........................................9
2.1.2 The base catalyzed epoxy/anhydride curing reaction.......................................10
2.2 Experimental .......................................................................................................13
2.2.1 Materials ..........................................................................................................13
2.2.2 Sample preparation..........................................................................................15
2.2.3 Wide angle x-ray diffraction (WAXD)................................................................15
2.2.4 Differential scanning calorimetry (DSC) ...........................................................15
2.2.5 Gas chromatograhy (GC) .................................................................................15
2.2.6 Temperature modulated differential scanning calorimetry (TMDSC) ................16
2.2.7 Chemorheological analysis ..............................................................................16
2.2.8 Fourier-Transform Raman Spectroscopy (FT-Raman) .....................................17
2.3 Results & discussion ...........................................................................................17
2.3.1 Phase behaviour of the binary MA-HHPA mixture............................................17
2.3.2 Curing reaction studied by differential scanning calorimetry (DSC) ..................19
2.3.2.1 Non isothermal DSC ....................................................................................19
2.3.2.2 Isothermal DSC ...........................................................................................26
2.3.2.3 Temperature modulated differential scanning calorimetry (TMDSC)............32
2.3.2.4 Relationship between glass transition temperature and conversion.............34
2.3.3 Curing reaction studied by chemo-rheology .....................................................35
2.3.4 Curing reaction studied by FT-Raman spectroscopy........................................40
2.3.4.1 Functional group analysis ............................................................................41
2.3.4.2 Rate curves and kinetics analysis................................................................51
2.3.4.3 Cure mechanisms........................................................................................54
2.3.5 Time-Temperature-Transformation (TTT) ........................................................55
2.4 Concluding remarks ............................................................................................58
2.5 References..........................................................................................................60

3 Deformation and fracture behaviour 63

3.1 Introduction to fracture toughness .......................................................................63

3.2 Linear elastic fracture mechanics (LEFM) ...........................................................64
3.2.1 Strain energy release rate criterion (G-Approach) ............................................64
3.2.2 Stress intensity factor (K-Approach).................................................................67
3.2.3 Fracture toughness test modes........................................................................68
3.3 Toughening of epoxy thermosets ........................................................................69
3.3.1 Epoxy toughening with soft particles (rubber toughening) ................................69
3.3.2 Epoxy toughening with hard inorganic particles ...............................................72
3.3.2.1 Crack pinning mechanism of hard particles .................................................73
3.3.3 Epoxy toughening with thermoplastics .............................................................75
3.4 Experimental .......................................................................................................79
3.4.1 Materials ..........................................................................................................79
v Contents

3.4.1.1 Hyperbranched polymers (HBP) modifier ....................................................79

3.4.1.2 Inorganic filler ..............................................................................................81
3.4.2 Sample preparation..........................................................................................82
3.4.3 Double torsion test (DT) ...................................................................................83
3.4.4 Tensile test.......................................................................................................85
3.4.5 Compression test .............................................................................................85
3.4.6 Dynamic mechanical analysis (DMA) ...............................................................86
3.4.7 Microscopy.......................................................................................................86
3.5 Results & discussion ...........................................................................................87
3.5.1 Effects of strain rate and temperature on the tensile deformation ....................87
3.5.2 Effect of the filler fraction on the tensile deformation ........................................89
3.5.3 Effect of strain rate and temperature on the compressive deformation.............91
3.5.4 Characterization of the fracture toughness behaviour ......................................94
3.5.4.1 Effect of the test temperature ......................................................................94
3.5.4.2 Effect of the strain rate.................................................................................98
3.5.4.3 Effect of the inorganic filler fraction..............................................................99
3.5.4.4 Effect of the inorganic filler size ...................................................................103
3.5.4.5 Effect of the filler/matrix adhesion................................................................106
3.5.5 Hyperbranched polymer toughening ................................................................107
3.5.5.1 Effects of the hyperbranched polymer end-groups ......................................107
3.5.5.2 Hyperbranched polymer modifier with a high epoxy functionality.................109
3.5.5.3 Hyperbranched polymer modifier with a low epoxy functionality ..................114
3.6 Concluding remarks ............................................................................................120
3.7 References..........................................................................................................122

4 Thermo-oxidative degradation behaviour 127

4.1 Introduction to thermo-oxidative degradation ......................................................127

4.1.1 Physical ageing of epoxy-networks ..................................................................127
4.1.2 Thermo-oxidative degradation of epoxy-networks............................................128
4.2 Principals of dynamic mechanical analysis (DMA) ..............................................131
4.3 Experimental .......................................................................................................135
4.3.1 Materials ..........................................................................................................135
4.3.2 Isothermal ageing experiments ........................................................................136
4.3.3 Dynamic mechanical analysis (DMA) ...............................................................136
4.3.4 Thermal gravimetric analysis (TGA) .................................................................136
4.3.5 Mechanical bending test ..................................................................................137
4.3.6 FT-IR micro attenuated total reflection (ATR) spectroscopy .............................137
4.4 Results and discussion........................................................................................138
4.4.1 Mechanical endurance during thermo-oxidative ageing ...................................138
4.4.1.1 Dynamic mechanical behaviour of the virgin state .......................................138
4.4.1.2 Effects of thermo-oxidative ageing on the dynamic mechanical behaviour ..141
4.4.1.3 Effects of thermo-oxidative ageing on the flexural behaviour.......................146
4.4.2 Short-term thermo-oxidative stability ................................................................149
4.4.2.1 Effects of the molar anhydride to epoxy ratio...............................................153
4.4.3 Long-term thermo-oxidative stability.................................................................154
4.4.3.1 Effects of inorganic fillers.............................................................................156
4.4.3.2 Modeling the isothermal weight-loss behaviour ...........................................159
4.4.3.3 The Double-Stage Mechanism (DSM) .........................................................159
4.4.3.4 The Fickian diffusion model .........................................................................165
4.4.4 Thermo-oxidative degradation studied by FT-IR spectroscopy.........................168
4.4.4.1 Thermo-oxidation profiles ............................................................................173
4.4.5 Thermo-oxidative degradation mechanism.......................................................176
vi Contents

4.5 Concluding remarks ............................................................................................179

4.6 References..........................................................................................................181

5 General conclusions and perspectives 185

6 Acknowledgments 188

7 Curriculum Vitae (CV) 189

vii Disclaimer

Disclaimer

The work contained in this thesis has not been previously submitted for a degree or diploma
at any higher education institute. To the best of my knowledge and belief, the thesis
contains no material previously published or written by another person except where due
reference is made.

Signature: (Jens Rocks)

Date: 4. October 2004

Introduction & scope of work 1

Chapter 1

1 Introduction & scope of work

1.1 Classification and history of epoxy thermosets

Epoxy monomers are molecules that contain at least two or more epoxy groups. The prefix
“epoxy” refers to a planar bridge consisting of an oxygen atom bonded to two carbon atoms in
a way that a three membered ring is formed, see Figure 1. Such a chemical structure is also
known as oxirane, ethoxyline or glycidyl group. Further specific information on the functional
epoxy group can be found elsewhere [1, 2]. Most uncured epoxies are thermoplastic resins
that have little value until they are reacted with crosslinking agents.
O
H2C CH R1
Figure 1: Epoxy functional group

The epoxy group can undergo an extremely wide variety of chemical reactions and combine
with a vast array of other materials to produce final products with wide ranging properties [3,
4]. One of the main features of the epoxy resin is that the ring opening reaction (cure) and the
formation of an interconnected three-dimensional molecular network, takes place in the
presence of many different bi- or multifunctional compounds, e.g., anhydrides, amines,
polyphenols etc. that are able to react with the epoxy functional group. The choice of a curing
agent depends on a variety of factors including cost, processing method, curing conditions,
environmental limitations and the mechanical, chemical, electrical and thermal properties
desired in the cured resin.
The reaction rate as well as the temperature at which an epoxy resin cures or hardens
depends upon the curing agent selected. In general, polyfunctional aliphatic amines (e.g.
diethylenetriamine or trimethylenetetramine), polysulfides and polyamido-amines are used
when ambient-temperature cures (T<80°C) are desired. In contrast, aromatic amines,
anhydrides, phenolics, ureas, imidazoles and other resinous hardeners generally require
processing at elevated temperatures (T>80°C) to effect a cure. Certain hardeners allow a
partial curing at room temperature, but require an elevated temperature to complete the cure.
Catalytic curing agents, such as Lewis acids and bases, function as initiators for epoxy resin
homopolymerization. A thorough explanation of the manifold epoxy resin chemistry can be
found in standard textbooks [3-6] and thermoset handbooks [7, 8].
2 Chapter 1

In their cured form, epoxy resins belong to the class of highly crosslinked polymers, also
called thermosets or thermosetting polymers, which are an important group of synthetic
polymers. The main difference between thermosets and other organic polymers such as
thermoplastics and elastomers is in their three-dimensional densely crosslinked structures,
i.e., an “infinite” molecular network with covalently bonded junction points having
functionalities greater than two. Cured resins do not melt, but soften to a rubbery state at
temperatures near and above the glass-transition temperature. In general, raising the cross-
link density results in a higher glass-transition temperature and a higher continuous-use
temperature; however, it reduces the flexibility of the cured resin.
Since their first commercial introduction in 1947, epoxy resins (EP) have experienced
tremendous growth on a commercial sale. Rather than becoming a commodity item, epoxy
resins have developed into engineering plastics, fulfilling the requirements in many different
applications. Today, their global share in the world plastic market is about 0.5%
corresponding to a consumption of roughly 1 billion metric tons per year [9] (Table 1).

This table is not available online. Please consult

the hardcopy thesis available from the QUT Library.

The main areas of epoxy monomer consumption are in coatings, surface protection and
electrical and electronics applications. Coatings are usually applied as liquid or solution type
varnishes in severe corrosive environments, as powder coatings for sheet steels in consumer
goods and by cathodic electro-deposition from aqueous bath onto car bodies. In the building
industry, epoxies are becoming more and more important for the repair and conservation of
concrete structures [10]. Electrical insulation is a traditional area for epoxy use [11]. High
voltage insulators, impregnated coil insulation and electrical laminates are the most striking
examples. Advanced structural composites based on glass, aramid or carbon fibers and
epoxy matrixes as well as structural adhesives [12] are also very important. The market
volume is, however, for the latter applications still very small. In Figure 2, the market shares
for different application areas are given.
Introduction & scope of work 3

This image is not available online. Please consult the hardcopy

thesis available from the QUT Library.

Figure 2: Epoxy resin consumption by end use in 1995 [13]

Today, the most frequently used epoxy resin is produced from epichlorohydrin and bisphenol-
A. Known as conventional or DGEBA-type resins (see structure I in Figure 3), these diglycidyl
ethers of bisphenol-A (and their higher molecular weight homologs) still account for about 80-
85% of the total worldwide epoxy consumption. Although DGEBA-type resins comprise the
majority of epoxy consumption, four other types of basic epoxies are of commercial
importance. These include, as shown in Figure 3, (II) brominated resins, (III) novolac resins,
(IV) cycloaliphatic resins and (V) phenoxy resins.
CH3
H2C CH CH2 O O CH2 CH CH2
(I)
CH3
O O
Br Br
CH3 (II)
H2C CH CH2 O O CH2 CH CH2
CH3
O O
Br Br
O O O
(III)
H2C CH CH2 O O CH2 CH CH2 O CH2 CH CH2

CH2 CH2
n
O
CH2 (IV)
O
O O

OH
CH3 CH3 (V)
HO O CH2 CH CH2 O OH
CH3 n CH3

Figure 3: Commercial types of basic epoxies (I) diglycidyl ether of bisphenol-A (DGEBA), (II)
brominated epoxy resins, (III) epoxy novolac (EN), (IV) cycloaliphatic epoxy resin (CEP), (V)
phenoxy resin
4 Chapter 1

Brominated epoxies are ignition-resistant compounds produced from epichlorohydrin (or low-
molecular weight DGBPA resin) and halogenated bisphenol-A (e.g., tetrabromobisphenol-A).
Brominated epoxies have grown in commercial importance over the past fifteen years and are
now used extensively in the production of laminates for printed circuit boards.
Novolac epoxies include primarily polyglycidyl ethers of phenol-formaldehyde or cresol-
formaldehyde resins. The diglycidyl ether of bisphenol-A is the simplest member of this
family. When cured, epoxy novolacs have better high-temperature properties and chemical
resistance than conventional epoxy resins. Epoxy novolacs are used in electrical laminates
and encapsulations, moldings and castings and reinforced plastics requiring high chemical
resistance.
Cycloaliphatic epoxy resins are usually non-glycidyl ethers that can be produced by the
peracetic acid epoxidation of cyclic olefins. Advantages of these resins over conventional or
novolac epoxies include lower viscosities, higher heat-distortion temperatures, lower dielectric
constants and excellent weatherability. Cycloaliphatic epoxy resins are particularly useful for
electrical applications (e.g., as replacements for porcelain in insulators) and for filament-
wound structures. The flexibilized versions are also recommended for high-performance non-
electrical applications.
Phenoxy resins are high-molecular-weight polymers (30’000-200’000 g/mole) derived from
epichlorohydrin and bisphenol-A. Even if they use the same raw materials as conventional
epoxies, they lack terminal epoxy groups. Although small relative to the total epoxy resin
market, phenoxy resins are used in coatings and adhesives and in binders for magnetic
tapes.
The remaining part of epoxy resins production comprises specialties. These include
multifunctional epoxies (e.g. glycidyl amines), aliphatics, glycidyl imides and glycidyl esters.
Together these types of epoxies are believed to comprise, at most, 5% of the epoxy market.
Research and development activities related to epoxy resins, however, by far exceed their
purely quantitative significance on the market. The main reason is a unique combination of
desirable properties, e.g. high glass transition temperature, high elastic modulus and
mechanical strength, excellent resistance to corrosion and chemicals, good dielectric
properties, high creep resistance and generally good performance at elevated temperatures,
making epoxies suitable for use in applications with high mechanical, thermal and chemical
demands. Recent developments in epoxy resin chemistry and applications have concentrated
more on further improvement of existing products rather than looking for new syntheses.
Introduction & scope of work 5

1.2 Scope and survey of this thesis

For more than three decades, silica, dolomite, wollastonite and alumina filled, anhydride
cured epoxy systems have been extensively used in the electrical power engineering
industry. The main application areas include medium and high voltage insulation for gas
insulated switchgears, generator circuit breakers, dry type distribution transformers, dry type
bushings and high voltage cable accessories. Required material properties in the power
electrical engineering industry include high electrical insulation resistance, high mechanical
strength and modulus, low creep under thermo-mechanical loads, as well as high damage
tolerance with reasonable toughness. Moreover, miniaturization and increased performance
demands have led to higher operating temperatures over the years. Today, continuous
operating temperature levels are at 105°C with guaranteed lifetimes up to 30 years.
Fierce competition on the electrical market has led to development of more compact power
equipment with higher performance ratings. However, this substantially requires materials
with continuous operating temperatures in the range of 150°C. Presently used insulation
organic materials cannot withstand these relatively high thermo-mechanical requirements.
Consequently, new insulating systems with higher glass transition temperatures (Tg) and
hence higher continuous operating temperatures are needed.
In 1974, Graham and O’Conner have shown that a non-catalyzed epoxy resin formulation can
be cured with a co-anhydride mixture at relatively low curing temperatures, while yielding a
high thermo-mechanical property profile (Tg>200°C) [14]. Later Vohwinkel and co-workers
extended the initial work of Graham and O’Conner and proposed a variety of further co-
anhydride/epoxy systems [15] that show a potential for highly demanding electrical insulation
applications. The high Tg of these materials is here combined with high mechanical strength,
high Young’s modulus (E), very good electrical insulation resistance and excellent processing
and rheological properties. The ultimate Tg of these systems can be well above 200°C
demonstrating that continuous operating temperatures of 150°C are feasible. Furthermore,
reduction in processing time and low temperatures for curing and post-curing make this new
class of epoxy materials very attractive for a wide range of industrial applications.
The main objective of the thesis is to identify the potential and limitations of these co-
anhydride/epoxy formulations for structural parts in the electro-technical industry. This
required combining a wide range of material issues including polymer chemistry,
polymerization mechanism and kinetics, rheology, thermodynamics, morphology and physical
and mechanical properties in order to establish a comprehensive material characterization
and more fundamental understanding of the material.
6 Chapter 1

In Chapter 2, the general curing reactions between anhydride and epoxy are reviewed. A
specific study of the liquid-solid transition occurring during the crosslinking reaction is
discussed. In particular the chapter focuses on the analysis of the chemical and structural
changes taking place during the in-situ polymerization by using real-time calorimetric,
spectroscopic Raman scattering, and chemo-rheological analytical methods. Kinetic analysis
from the conversion of the functional groups and the obtained empirical rate curves are
discussed in the light of proposed reaction mechanisms for the thermal cure behaviour of the
investigated uncatalyzed co-anhydride/epoxy-systems.
Chapter 3 discusses the thermo-mechanical deformation and fracture behaviour of the new
materials. The possibility of overcoming the intrinsic brittleness of the current highly
crosslinked networks by selecting alternative toughening additives is illustrated by using
hyperbranched thermoplastics with functionalized end-groups. Furthermore, the influence of
inorganic particles as externally added filler is elaborated in more detail. The obtained
morphologies of the polymer composites and the effects on the deformation and toughening
probability are outlined.
Chapter 4 presents a specific study on the long-term behaviour of the material during thermo-
oxidative ageing. In particular, structural network changes occurring during ageing at elevated
temperature are discussed by taking into account the changes of intrinsic chemical
structures, as well as gravimetric, and viscoelastic alterations. On the basis of experimental
observations a model allowing a reasonable prediction of an expected long-term property
profile is created and verified.

1.3 References

[1] Lwowski W., Comprhensive Hetrocyclic Chemistry, Pergamon Press, Oxford, (1984).
[2] Lewars E.G., The Chemistry of Ether Linkage, Wiley Interscience, New York, (1967).
[3] May C.A., Epoxy Resins, Marcel Dekker, New York, (1988).
[4] Ellis B., Chemistry and Technology of Epoxy Resins, ed.1, Blakie Academic
Professional, Glasgow, (1993).
[5] Hamerton I., Recent Development in Epoxy Resins, Rapra Technology, Shawbury,
(1996).
[6] Lee H., Neville K., Handbook of Epoxy Resins, McGraw-Hill, New York, (1967).
[7] Franck A., Biederbick K.H., Kunststoff Kompendium, Vogel Publisher, Würzburg,
(1990).
[8] Saechtling H., Kunststoff Taschenbuch, Hanser Publisher, Munich, (1986).
[9] Merino J.A., Kunststoffe, 91, 352 (2001).
[10] Selwitz C., Epoxy Resins in Stone Conservation, The Getty Conservation Institute,
Marina del Rey, (1992).
[11] Beyer M., Epoxidharze in der Elektrotechnik, Expertenverlag, Ehningen, (1983).
[12] Kinloch A.J., J. Mater. Sci., 15, 2141 (1980).
[13] Goodman S.H., Handbook of Thermoset Plastics, 2nd ed. , Noyes Publications, New
Jersey, (1998).
Introduction & scope of work 7

[14] Graham J.A., O`Conner J.E., Adhes. Age, 21, 30 (1978)

[15] Patent Application DE 198 28 248; Vohwinkel F. (1998).
8 Chapter 2

Chapter 2

2 Epoxy-anhydride crosslinking reaction

2.1 Introduction

The curing of epoxy resins is an important reaction for the manufacture of structural parts, in
particular for electrically insulating materials. A cure reaction can be described as initial
chain formation, followed by linear growth and branching, gelation, and finally vitrification to
a crosslinked, infinite network.
Presently, multifunctional aromatic or aliphatic amines are mainly used as curing agents for
epoxy resins. Consequently, the curing behaviour and kinetics of amine-cured epoxy
systems have been intensively studied by many different analytical methods such as
differential scanning calorimetry (DSC) [1], gel permeation chromatography (GPC) [2],
nuclear magnetic resonance (NMR) [3], and vibrational spectroscopy [4, 5], rheological
analysis [6], and dielectric thermal analysis [7, 8], respectively, yielding an extensive
knowledge of the reaction mechanisms involved. However, the use of amine curing agents
has the general disadvantage that new polar functional groups are created and
subsequently relatively high amounts of polar groups, e.g. water, can be absorbed. Hence,
this behaviour can significantly influence the final property profile (i.e. mechanical and
dielectric properties) of the glassy material and restricts the use of the materials in numerous
technical applications such as medium and high voltage insulation.
The formation of polar groups does not occur when carboxylic acid anhydrides are used as
the curing agent. Carboxylic acid anhydrides were among the first curing agents used with
epoxy resins [9] and their generalized structure is given in Figure 4. Today the group of
anhydrides has achieved a significant commercial importance due to a combination of their
cure profiles and cured properties. Consequently, after amines, acid anhydrides constitute
the next most commonly used class of curing agents.
O
R1 C
O
R2 C
O
Figure 4: Anhydride functional group
 Epoxy-anhydride crosslinking reaction 9

In contrast to amine curing agents, anhydride reactions with epoxy are usually characterized
by a low exothermal reaction, low shrinkage, low internal stresses, reduced water
absorption, high glass transition temperatures and excellent electrical properties. Moreover,
these properties are retained even under moderate to high operating temperatures due to
the good thermal stability [9-11]. On the other hand most anhydride-epoxy curing reactions
are less reactive and relatively high reaction temperatures (≥120°C) are required to initiate
and propagate the curing reaction [12-14]. This results in very expensive processing and
handling steps. Although the selection of anhydride/epoxy systems for many technical
applications is essential, the formation and structure of the cured products is much less
understood than in epoxy/amine-systems. Especially, information on the curing behaviour of
multifunctional glycidyl resins in the presence of cyclic carboxylic acid anhydrides is rather
limited.
The objective of this chapter is to study the curing kinetics and mechanism of a novel
epoxy/co-anhydride-system by different analytical methods. Dynamic calorimetric
experiments are performed and the obtained results are used for the extraction of the
appropriate kinetic parameters for the analysis of cure. Isothermal calorimetry, chemo-
rheometry, and FT-Raman spectroscopy are carried out, in order to compare the different
analytical methods and outline underlying assumptions and limitations. Moreover, real-time
vibrational FT-Raman spectroscopy, as a complementary method to DSC and rheometry,
provides additional functional group information that is used to identify the curing
mechanism of the epoxy/maleic anhydride/hexahydrophthalic anhydride-system.
Finally, the complex physical changes occurring during cure of the epoxy resin are
visualized by means of a principal Time-Temperature-Transformation (TTT) diagram.

2.1.1 The uncatalyzed epoxy/anhydride curing reaction

The epoxy group is characterized by its reactivity towards nucleophilic and electrophilic
species and thus it is reactive to a wide range of reagents. The formation of polyester from
the uncatalyzed reactions of epoxy monomers and cyclic carboxylic acid anhydrides has
been intensively studied during the last years [15]. The research resulted in a generally
accepted stepwise reaction mechanism that occurs during the non-catalyzed polymerization
[16-19] as schematically shown in Figure 5. The mechanism involves the presence of
hydroxyl groups, which act as an initiator for the reaction. The hydroxyl groups are mostly
present as substituents on a fraction of the epoxy resin molecules. An attack of these
hydroxyl groups on the anhydride molecules forms a monoester having a free carboxyl
group (reaction I). The monoester then reacts with the epoxide to yield a diester and a new
secondary hydroxyl group (reaction II), thus starting the cascade of curing reactions
10 Chapter 2

(reaction III). The evidence that the quantities of mono– and diester produced are equal to
the amount of consumed anhydride confirmed these reaction sequences. Moreover, the fact
that no significant difference in the consumption of epoxy and anhydride groups [20] has
been observed at least up to gelation supports the proposed reaction scheme.
The observation that the number of epoxy groups at higher conversions decreases faster
than the increase in the diester groups can be attributed to the etherification reaction. This
reaction occurs between the epoxy and hydroxyl groups under the catalytic influence of
anhydride or carbonyl groups (reaction IV), and constitutes a significant side reaction. Other
side reactions, such as the epoxy homopolymerization, are also possible but these are
considered of little significance here as the curing temperatures are well below 180°C for the
studied TGDDM system [5].
O
k1 COO R1
(I)
O + R1 OH
COOH
O
COO R1 COO R1
O k2 (II)
+ CH2 CH CH2 R2
COOH COOCH2 CH OH
C R2
H2

O
COO R1 O (III)
COO R1 HO

COOCH2 CH OH + O
COOCH2 CH O
C R2 O C R2 O
H2 H2

OH
O (IV)
k3
H2C CH CH2 R1 + R2 OH R2 O CH2 CH CH2 R1

Figure 5: Schematic mechanism (reaction (I) to (III)) of uncatalyzed epoxy/acid-anhydride

addition reaction as suggested by Fisch and Hofmann [16, 21, 22]. The hydroxyl group is
present in the epoxy resin. The etherification reaction (reaction (IV)) occurs at higher
temperatures and is catalyzed by anhydride or carbonyl groups.

2.1.2 The base catalyzed epoxy/anhydride curing reaction

To accelerate the curing reaction for industrial processing, strong Lewis bases, such as
tertiary amines, imidazoles or ammonium salts are usually used to catalyze the cure of
epoxy resins with cyclic carboxylic acid anhydrides. A review of some of the postulated
reactions for the catalyzed epoxy/anhydride polymerization is given and discussed in this
section.
The initiation mechanism is still a matter of controversy and several reaction models are
available. Irrespective of the different proposed mechanisms three types of initiation have
 Epoxy-anhydride crosslinking reaction 11

been proposed so far [15, 23, 24]: (1) ionic mechanism according to Fischer, (2) initiation
with the participation of pre-existing proton donor, and (3) Tanaka’s mechanism of initiation
by the proton donor arising in the reaction.
According to Fischer [25] and several other workers, the reaction catalyzed by the presence
of tertiary amines is initiated by the activation of the anhydride. It was proposed that the
anhydride undergoes a nucleophilic attack by the lone electron pair of the tertiary amine,
giving rise to a zwitterion containing a quaternary ammonium salt and a carboxylate anion
as shown in Figure 6 in reaction (V). This activated tertiary amine complex may be
considered as the real initiator of the chainwise co-polymerization. Further propagation
occurs either through the reaction of the generated carboxylate with an epoxy group and the
accompanied formation of new alkoxide anion (reaction VI) or with another anhydride under
regeneration of a carboxylate (reaction VII), see Figure 6.
The generated alkoxide group is supposed to react at a very fast rate with another anhydride
group and regenerate species containing carboxylate as active center. This suggests the
initiation of an anionic strictly alternating copolymerization as later shown by Trappe et al.
[26]. The fact that the presence of alkoxide anions could not be detected by NMR, was
explained by their fast reaction rate with anhydrides [23, 27].
The presence of termination and chain transfer steps has been proposed in the literature
[10, 15, 28] to account for the fact that molecular weights are lower than those expected for
an ideal living polymerization. A possible mechanism involving the regeneration of the
tertiary amine is shown in Figure 6, reaction (VIII). The regenerated tertiary amine becomes
available to initiate a new chain.

This image is not available online. Please consult the hardcopy

thesis available from the QUT Library.
12 Chapter 2

Figure 6: Schematic mechanism of the tertiary amine catalyzed anhydride/epoxy co-reaction

according to Fischer [25].

Tanaka and Kakiuchi [20] proposed that the curing reaction requires the presence of proton
donor (HA) as a catalyst, which forms a complex with the amine according to the reaction
scheme illustrated by reaction (IX) to (XII) in Figure 7:

This image is not available online. Please consult the hardcopy thesis available
from the QUT Library.

Figure 7: Schematic reaction mechanism of tertiary amine catalyzed anhydride/epoxy co-

reaction according to Tanaka and Kakiuchi [20]

Additionally, Tanaka and Kakiuchi [29] stated, that the reaction might also be initiated in the
absence of an external proton donor because of the formation of a proton donor during the
isomerization of the epoxy to a derivative of alkyl alcohol as shown in reaction (XIII) in
Figure 8.
This image is not available online. Please consult the hardcopy
thesis available from the QUT Library.

Figure 8: Initiation step of a catalyzed anhydride/epoxy co-reaction according to Tanaka and

Kakiuchi [29]
 Epoxy-anhydride crosslinking reaction 13

Luston and Manasek [30, 31] also proposed the formation of allyl-alcohol derivatives by
isomerization of epoxide. The hypothesis, which was based on conductivity measurements,
presumed that the active site for the polymerization is a three-component complex that
contains amine, epoxide, and anhydride as shown in reaction (XIV) and (XV) in Figure 9.
However, in contradiction Antoon and Koenig [24, 32] did not confirm the formation of allyl-
alcohol groups during the base-catalyzed reaction between an epoxy and anhydride by
using different spectroscopic methods.

This image is not available online. Please consult the hardcopy thesis
available from the QUT Library.

Figure 9: Initiation step of a catalyzed anhydride/epoxy co-reaction according to Luston and

Manasek [30]

Although the various epoxy/anhydride/amine-systems in the above discussion are extremely

similar, a generally accepted reaction mechanism of the complicated base catalyzed curing
reaction between anhydrides of epoxies is not available at the present time. However, most
authors agree that the reaction mechanism of the epoxy/anhydride catalyzed with tertiary
amine is an anionic one and that propagation proceeds via two reactions: (1) acylation and
(2) the reaction between epoxide and carbonyl derivatives [20, 23, 25, 27, 29, 30, 33-35].
There is still no definite validation regarding the question whether the initiator remains
chemically bound during the whole course of the reaction. Some authors describe a release
of the initiator [28, 29] but there is disagreement on this point [36, 37].
The fact that the competing epoxy-homopolymerization is effectively suppressed has been
independently reported in literature [20, 24, 25].

2.2 Experimental

2.2.1 Materials

A commercial multifunctional aminoglycidyl-resin, AralditeMY721 (Vantico, Switzerland)

mainly based on N,N,N’,N’-tetraglycidyl-4,4’-diamino-diphenylmethane (TGDDM) was used
as epoxy monomer for all investigations. Araldite®MY721 is most commonly synthesized by
a reaction that involves several sequences as shown in Figure 10. In this sequence the
14 Chapter 2

stepwise reaction of 4,4’-diaminodiphenylmethane (DDM) with a slight excess of

epichlorohydrin yields chlorohydrin which is subsequently dehydrochlorinated with sodium
hydroxide to form epoxy groups [38, 39].

NH2 CH2 NH2 + 4 CH2 CH CH2 Cl

Cl CH2 CH CH2 N CH2 N CH2 CH CH2 Cl

2 2
OH OH

4 NaOH

CH2 CH CH2 N CH2 N CH2 CH CH2

2 2
O O
Figure 10: Idealized reaction scheme for the synthesis of N,N,N’,N’-tetraglycidyl-3,3’-4,4’-
diamino-diphenylmethane (TGDDM)

An anhydride hardener mixture consisting of maleic anhydride (MA) and hexahydrophthalic

anhydride (HHPA), both supplied by Fluka, was used to cure the TGDDM resin. MA is a
white crystalline solid with a theoretical molecular weight of 98.16 g/mole and a melting point
of 53°C. HHPA is a waxy solid having a theoretical molecular weight of 154.17 g/mole and
melting point of about 35°C. The chemical structures of the used materials are shown in
Table 2. The mixing ratio, unless other stated, was 100/48/32 (TGDDM/MA/HHPA) by
weight, reflecting a molar anhydride to epoxy ratio, r, of 0.8.

Table 2: Chemical structures and material composition of the materials used

Chemical name /Tradename Chemcial structure Molecular Weight Molar
weight fraction fraction
[g/mole] [%] [%]
O O
*
N,N,N’,N’-Tetraglycidyl-4,4’- 468 55.55 55.09
N CH2 N
diaminodiphenylmethane
(TGDDM); AralditeMY721
O O

O O O
Maleic anhydride (MA) 98.16 26.67 31.53

O
Hexahydrophthalic anhydride 154.17 17.78 13.38
(HHPA) O

O
*) Determined by potentiometric titration according to DIN 16945 standard
 Epoxy-anhydride crosslinking reaction 15

2.2.2 Sample preparation

The epoxy monomer was preheated to 65°C to reduce the viscosity of the MY721-resin,
before the premixed hardener-mixture was added and dissolved with an Ultraturraxhigh
shear mixer at 13.500 rpm for 5 minutes to obtain a homogeneous mixture. To prevent a
temperature rise and any further reactions during mixing, all samples were cooled during the
mixing process in ice water. Afterwards, the mixture was degassed for 3 minutes at
pressures below 100 mbar and stored in liquid nitrogen in order to prevent further curing.

2.2.3 Wide angle x-ray diffraction (WAXD)

X-ray diffraction graphs were recorded at room temperature using a Siemens D-5000
instrument, operated at 50 KV and 50 mA with a Ni-filtered Cu-Kα radiation source. The
scanning speed and step size were 0.05°/5s and 0.05°. Intensities were quantified with a
Braun position sensitive counter. The samples were placed on a background-free
amorphous glass plate and analyzed as produced.

2.2.4 Differential scanning calorimetry (DSC)

The epoxy mixture (10±2 mg) was filled in sealed aluminum pans with a pin-hole in the lid
and stored in liquid nitrogen to prevent further curing. All measurements were carried out on
a Perkin-Elmer differential scanning calorimeter, Model DSC-7. The continuous heating
scans were performed at heating rates (ϕ) of 2.5, 5, 7.5, 10, 15 and 20°C/min, respectively,
in a temperature range from –60°C to 300°C.
Isothermal scans were accomplished at 100, 110, 120, 140 and 160°C. The reaction was
considered complete when the rate curve leveled off to a baseline. The total area under the
exothermal curve, based on the extrapolated baseline at the end of the reaction, was used
to calculate the isothermal heat of cure at a given temperature.
The experimental data were analyzed for the kinetic models using a least squares curves
fitting routine based on the Marquardt-Levenberg algorithm. In each case the fitted data
were verified for the presence of a global minimum, so that the values of the kinetic rate
parameters obtained from the fit were “best values”.

2.2.5 Gas chromatograhy (GC)

16 Chapter 2

A two column micro gas chromatograph (MTI M200H) equipped with a 8 m Poraplot Q and a
10 m molecular sieve (5 Å) column was used to separate the trapped gaseous products
during isothermal cure at 65°C. Helium with a flow rate of 10 ml/min was used as a carrier
gas. Continuous data acquisition was carried out every 3 minutes until the samples reached
the point of gelation.

2.2.6 Temperature modulated differential scanning calorimetry (TMDSC)

Temperature modulated DSC (TMDSC) is a novel thermal analysis technique, which clearly
separates thermal transitions that have reversing and non-reversing components with
respect to a small amplitude temperature oscillation [40, 41]. This is accomplished by
imposing a sinusoidal temperature modulation on top of a constant underlying heating rate
or constant temperature. When the temperature is oscillated around a constant underlying
heating rate or temperature, the response of the sample consists of an oscillating heat flow,
which has both reversing and non-reversing components. A discrete Fourier transformation
is then used to deconvolute the signal into a conventional DSC curve and to separate the
reversing and non-reversing components. Additionally, the separation of the heat flow into
its components facilitates determination of heat capacity from a single MDSC experiment as
opposed to multiple conventional DSC experiments.
TA Instruments DSC 2910 Differential Scanning Calorimeter with the MDSC capability was
used to study the epoxy systems. In all experiments the cooling accessory was used to
facilitate accurate temperature control and to achieve sub-ambient temperatures at –60°C. A
temperature ramp at a constant heating rate of 1°C/min and an underlying modulation with a
0.2°C amplitude, and a period of 60 seconds under nitrogen purge at 40 ml/min was used in
all experiments.

2.2.7 Chemorheological analysis

A dynamic rheometer (Bohlin CVO75) was used to determine the rheological characteristics
of the studied epoxy formulations. Complex viscosity, η*, storage modulus, G’ and loss
modulus G’’ were obtained by oscillatory shear measurements. All measurements were
carried out in a plate–plate configuration (diameter=40 mm; gap=0.3 mm) at temperatures of
30, 40, 50, 60, 70, 80, 90, 100 and 110°C, respectively. Isothermal dynamic tests were
performed to determine the time to gelation, as the time where the storage (G’) and loss
(G’’) moduli cross over (tanδ=1), at a constant frequency, ω. All tests were made at a
frequency of 1 Hz and a strain of 3%. The temperature reached the set point 1min after
sample insertion. The steady state temperature was kept within 0.5°C of the set point.
 Epoxy-anhydride crosslinking reaction 17

2.2.8 Fourier-Transform Raman Spectroscopy (FT-Raman)

All spectroscopic experiments were carried out on a Perkin-Elmer System 2000 NIR FT-
Raman spectrometer, powered by a diode pumped Nd-YAG laser (λ=1064 nm) as an
excitation source and a room temperature InGaAs photoelectric detector. The backscattered
radiation was collected at 180° to the excitation. Typical spectra were recorded in a range of
200 to 3800 cm-1 at a laser power of 320 mW. Every 4 minutes a spectrum was collected
consisting of 32 co-added scans with spectral resolution of 8 cm-1. The premixed epoxy
mixture (4±0.2 g) was filled in a NMR sample tube and placed in the preheated heating
block under isothermal conditions at 55, 65, 75, 100, and 120°C, respectively. The steady
state temperature was controlled with an accuracy of ±0.2°C. Spectra were recorded from
the moment the tube was introduced in the heating block. To reach the ultimate conversion
some samples were also post-cured at 150°C for 1 hour.

2.3 Results & discussion

2.3.1 Phase behaviour of the binary MA-HHPA mixture

The phase behaviour of the various MA-HHPA mixtures was analyzed using continuous
heating DSC analysis and X-ray diffraction (XRD) techniques at room temperature as shown
in Figure 11. The DSC scans in Figure 11(a) show two phenomena. First, the continuous
decrease in the melting temperature of the HHPA-rich phase with increasing HHPA content.
Second, the presence of a constant melting temperature at round 7°C, indicating that the
phase diagram (see Figure 12) of both anhydrides has a eutectic point between a molar MA
fraction of 0.2 and 0.4. The XRD data in Figure 11(b) also reveal a depression in the melting
point of the eutectic temperature below room temperature. Consequently, only a broad
amorphous peak centered at around 2θ=17° was observed between the molar maleic
anhydride fractions of 0.2 to 0.5 in the MA/HHPA system.
 18 Chapter 2

a) b)

0.0 0.0

0.2
0.2
Heat Flow [mW/mg]

Count rate [a.u.]

Molar fraction MA

Molar fraction MA
0.4
0.4

0.6 0.6

0.8
0.8

1.0 1.0
-20 0 20 40 60 10 15 20 25 30
Temperature [°C] 2θ [°]

Figure 11: (a) Continuous heating DSC-thermograms (ϕ=10°C/min)of the system maleic
anhydride and hexahydrophthalic anhydride (MA/HHPA) as a function of the MA fraction,
indicative of eutectic phase behaviour. (b) XRD-data of the system maleic anhydride and
hexahydrophthalic anhydride (MA/HHPA) as a function of the MA fraction measured at room
temperature.

The behaviour of the experimentally determined melting point as a function of the MA-
content in the binary MA/HHPA phase diagram is shown in Figure 12. The continuous lines
in this plot represents the decrease of the melting point for the ideal situation, and were
calculated using the Flory-Huggins equation (2.1), assuming zero interaction energy, and
using the theoretical melting points (MA=55,2°C, HHPA=31.8°C) and melting enthalpies
from the pure anhydrides (MA=14038 J/mol and HHPA=15456 J/mol):
1 R 1
= ln ϕ − (2.1)
Tm Tm ∆H 0

The excellent agreement between the experimental and predicted phase behaviour in Figure
12, confirms the ideal mixing assumption (zero interaction energy), which was used for the
theoretical prediction.
 Epoxy-anhydride crosslinking reaction 19

molar fraction xHHPA

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0
60

50

40

Temperature [°C] 30 L

L+β
20
L+α
10

0
α+β

-10
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

molar fraction xMA

Figure 12: Partial phase diagram of the binary MA/HHPA-system based on DSC data,
indicative of eutectic behaviour. The solid lines represent calculated values based on the
Flory-Huggins relation eqn (2.1)

2.3.2 Curing reaction studied by differential scanning calorimetry (DSC)

Thermal curing kinetics of epoxy systems has been the subject of many studies in the past
[1, 42-51]. The most common way to study curing is to use differential scanning calorimetry
(DSC). DSC is a powerful technique that provides information about the exothermicity of the
epoxy ring opening reaction. Assuming that there are no other enthalpic events contributing
to the heat flow than the chemical curing reaction, the absolute recorded value of heat flow,
dH / dt , is proportional to the reaction rate or rate of conversion, dα/dt:

dH dα (2.2)
~
dt dt
Considering the validity of this basic assumption, the DSC technique can be used to study
the curing kinetics of the investigated thermosets.

2.3.2.1 Non isothermal DSC

The overall cure behaviour of the MY721 in combination with only pure MA or HHPA and the
mixture of both hardeners was followed by DSC with a continuous heating rate of
ϕ=10°C/min. The obtained results are shown in Figure 13. It was observed that all systems
react in an exothermic manner. The combination of both hardeners results in a single
exothermal event and no overlapping exothermic doublet peak was detected in the kinetic
DSC-scan, which confirms the existence of a quasi-eutectic hardener mixture. Furthermore,
20 Chapter 2

if only MA is used as hardener, the exothermic peak temperature (Tm∼160°C) is lower than
when HHPA (Tm∼220°C) alone is used as hardener showing that MA reacts faster with
MY721.
Besides, directly after adding maleic anhydride to the epoxy resin a visible change in color
from light yellow to a deep red-brown was observed. It is proposed that the observed
discoloration can be related to the interaction between MA and the tertiary amine structure
of the used TGDDM-molecule (see Table 2). This tertiary amine-activated color change in
maleic anhydride containing polymers has been described for the first time by Martz et al.
[52]. The authors reported that all anhydrides, which contain a carbon-carbon unsaturation
in direct conjunction with the anhydride ring, e.g. maleic anhydride, give in combination with
tertiary amines dark colored solutions. Later McCormick and Chang [53] proposed that the
color change mechanism was based on the formation of zwitterions intermediates; while
Panarin and Sevetlova [54] hypothesized that maleic anhydride interacts with the tertiary
amine to give charge transfer complexes.
Normalized Heat Flow

exotherm

MA
HHPA
MA/HHPA (3/2 w/w)

50 100 150 200 250 300

Temperature [°C]

Figure 13: Continuous DSC heating thermogram (ϕ=10°C/min) of pure anhydride (MA and
HHPA) and co-anhydride (MA/HHPA) MY721-resin system. The stoichiometric
anhydride/epoxy ratio is r=0.8 for all systems.

The influence of different heating rates on the curing behaviour of the MY721/MA/HHPA-
system is illustrated in Figure 14. To compare the different runs, the measured heat flow,
dH dt [W], was normalized by the sample weight, m [g], and the heating rate, ϕ [°C/s]. The
resulting unit is the specific heat, [J/g°C]. The exothermal peak area in a plot of the specific
heat as a function of temperature represents the reaction enthalpy of curing. In addition, the
broad exothermal reaction above 250°C, reflects the onset of thermal degradation reactions,
which is in accordance with earlier results from Morgan et al. [5] for TGDDM-based epoxy
 Epoxy-anhydride crosslinking reaction 21

systems. The increased heating rate shifted the onset-, peak- and offset- temperatures of
the curing reaction to higher temperatures. Also, it reduced the peak height, the peak area
and therefore the exothermic reaction enthalpy (∆Htot), see table in Figure 14. This kind of
dependency is in accordance with other studies on epoxy-systems [55-58], but the decrease
in ∆Htot with increased heating rate was less than 20%, which is much less than the change
of ∼40% observed for the MY721/MA/HHPA-system here.

exothermal
0
Specific Heat dH/(dT m) [J/gK]

-2

-4
20 °C/min
ϕ [°C/min] -∆H [J/g]
-6 20 -250
10 °C/min 15 -267
10 -291
-8
5 -377
2.5 -412
2.5 °C/min
-10
60 80 100 120 140 160 180 200 220 240 260 280 300
Temperature [°C]

Figure 14: Dynamic DSC scans at different heating rates (ϕ=2.5, 10 and 20°C/min) of the
unreacted MY721/MA/HHPA-system(r=0.8), directly after mixing.

Consequently, the large difference in the total reaction enthalpy (∆Htot) values is attributed to
the specific behaviour of MA in the investigated system. This can been seen in Figure 15
which shows the behaviour of ∆Htot as a function of heating rate for the MY721/HHPA-,
MY721/MA- and MY721/MA/HHPA-system with a fixed anhydride to epoxy ratio of r=0.8.
For the MY721/HHPA-system an increase in ∆Htot of about 15% with increased heating rates
ranging from 2 to 20°C/min can be observed. On the other hand the MY721/MA-system
shows a significant decrease in ∆Htot by about 134% when varying the heating rate in the
same range. This significant increase in the total reaction enthalpy at decreased heating
rates may be attributed to the higher temperatures attained in these samples, due to the
decomposition of maleic anhydride. In addition, maleic anhydride may homopolymerize [59]
in the presence of nitrogen bases, i.e. the TGDDM structure, so providing a contribution to
the total heat flow that is not directly linked to the curing reaction.
22 Chapter 2

1200
MY721/MA
MY721/MA/HHPA
1000 MY721/HHPA

-∆H [J/g]
800

600

400

200
0 2 4 6 8 10 12 14 16 18 20 22
Heating rate [°C/min]

Figure 15: Total reaction enthalpy (∆Htot) as a function of heating rate (ϕ) for different
anhydride cured MY721-systems with a molar anhydride to epoxy ratio, r=0.8. Solid lines
represent the extrapolation to ϕ=0°C/min. The dotted line represents the extrapolation of the
MY721/MA/HHPA-system to ϕ=0°C/min considering only heating rates at 10, 15, and
20°C/min.

Evidence on a significant amount of volatiles that were already lost at relatively low curing
temperatures at 65°C was given by in-situ gas chromatography (GC), see Figure 16. During
cure mainly water (H2O) and carbon dioxide (CO2) were released, which can be attributed to
the spontaneous decarboxylation of maleic anhydride, since no H2O and only very small
traces of CO2 were detectable by the same GC-method when using only HHPA as pure
curing agent (Figure 16). This observation is in agreement with previous studies [59-61],
although the exact mechanism remains unclear. Davie et al. [61] postulated that the
decomposition of liquid maleic anhydride is activated by the presence nitrogen bases (here
the tertiary amine structure of TGDDM structure, see Table 2).
The significant shift in the baseline of the thermograms in Figure 14 at the higher heating
rates may reflect the artifacts due to the side reactions described above. As will be noted
later, the significant differences between dynamic DSC methods and isothermal techniques
can arise from high temperature reaction pathways not encountered during the normal
isothermal cure process.
 Epoxy-anhydride crosslinking reaction 23

800
800

Relative weight loss H2O [ppm]

Relative weight loss CO2 [ppm]
700
700
600
600
500
500
400
400
300 300

200 200

100 100

0 0
0 50 100 150 200
Time [min]

Figure 16: Released carbon dioxide (, ) and water (Ο, ) as determined by in situ gas
chromatography (GC) during isothermal cure at 65°C. Open symbols reflect the
MY721/MA/HHPA-system and closed symbols denote the MY721/HHPA-system with a
molar anhydride/epoxy ratio, r=0.8.

The single heating method as proposed by Barrett [42], based on the following general
expression used in the analysis of dynamic DSC data, where the rate of conversion (dα/dt)
and the rate constant (k) lead to the following expression:
dα
= kf (α ) (2.3)
dt
At a given time during the DSC scan (t) and temperature (T), f(α) is a function of the
fractional conversion (α) and is assumed for a reaction of n-th order to be as follows:

f ( α ) = (1 − α )n (2.4)

The apparent rate constant (k) is assumed to have an Arrhenius dependence on

temperature, as given in eqn (2.5):
 − Ea 
k = A exp  (2.5)
 RT 
where A is the frequency factor, Ea the apparent activation energy, R the universal gas
constant and T the absolute temperature.
The combination of eqns (2.3) and (2.4) and transformation into the logarithmic form leads to
the following expression:
 ddtα 
 = ln(A ) − a
E
ln n 
(2.6)
 (1 − α )  RT
If the reaction order (n) is known Barrett [42] proposed a plot of the left-hand side of eqn
(2.6) against the inverse temperature, 1/T. The appropriate value for the apparent activation
energy can than be obtained by the linear regression.
24 Chapter 2

To calculate the reaction order (n), eqn (2.7) can be rearranged and the derivative taken
with respect to ln(1-α), yielding to the Freeman-Carroll relation in term of fractional
conversion [45]:
d ln (dα / dt ) E  d(1/ T ) 
=− a  d ln(1 − α )  + n
(2.7)
d ln (1 − α ) R  

By use of the Freeman-Carroll relationship (eqn (2.7)) the reaction order (n) and the
activation energy (Ea) was calculated. The kinetic parameters obtained for the Freeman-
Carroll method by linear regression for different constant heating rates are listed in Table 3.
The average values for the activation energy (Ea) and the reaction order (n) are
Ea=113.2±13.3 kJ/mole and n=2.0±0.2, respectively.
With the calculated reaction order of n=2.0, the Barrett relation (eqn (2.6)) was plotted as
shown in Figure 17. The linear regression yielded the activation energies (Ea) and the
frequency factors (ln A), which are tabulated in Table 3 as well. The average of the overall
activation energy and the frequency factor is Ea=112.6±5.3 kJ/mole and ln A=27.4±1.6,
respectively.

2
d(ln(da/dt))/d(ln(1-α)^n)

-1

-2

-3

-4
-3 -3 -3 -3 -3 -3
2.1x10 2.2x10 2.3x10 2.4x10 2.5x10 2.6x10
1/T [1/K]

Figure 17: Typical Barrett plot (eqn (2.6)) of the MY721/MA/HHPA-system with an anhydride
to epoxy ratio of r=0.8. Experimental data were obtained from a continuous heating scan
(ϕ=10°C/min).
 Epoxy-anhydride crosslinking reaction 25

Table 3: Reaction order, n, and activation energy, Ea, calculated using the Freeman-Carroll
relation eqn (2.3). The average value of the reaction order, n=2.0, was used to calculate the
activation energy and the frequency factor using the Barrett relation eqn (2.6).
Freeman-Carroll Relation Barrett Relation with n=2.0
Heating
rate, φ Activation Activation
Reaction Correlation Frequency Correlation
[°C/min] energy, Ea 2 energy, Ea 2
order, n coefficient, R factor, ln A coefficient, R
[kJ/mole] [kJ/mole]
20 2.3 135.1 0.9975 112.8 26.6 0.9964
15 1.9 100.9 0.9986 107.2 26.0 0.9996
10 1.9 114.4 0.9956 116.8 28.4 0.9998
5 1.8 111.0 0.9985 118.8 29.7 0.9999
2.5 2.0 104.5 0.8876 107.5 26.5 0.9994

The calculated values for the activation energies obtained by the Barrett- and the Freeman-
Carroll relations are consistent. This is not too surprising, since the Freeman-Carroll relation
(eqn (2.7)) is only the derivative with respect to ln(1-α) of the Barrett eqn (2.6). The relatively
high values of the obtained overall activation energies have been noted before [50, 51, 57].
The discrepancy can be probably attributed to the fact that a single pair of Arrhenius
parameters may not adequately describe the temperature dependence of the epoxy cure
rate. Furthermore, the two single-heating models are predicated on the assumption that
simple n-th order kinetic is able to describe the entire cure reaction of the investigated
epoxy-anhydride system. It is known that many systems show autocatalysis (described by
eqn below) and so the value of the above methods for analyzing a single-temperature-
scanned DSC curve is rather limited.
Other phenomenological multiple-heating methods, as originally proposed by Kissinger [44],
and by Ozawa [46], have shown that these methods are also effective for both n-th order
reaction and autocatalysis. Both models are fairly similar and based on the assumption that
the reaction rate at a constant extent of conversion is only a function of temperature (iso-
conversion approach) and therefore eqn. (2.8) applies:
 d ln( dα / dt )  E
 −1  =− a (2.8)
 dT  R
The Kissinger method describes the relation, that for a crosslinking reaction, the extent of
conversion at the peak temperature, Tm, is constant and independent of the heating rate (ϕ),
thus eqn (2.9) yields:
( )= −E
d ln Tφ2 (2.9)
( )
m a
1
d Tm
R
In contrast Ozawa’s approach yields a simple relation between the activation energy (Ea),
heating rate (ϕ), and iso-conversion temperature:
26 Chapter 2

R ∆ ln φ
Ea = − ⋅ (2.10)
1.052 ∆ T1
m

Since both methods are independent of the reaction order, n, the calculated activation
energies from iso-conversional models are able to provide the activation energy over the
entire course of reaction as shown in Figure 18.

75

Peak conversion
70
Ea [kJ/mol]

65

60

55
0.0 0.2 0.4 0.6 0.8 1.0
Conversion α [-]

Figure 18: Activation energy (Ea) as a function of fractional conversion (α) as calculated by
the Ozawa method eqn (2.10) for the MY721/MA/HHPA-system (r=0.8).

However, both theories reduce the complexity of a multi-step curing reaction to a single
reaction path by ignoring the reaction order, n, thus the calculated activation energies
Ea=63.5 kJ/mole (Kissinger) and Ea=67.3 kJ/mole (Ozawa), respectively can only be
considered as approximate values. Therefore, although the Ozawa and Kissinger methods
yield almost identical values, they are at variance with the values obtained by the Barrett
and Freeman-Carroll relations. Such deviation has been previously reported by Rogers and
Smith [51], in the calculation of the activation energy for the decomposition of hexahydro-
1,3,5-trinitro-s-triazine where a divergence of up to 46% between the methods was
observed.

2.3.2.2 Isothermal DSC

Isothermal DSC experiments have been used in several studies to determine various kinetic
parameters [1, 50, 55, 62-71]. To describe the curing kinetics of thermosetting materials two
general kinetic models, the n-th order and autocatalysis, can be applied. The curing kinetics
of a reaction with an autocatalytic and a noncatalytic contribution was for the first time
analytically described by Kamal and Sourour [65, 70]. Later this approach was modified by
 Epoxy-anhydride crosslinking reaction 27

Kenny et al. [66, 67, 72, 73], who developed and proposed the generalized mathematical
relation by considering the maximum possible conversion, αmax , as shown in eqn (2.11):
dα
dt
( )
= k( T ) f ( α ) = k 1 + k 2 α m ⋅ (α max − α )
n (2.11)

where m and n are the empirical power-law exponents, whose sum represents the overall
reaction order. In eqn (2.11) k1 indicates the reaction rate, which is independent of the
reaction advancement and is therefore related to noncatalytic rate constant as opposed to
k2, which is the autocatalytic rate constant. Furthermore, the specific reaction rates, k1 and
k2, are assumed to have an Arrhenius temperature dependence as shown in (2.12):
 Ea 
k i = A i exp − i  (2.12)
 RT 
where Eai and Ai denotes the activation energy and the frequency factor of the part
reactions, respectively, R the gas constant, and T the absolute temperature.
By a linear regression of ln ki versus 1/T the activation energies and frequency factors of the
noncatalytic and autocatalytic reactions can be calculated separately, and used to classify
the polymerization process. The modified Kamal approach is considered to be generally
valid and has been employed for a wide range of thermosetting systems [74].
In Figure 19 the conversion of the isothermal curing reactions at temperatures of 100, 110,
120, and140°C respectively is shown as a function of the curing time. At constant curing
temperatures below 100°C the reaction rate is very slow and a meaningful analysis of the
heat flow signal was not possible. The conversion, α, was calculated according to eqn (2.13)
and normalized to the total reaction enthalpy (∆Htot) obtained from extrapolated dynamic
scans with ϕ=0°C/min (see dynamic DSC-scans). After each isothermal run, the sample was
quenched to room temperature and afterwards reheated at 10°C/min to 250°C in order to
determine the residual heat of reaction, ∆HR.
∆Htot − ∆HR ∆H
α= = (2.13)
∆Htot ∆Htot
The absolute value of α is sensitive to the magnitude of ∆Htot, which as seen in Figure 15 is
not constant for the investigated system, but dependent on temperature scan rate (ϕ). This
dependence on ϕ is unusual, because the cure reaction is interfered by other exothermal
events. Thus the uncertainty in α is potentially larger than normally expected [1, 56]. To
minimize this general DSC limitation the extrapolation of total reaction enthalpy to
ϕ=0°C/min was performed by only using heating rates of 10, 15, and 20°C/min (see Figure
15 dotted line). At these relatively high heating rates the interference with the simultaneously
occurring decarboxylation of maleic anhydride was greatly minimized and the obtained ∆Htot
value of 301 J/g was used for further calculations. This obtained ∆Htot value is in the same
28 Chapter 2

order of magnitude than other reported values for anhydride cured epoxy-systems, ranging
from ∆Htot= 306 J/g [75] to ∆Htot= 389 J/g [55].

1.0
140 °C
120 °C
110 °C
0.8
100 °C

Conversion α [-]
0.6

0.4

0.2

0.0
0 15 30 45 60 75 90 105
Time [min]

Figure 19: Isothermal conversion, α, as a function of time at curing temperatures of 100,

110, 120, and 140°C, normalized with the total reaction enthalpy of ∆Htot = 301 J/g as
obtained form the extrapolated dynamic DSC scan to ϕ=0°C/min.

From the initial slope of the curves in Figure 19 it can be seen that the reaction is
progressing faster at higher temperatures, which is even more obvious in Figure 20 where
the rate of conversion, dα/dt, is shown as a function of conversion, α. The curves in
Figure 20 show a maximum in conversion rate, which is a typical indication for autocatalytic
reactions or reactions including an autocatalytic step [76].

0.18

0.16
140°C
Rate of conversion dα/dt [sec ]
-1

0.14

0.12

0.10

0.08

0.06 120°C
110°C
0.04

0.02 100°C
0.00
0.0 0.2 0.4 0.6 0.8 1.0
Conversion α

Figure 20: Rate of conversion, dα/dt, as a function of conversion, α, at curing temperatures.

The solid line represents the experiment and the dotted line the model relation described by
(2.11) with parameters obtained by least squares curve fit. The (♦) symbol indicates the
maximum rate of conversion. All curves are normalized with the total reaction enthalpy of
∆Htot= 301 J/g obtained form extrapolated dynamic DSC scans to ϕ=0°C/min.
 Epoxy-anhydride crosslinking reaction 29

The appropriate kinetic parameters were determined using the modified Kamal relation eqn
(2.11) by best curve fitting. A good agreement of the analytical model (dotted line) with the
experimental data (solid line) was observed (Figure 20). A constant extent of reaction at the
maximum rate of conversion was observed between 0.29<α<0.35. Moreover, it can be seen
that at early stages of cure, before gelation and vitrification occur, the reaction takes place in
the liquid phase and is controlled by chemical kinetics. The kinetic described by the modified
Kamal relation (eqn. (2.11)) coincides with the experimental data. As the reaction
progresses a deviation from the symmetric reaction rate curves appears, at conversions
above 0.6 to 0.7. This observation can be related to the change of the reaction mechanism
from a kinetic to a diffusion-controlled reaction due to the hindered mobility of reactive
groups [77].
An interpretation of the maximum in the curve on the chemical level is that the increase in
the reaction rate up to the maximum reaction rate is governed by the accelerating effect of
typical catalysts, e.g. tertiary amines, and is mainly described by the apparent reaction order
for the autocatalytic part (m in eqn (2.11)). Due to the absence of externally added catalysts
it is believed that the TGDDM-monomer might provide the tertiary amine sites on its own
structure (Table 2), which act as an internal catalyst for the anhydride ring. The decreasing
slope of the curve beyond the maximum is dominated by the consumption of the reactive
species of the mixture and is predominately influenced by the reaction order for the non-
catalytic reaction (n in eqn (2.11)).
The obtained values for the reaction orders and the reaction rates of the phenomenological
Kamal equation, m, n, k1 and k 2 , are plotted in Figure 21 with respect to the curing
temperature. The reaction order, n, increases linearly with the curing temperature whereas,
m, remains almost on a constant level. A similar dependency has been observed by
Mezzenga et al. [69] for an epoxy/amine system (DGEBA/isophorone diamine), whereas
other studies state that the overall reaction order (m+n) remains at a constant level of 2 [78],
depending on the epoxy/amine systems. The calculated arithmetic mean over the
investigated curing temperature range was n=1.04 and m=0.4.
As seen from Figure 21(b), k1 values are small compared with those obtained for k2
indicating that the latter reaction rate is the rate determining part reaction. The reaction rate
for the autocatalytic reaction, k2, increases with temperature, whereas the reaction rate
corresponding to the noncatalytic reaction, k1, remains almost unaffected by temperature.
30 Chapter 2

a)
1.2

n
1.0

reaction order (m, n)

0.8

0.6

m
0.4

0.2
90 100 110 120 130 140 150
Curing Temperature [°C]

-2 ln k2
Y =14.3 - 51.16 X

-4
ln(k1, k2)

-6 ln k1
Y =-7.6 - 0.04 X

-8

0.29 0.30 0.31 0.32 0.33 0.34

-1
1/RT*1000 [K ]

Figure 21: Dependence on curing temperature of (a) reaction exponents n and m and (b)
reaction rates k1 and k2 in the modified Kamal eqn (2.11) as obtained by best fitting of
experimental data. All curves are normalized with the total reaction enthalpy of ∆Htot= 301
J/g obtained form extrapolated dynamic DSC scans to ϕ=0°C/min.

The activation energy related to the cure reaction, obtained from eqn (2.12), is E1=0.04
kJ/mole and E2=51.2 kJ/mole and the corresponding logarithm of the frequency factor is
lnA1=-7.7 and lnA2=14.3, respectively. Herein, the subscripts ‘1’ and ‘2’ denote the
noncatalytic and the autocatalytic part reaction, respectively. The calculated activation
energy agrees reasonably well with other reported values for isothermal anhydride/epoxy
curing reactions that vary between 58 to 87 kJ/mol [20, 55, 56, 79, 80].
Even more obvious is the negligible contribution of the noncatalytic part during the complete
curing reaction in Figure 22. Here the catalytic and noncatalytic contributions from the
modified Kamal equation are separated and plotted in comparison to experimental results.
 Epoxy-anhydride crosslinking reaction 31

The observation that the overall reaction is dominated by the autocatalytic contribution is
consistent with studies by Montserrat et al. [81, 82]. They used a similar model to describe
the curing reaction of DGEBA and methyl-tetrahydrophthalic anhydride in combination with a
tertiary amine catalyst and also found that the noncatalytic contribution was neglible.

0.07 Experiment 120°C (solid line)

0.06
Rate of conversion dα/dt [sec ]
-1

0.05

0.04

0.03
Model: autocatalytic (dotted line)
0.02
Model: non-catalytic (dashed line)
0.01

0.00

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Conversion α

Figure 22: Conversion rate, dα/dt, as a function of conversion, α. Comparison of the

autocatalytic (dotted line) and non-catalytic (dashed line) model contributions according to
eqn (2.11) to the experimental results (solid line) measured for the MY721/MA/HHPA-
system (r=0.8) at 120°C

The nature of the pseudo-autocatalytic rate curve appearance is not fully understood. Kamal
and Chanda [68], and Ghaemy and Riahy [62] reported earlier an autocatalytic rate cure
profile for an anhydride/epoxy curing reaction. However, the uncatalyzed anhydride/epoxy
curing reaction as proposed by Fischer [25] and schematically shown in Figure 6, cannot be
strictly assigned to an autocatalytic reaction scheme. Thus, the formation of a pseudo-
autocatalysis conversion rate profile can only explained by the significant difference in the
appropriate reaction rate constants, k2>>k1 [83] from the involved reaction steps as shown in
reaction (V) to reaction (VI) Figure 6. Hence the reaction does not follow simple n-th order
kinetic profile straight from the beginning of the reaction.
Moreover, it can be stated that the kinetic parameters obtained from isothermal experiments
are inconsistent with the values obtained from the dynamic scans. Riccardi et al. [50]
observed similar differences between the isothermal and dynamic DSC measurements for a
diglycidyl ether/ethylenediamine-system. They explained the serious disagreement between
the kinetic parameters obtained from dynamic and the isothermal DSC measurements with a
change in the curing mechanism with varying curing temperature. At low temperatures, the
dynamic experiment is comparable with the conditions of isothermal DSC measurements,
where the autocatalytic reaction is dominant. This is also consistent with the fit of our data to
32 Chapter 2

the modified Kamal equation (Figure 21). At higher isothermal curing temperatures,
comparable to the dynamic DSC conditions, the noncatalytic part of the reaction becomes
more significant so that apparent n-th order kinetics is obtained. The kinetic models used for
dynamic DSC analysis do not separate the different kind of reaction mechanism and just
deal with general constants and reactions. The parameters obtained from dynamic-DSC
models have consequently no mechanistic meaning.

2.3.2.3 Temperature modulated differential scanning calorimetry (TMDSC)

As discussed in the sections before, conventional DSC does not clearly distinguish the
reaction rates of the two different used anhydrides in the co-anhydride mixture.
Consequently, an unambiguous identification of the network structure is limited, because the
signal obtained from conventional DSC was too low to separate possible sub-transitions
from the baseline. For this reason temperature modulated differential scanning calorimetry
(TMDSC) experiments were performed to study the thermal cure reaction (Figure 23(a)) and
the final network-structure (Figure 23(b)) of the MY721/MA/HHPA-system. Typical Fourier
deconvoluted TMDSC thermograms for a constant heating ramp at 1°C/min with underlying
oscillation amplitude of 0.2°C for a period of 60 seconds, are shown in Figure 23.

a)
0.5 0.0250
Total heat flow / non-rev. heat flow [W/g]

rev. heat flow

0.4
0.0225 reversing heat flow [W/g]

0.3

0.0200
0.2

total heat flow

0.1 0.0175

0.0
non-rev. heat flow
0.0150
0 50 100 150 200 250
Temperature [°C]
 Epoxy-anhydride crosslinking reaction 33

b)
0.25
0.8 Tg2

Total heat flow / non-rev. heat flow [W/g]

rev. heat flow
0.7
0.6
0.20

rev. heat flow [W/g]

0.5
0.4 Tg1
0.3 total heat flow
0.15
0.2
0.1
0.0
non-rev. heat flow 0.10
-0.1
-0.2
0 50 100 150 200
Temperature [°C]

Figure 23: Deconvoluted temperature modulated DSC-thermogram with total heat flow (ˆ),
reversing heat flow (Š) and non-reversing heat flow (‰) for (a) unreacted and (b) totally
cured MY721/MA/HHPA-system (r=0.8).

For the uncured MY721/MA/HHPA-system the non-reversing heat flow component (Figure
23(a)) clearly indicates that MA and HHPA obey different reaction rates during the
crosslinking reaction. Conventional DSC-experiments of the pure anhydride/epoxy mixtures
(cf. Figure 13) have confirmed that the low reaction peak temperature (T~85°C) can be
attributed to the MY721/MA reaction, and the high reaction peak temperature (T∼130°C) to
the MY721/HHPA reaction, respectively.
After cooling the sample back to room temperature the same sample was subjected to a
second TMDSC-run under comparable conditions. The total heat flow and its reversing and
non-reversing components are shown in (Figure 23(b)). The glass transition temperatures
could not be accurately determined from the total heat flow curve, since reversing (glass
transition) and non-reversing (enthalpic relaxation) events significantly overlap. However,
upon deconvolution of the signal into the appropriate reversing heat flow and non-reversing
heat flow, two Tg’s (Tg1∼105°C and Tg2∼220°C) can be determined from the reversing heat
flow component. This observation provides reliable indication that the studied
MY721/MA/HHPA-system consists of a heterogeneous network structure. Due to the
different used anhydrides (co-anhydride) it is proposed that at least two independent sub-
networks consisting of MY721/MA and MY721/HHPA are present in the final structure.
34 Chapter 2

2.3.2.4 Relationship between glass transition temperature and conversion

The variation of the glass transition temperature (Tg) with the extent of reaction (α) of the
MY721/MA/HHPA-system was determined, in order to establish a full analytical expression
of the reaction kinetics when it becomes diffusion controlled on vitrification. It has been
demonstrated [84, 85] that a unique relationship between Tg and α is valid when the curing
reaction involves the formation of predominately one type of chemical linkages or when
competing reactions yielding different structures have similar activation energies or occur
sequentially. Considering these fundamental assumptions DiBenedetto [86], and Nielsen
[87], proposed a theoretical model based on the lattice energy and the segmental mobility of
the involved species to predict the glass transition temperature as function of conversion as
given in eqn (2.14):
Tg − Tg0 (E x / E m − Fx / Fm ) α gel
= (2.14)
Tg0 1 − (1 − Fx / Fm ) α gel
Where Tg0 is the glass transition temperature of the unreacted system (α=0), Fx/Fm is
corresponding to the ratio of segmental mobility. The subscripts ‘m’ and ‘x’ denote the
unreacted and crosslinked states, respectively. The ratio of the energies associated with an
element in the lattice space for the crosslinked and un-crosslinked resins (Ex/Em) can be
approximated [86] by eqn (2.15):
2
Ex d (M0 ) x  δ x  (2.15)
= m  
E m d x (M0 )m  δ m 
Where d represents the density, M0 is the molecular weight of the constituent repeating unit,
and δ is the solubility parameter. Later Enns and Gillham [88] demonstrated that a sufficient
agreement between a curve fit and numerous experimental data could be obtained when
both Fx/Fm and Ex/Em parameter were determined as adjustable values. From eqn (2.14) the
ratio of both parameters is then given by eqn (2.16):
E x / E m Tg∞
= (2.16)
Fx / Fm Tg0
Where Tg∞ is defined as the maximum glass transition temperature, obtained at the ultimate
conversion (αmax).
Introducing eqn (2.14) in eqn (2.16), and substituting the ratio of sequential mobility, Fx/Fm by
the constant λ, results in eqn (2.17), which is referred as the modified DiBenedetto equation:
Tg − Tg0 λα
= (2.17)
Tg∞ − Tg0 1 − (1 − λ )α
Where λ must be considered as an adjustable parameter with a value between 0 and 1. Tg
values of the MY721/MA/HHPA-system cured isothermally for various periods of times and a
comparison of the experimentally data of the MY721/MA/HHPA-system with eqn (2.14)
using λ=0.193 obtained from nonlinear least square fit are given in Figure 24. The non-
 Epoxy-anhydride crosslinking reaction 35

linearity of the Tg behaviour as a function of conversion is in agreement with earlier work on

thermosetting materials [89, 90]. Moreover, the general non-linear shape of the resulting
curve shows that the relative increase of the glass transition temperature is greater at higher
extents of conversion. This fact is consistent with the steep increase in the crosslink density
beyond the gel point.

250

200
λ=1

150
Tg [°C]

100

50

0 λ=0

-50
0.0 0.2 0.4 0.6 0.8 1.0
Conversion [-]

Figure 24: Glass transition temperature (Tg) as a function of the degree of conversion (α), for
the MY721/MA/HHPA-system (r=0.8) as determined by dynamic DSC (ϕ=10°C/min) on
partially isothermal cure samples. The solid line represents the best fit of data to the
modified DiBenedetto eqn (2.17).

Consequently, if one can assume the validity of the relationship between the glass transition
temperature and the conversion the studied MY721/MA/HHPA-system can be expressed as:
45.626α
Tg = − 20 (2.18)
1 − 0.807α
or
Tg + 20
α= (2.19)
61,78 + Tg 0.807

2.3.3 Curing reaction studied by chemo-rheology

The successful processing of thermosetting materials requires accurate knowledge and
precise control of the viscosity development during the crosslinking reaction. Especially the
changes in viscosity during the fluid to solid transformation, at a certain degree of
conversion, determine many aspects of the processing, e.g. equipment, temperature window
and deformation regime of processing.
As expected, the initial complex shear viscosity (η*0), directly after mixing and before cure,
was reduced from the initial 130 mPas to 35 mPas with increasing cure temperature from
36 Chapter 2

40°C to 100°C. The obtained data agree reasonably well with an Arrhenius temperature-
dependent viscosity as depicted in Figure 25 and η*0 can be described as:

 3.21x10 3 
η* 0 (T ) = 4.18 x 10 −6 exp 
 [Pas] (2.20)
 RT 

-1.5

-2.0

-2.5
ln (η 0, Pas)
∗

-3.0

-3.5
-6 3
η0= 4.18 x10 exp(3.21x10 /RT)

-4.0
2.6 2.8 3.0 3.2
-1
1000/Temperature [K ]

Figure 25: Arrhenius plot of the initial complex shear viscosity (η*0) as a function of
temperature for the MY721/MA/HHPA-system (r=0.8).

The prediction of the onset of the gelation phenomenon by dynamic rheometry has been the
subject of many studies in the past. For the investigated epoxy systems, the crossover point
between the storage (G’) and the loss (G’’) moduli (tanδ =1), as originally introduced by
Tung and Dynes [91] and widely established as a reliable method [92-97], has been used
here as a criterion to determine the gel-point. In this work the effect of the frequency was not
investigated, because satisfactory information on the thermal cure behaviour of the system
was obtained by comparing the results with respect to the control system at a fixed
frequency value. An example of an isothermal cure of the MY721/MA/HHPA-system at a
curing temperature of 40°C is shown in Figure 26.
 Epoxy-anhydride crosslinking reaction 37

6
10

G' [Pa]
10
4 G'' [Pa] 10

Dynamic moduli G', G'' [Pa]

tan delta

2
10

Loss factor
1

0
10
0.1
-2
10

-4
0.01
10
0 2000 4000 6000 8000
Time [sec]

Figure 26: Dynamic rheological spectra of storage modulus (G’), loss modulus (G’’), and loss
factor (tanδ) during isothermal cure of the MY721/MA/HHPA-system (r=0.8) at 75°C to
determine the time of gelation at G’=G’’

A better view over the overall cure process can be seen in the plot of the complex shear
viscosity (η*), the storage modulus (G’), and the loss modulus (G’’) as a function of reaction
time and isothermal cure temperatures for the MY721/MA/HHPA-system as illustrated in
Figure 27. As the reaction starts the storage modulus (G’) rapidly increases whereas the
loss factor (tanδ) shows a reverse trend near the gel point. Consequently, the time to
gelation decreased significantly with increasing temperature, due to an increasing reaction
rate of the epoxy system.
In order to determine the extent of conversion at the point of gelation (αgel) isothermal FT-
Raman experiments (cf. chapter 2.2.8) were combined with the rheological experiments.
Figure 28 shows the complex viscosity versus conversion data for the MY721/MA/HHPA-
system during an isothermal cure at 75°C. The extent of conversion at the point of gelation,
as marked by an arrow, was obtained at αgel~0.68. Assuming that the epoxy group and the
cyclic anhydride are both bi-functional [34, 35, 98] and that the conversion of the
functionalities is a random reaction, general mean field theories [99-101] allow the prediction
of the theoretical extent of conversion at the gel point between the tetra-functional epoxide
(TGDDM) with an mono-anhydride, resulting in a value that approaches αgel~0.43. The
experimentally determined conversion level at gelation is significantly higher than the
theoretical prediction (or theoretical gel times are significantly lower). However, higher
experimental values have been previously reported in the literature [88] emphasizing the
limitations of the underlying assumptions of the theoretical model and the complexity in the
reaction chemistry in anhydride-cured systems compared to amines-cured resins.
38 Chapter 2

5
a) 10
50°C 30°C
4
10 70°C
90°C

Complex Viscosity [Pa*s]

110°C
3
10

2
10

1
10

0
10

-1
10

2 3 4 5
1x10 1x10 1x10 1x10
Time [sec]

Time [h]
5
0.28 2.78 27.78
b) 10

4
10
110°C
Storage modulus [Pa]

3 90°C
10
70°C
2
10 50°C
1 30°C
10

0
10

-1
10

-2
10
3 4 5
10 10 10
Time [sec]

Time [h]
0.28 2.78 27.78
c) 3
10

2
Loss modulus [Pa]

10 30°C
50°C
70°C
1
10 90°C
110°C

0
10

3 4 5
10 10 10
Time [sec]

Figure 27: Dynamic rheological spectra versus time during isothermal cure from 30°C to
120°C of the MY721/MA/HHPA-system (r=0.8). (a) Dynamic shear viscosity (η*) the dots
indicate the time to gelation (tgel); (b) dynamic storage modulus (G’); and (c) dynamic loss
modulus (G’’)
 Epoxy-anhydride crosslinking reaction 39

100000

10000

Complex viscosity [Pa*s] 1000

αgel =0.675
100

10

0.1

0.01
0.0 0.2 0.4 0.6 0.8 1.0
Conversion α

Figure 28: Complex shear viscosity (η*) as a function of extent of conversion (α) for the
MY721/MA/HHPA-system (r=0.8) at an isothermal cure at 75°C

The apparent activation energy (Ea) for the crosslinking reaction can be calculated from the
knowledge of the gel time at different temperatures. An increase in conversion in the region
where the crosslinking is kinetically controlled can be expressed by a constitutive equation
as:
dα  − Ea 
= A (rheo ) f (α ) exp  (2.21)
dt  RT 
Integrating eqn (2.21) at constant temperature and taking natural logarithms leads to eqn
(2.22):
dα E α gel

f (α )
ln (1 / t gel ) = ln
+ ln A (rheo ) − a
RT ∫
0
(2.22)

where Ea is the crosslinking activation energy, tgel is the time to gelation, R is the universal
gas constant, and T is the cure temperature. Since the time to gelation represents a specific
extent, and the gel-time is inversely related to the rate of reaction, the gel-time-temperature
profile follows an Arrhenius relationship, where the cure is represented by eqn (2.23):
Ea
ln (1 / t gel ) = ln A (rheo ) −
(2.23)
RT
There is a linear relationship between ln(1/tgel) and the inverse of temperature for an
isothermal curing reaction. From the slope of this relationship the apparent activation
energy, Ea, of the reaction can be calculated. Figure 29 shows ln(1/tgel), obtained from the
point where tanδ=1, as a function of the reciprocal temperature, giving: ln A(rheo)=11.5 and
Ea=58.7±1.9 kJ/mole.
40 Chapter 2

-7

-8

-9

ln (1/tgel)
-10

-11
Y =11.49 - 7069 X
-12
Ea= 58.7 kJ/mol
-13
0.0026 0.0028 0.0030 0.0032 0.0034
-1
1/T [K ]

Figure 29: Arrhenius plot of ln(1/tgel) as a function of inverse temperature (1/T) for the
MY721/MA/HHPA-system (r=0.8) for determination of the activation energy for gelation (Ea).

The value for the overall activation energy obtained by rheological measurements agrees
reasonably well with that from studies of the chemical reactions measured by isothermal-
DSC as discussed under section 2.3.2. The small differences between absolute values
might be caused by different thermal temperature gradients, related to the different sample
sizes and the different instrumental configurations.

2.3.4 Curing reaction studied by FT-Raman spectroscopy

The objective of this section was to study the low temperature curing behaviour of the same
epoxy/co-anhydride mixture by means of vibrational FT-Raman spectroscopy under
isothermal conditions. As a complementary technique to mid infrared spectroscopy, FT-
Raman is able to provide fundamental vibrational information. Based on the different
selection rules of these two spectroscopic techniques, polar bonds tend to yield strong IR
and weak Raman bands, whereas non-polar groups give rise to strong Raman and weak IR
bands. FT-Raman also allows one to monitor the curing reaction over the complete
conversion range, although the material undergoes several state transitions during curing,
i.e. gelation and vitrification. These transitions have only a limited influence on the
vibrational spectrum. The quantitative aspects of FT-Raman spectroscopy permit quasi real
time in situ studies of the crosslinking reaction and the direct calculation of the concentration
profiles of various reactive species. Further advantages compared to mid-IR are the lack of
interferences from water vapor and glass, which is very convenient because glass sample
containers can be used for the experiments.
 Epoxy-anhydride crosslinking reaction 41

Conversion of the functional groups is followed by means of real-time FT-Raman

spectroscopy and the results are validated against empirical relationships describing the
conversion versus time as a function of cure temperature. Kinetic analysis from the
conversion of the functional groups is used to obtain the empirical rate curves. Moreover, it
is attempted to propose a reaction mechanism for the uncommon cure behaviour of the
uncatalyzed MY721/MA/HHPA-system

2.3.4.1 Functional group analysis

The spectra of the unreacted MY721 epoxy monomer as well as the two carboxylic acid
anhydrides can be seen in Figure 30. Detailed peak assignments of the main absorption
bands are listed in Table 4.

Table 4: Raman spectral peak assignments of (a) carboxylic acid anhydrides and (b) epoxy
resin [102-107]

This image is not available online. Please consult the hardcopy thesis available
from the QUT Library.

The band assignments were made using standard spectroscopy literature [102-105] and by
considering results of previous Raman studies [106, 107]. In the spectrum of MY721 five
bands attributable to the epoxy ring are easily observed. These are located at 3072 cm-1 [νas
(CH2)] and 3008 cm-1 [νs(CH2)] due to C-H groups of the epoxy ring and at 1260 cm-1
[breathing of the epoxide ring], 902 cm-1 [νas of epoxide ring], and 866 cm-1 [νs of epoxide
ring] due to vibrations of the entire epoxy ring. In the spectra of the crystalline anhydrides
the bands of major interest are those due to the vibrations of cyclic anhydride groups, such
42 Chapter 2

as the intense symmetric and antisymmetric stretch of the carbonyls at 1840 and 1780 cm-1,
respectively. This is because the anhydride group is the reactive center and the bands
would be expected to decrease during curing with the epoxy.
Real-time FT-Raman spectroscopy was performed to monitor the curing process. It was
assumed and later confirmed by the experiments that the aromatic hydrocarbon groups,
corresponding to the characteristic vibration at 1615 cm-1 [aromatic quadrant stretch], are
not involved in the curing reaction and hence the peak area does not change over the entire
curing period. Therefore, the invariant peak at 1615 cm-1 was used for all spectra as an
internal normalization standard [106, 107].
From the recorded spectra, the fractional consumption of functional groups,α, can be
calculated by eqn (2.24):
 I( t ) I (0 ) 
α(t ) = 1 −  1615 / 
I1615 
(2.24)
I
where Io, It, are the relative intensities (peak height) of reactive species at time zero, and
after a certain time interval, t, of the cure cycle. Due to the epoxy excess (r<1), the
concentration curves of epoxy and anhydride groups can be directly compared only if we
consider the absolute conversion. In terms of spectroscopic parameters, the absolute
conversion is given by C0α(t), where C0 refers to the initial concentration of the reactive
species.
Curing data obtained after different reaction times and reaction temperatures are shown for
the MY721/MA/HHPA-system with a molar anhydride to epoxy ratio, r=0.8, at 65°C in Figure
31 and Figure 32. Interferences from thermal emission bands or fluorescence, as reported
by Walton and Williams [107], were not obtained. Subsequently, the spectral changes were
monitored quantitatively and detailed results are discussed in the following sections.
 Epoxy-anhydride crosslinking reaction 43

a)

Raman intensity [a. u.]

b)

c)

3500 3000 2500 2000 1500 1000

-1
Raman shift [cm ]

Figure 30: FT-Raman spectra of unreacted anhydride hardeners (a) hexahydrophthalic

anhydride; (b) maleic anhydride and (c) epoxy resin AralditeMY721

t = 0 m in
Raman intesity [a. u.]

t = 1 8 0 m in

p o s tc u re d 1 h @ 1 2 0 °C

3000 2500 2000 1500 1000

-1
R a m a n s h ift [c m ]

Figure 31: FT-Raman spectra of MY721/MA/HHPA at various extents of cure at 65°C and
after 1 hour post-curing at 120°C
44 Chapter 2

30

formation

1733
20

1643
2959
3054
10

0
consumption

3004
-10

3114

868
1067
1260
-20

1590

636
-30

-40

-50

1847
-60
3200 2800 2400 2000 1600 1200 800
-1
Wavenumber [cm ]

Figure 32: Difference of FT-Raman spectra obtained by spectrum “after 1hour at 100°C”
minus spectrum “uncured” directly after mixing of the MY721/MA/HHPA-system. Changes in
peaks marked with a symbol are discussed in the text.

Anhydride:
The peaks from the mixture at 3111, 1850, 1780, 1060, 868, and 636 cm-1, respectively, can
be assigned to the two anhydrides used in the co-anhydride mixture. Peaks corresponding
to the individual MA and HHPA components are not reliable due to peak overlap. Hence,
although a clear band due to MA is seen at 3111 cm-1, in the presence of the epoxy, the
respective functional groups consumption of the two anhydrides, hexahydrophthalic
anhydride and maleic anhydride, cannot be further differentiated by FT-Raman
spectroscopy. However, it can be expected that both anhydrides compete for the reaction
with the epoxy groups as shown in Figure 6, even when the two anhydrides show similar
reactivity as reported in [108] for the same system.
All anhydride functional group peak intensities decrease progressively throughout the entire
cure. The total conversion of the functional anhydride groups can be gained using the typical
in-phase (symmetric) and out-of-phase (anti-symmetric) carbonyl stretching vibrations at
1850 and 1780 cm-1 [νs(C=O) and νas(C=O)], and 1060 cm-1 [νas(C-O-C)], respectively. In
this study the reaction of anhydrides was followed by means of the 1850 cm-1 [νs(C=O) of
the anhydride ring] as proposed in the past [24] and the results are illustrated in Figure 33(a)
and (b). To convert peak areas for the evolution of anhydride concentration, [Anh], as shown
in Figure 33(b), the initial anhydride concentration of 3.89 mol/kg was considered. It can be
seen that the anhydride peaks have not fully extinguished at lower curing temperatures and
 Epoxy-anhydride crosslinking reaction 45

hence the anhydride concentration curve stabilized on a quasi-stationary plateau after a

certain time period. This observation can be explained by the fact that the anhydride
consumption is limited due to a physical liquid-to-solid transformation of the material. From
this point, the reaction becomes diffusion controlled and a significant decrease of the
reaction rate is observed; see Figure 33(b). To reach a complete conversion of the reactive
groups isothermal curing temperature above 75°C or extended post curing steps are
necessary.

a) 70 b)

60 1 4

Anhydride concentration [mol*kg ]

-1
50
Raman intensity [a. u.]

3
40 2

30 2
3
20
1
4 1
10 2
3

0 4
0
1900 1890 1880 1870 1860 1850 1840 1830 1820 0 200 400 600 800 1000 1200
-1
Raman shift [cm ] Time [min]
Raman intensity [a.u.]

c)
Cu

500
rin

3 400
gt

300
1
im

200
e[

4 100
mi

2
n]

0
1800

1750

1700

1650

1600

1550

-1
Wavenumber [cm ]

Figure 33: (a) FT-Raman spectra of the anhydride peak [1850cm-1] at (1) 0, (2) 30, (3) 60,
(4) 120 min of cure at 75°C; (b) Plot of anhydride group concentration versus cure time at (1)
55°C, (2) 65°C, (3) 75°C, (4) 100°C; (c) FT-Raman spectra as a three-dimensional plot of
the intensity versus time for polymerization temperature of 65°C

In particular the maleic anhydride ring opening reaction and the related changing
substitution pattern can be followed by the disappearance of the characteristic C=C
vibrations from maleic anhydride 1590 cm-1 [ν(C=C)] as shown in Figure 34(a) and Figure
34(b). The simultaneously resulting formation of the appropriate acid group can be clearly
46 Chapter 2

seen by the broad growing intensity situated at 1640 cm-1 [ν(C=COOH)] as shown in Figure
34(a) and Figure 34(c). This observation provides ample evidence that the unsaturation of
maleic anhydride is not reactive at low temperatures.

a) b)
70

60 2.5

C=C concentration [mol*kg ]

-1
Raman intensity [a. u.]

50 2.0

40 1.5
1
30 4
1.0 3
2
20 3 2
0.5 1
3
10
2 4
1 0.0
0
1660 1640 1620 1600 1580
0 100 200 300 400 500 600
-1
Raman shift [cm ] Time [min]

c)

100
Extent of reaction (1640 cm ), [%])

3
80
-1

60

1
40

20

0
0 200 400 600 800
Time [min]

Figure 34: (a) FT-Raman spectra at (1) 0, (2) 30, (3) 60, (4) 120 min of cure at 75°C. (b) plot
of MA [νC=C; 1590 cm-1] fractional group consumption and (c) new formed acid
[ν(C=COOH); 1640 cm-1] peak versus cure time at as a function of isothermal curing
temperature (1) 55°C, (2) 65°C, (3) 75°C.

Epoxide:
A typical decrease of peak intensities corresponding to the epoxy group consumption at
3072, 3008 and 1260 cm-1 can be observed during the entire curing reaction. The epoxide
ring breathing vibration at 1260 cm-1 is well-resolved and essentially Lorentzian shape, as
seen in Figure 35(a) and (c). Hence this peak was utilized to monitor the epoxy conversion
as proposed previously [106, 109]. To enable peak height measurements a baseline
correction was performed and an initial epoxide concentration of 4,72 mol/kg was
considered to create the epoxide group concentration profile. A plot of the epoxide group
concentration, [EP], over time for different temperatures is given in Figure 35(b). A sigmoidal
curve is obtained for all curing temperatures. It can be seen that at curing temperatures
 Epoxy-anhydride crosslinking reaction 47

below 100°C the crosslinking reaction is quenched by a liquid to solid transition before it can
reach the 100% conversion. This behaviour is consistent with the observed anhydride
consumption behaviour. At an isothermal curing temperature of 55°C almost 68% of epoxy
groups were converted, while during a curing at 75°C, 76% of epoxide groups were
converted. However, it can be expected that the polyesterification reaction will continue also
into the glassy state, but at a significantly decreased reaction rate.

a) 40 b)

Epoxy concentration [mol*kg ]

-1
1 4
Raman intensity [a.u.]

30 2
3

3 1
2
2
20 4
1 3

4
0
1290 1280 1270 1260 1250 1240 1230 1220 0 200 400 600 800 1000 1200
-1 Time [min]
Raman shift [cm ]

c)
Raman intensity [a.u.]

10

2
Cu

500
rin

400
gt

300
im

200
e[
mi

100
n]

0
1280

1270

1260

1250

1240

1230

1220

-1
Wavenumber [cm ]

Figure 35: (a) FT-Raman spectra of the epoxy peak [1260 cm-1] at (1) 0, (2) 30, (3) 60, (4)
120 min of cure at 75°C; (b) plot of epoxy group concentration versus cure time at (1) 55°C,
(2) 65°C, (3) 75°C, (4) 100°C; (c) FT-Raman spectra as a three-dimensional plot of the
intensity versus time for polymerization temperature of 65°C

Ester group:
The relative increase of bands due to ester group formation upon cure can be observed at
2963 cm-1 [νas(CH2) adjacent to the ester group], and 1730 cm-1 [ν(C=O) aromatic ester].
The latter total aromatic ester peak contains both monoester and diester contributions and
was used to determine the extent of curing reaction according to eqn (2.25) as:
48 Chapter 2

I1730 (t ) I1730 ( 0 ) (2.25)

1615
− 1615 ) (
I I
β = 1730
I (∞ ) I1730 (0 )
( 1615 − 1615 )
I I
where β is the extent of the esterification reaction, It, the Raman intensity at time, t, and I0
corresponds to the intensity at t=0, and I∞ is the intensity at time of curing completion, t=∞. It
was assumed that the ultimate conversion (t=∞) of reactive groups was reached after a
post-curing step at 150°C for 1 hour, which defines 100% cure.
The total formation of aromatic ester groups (β) during the curing reaction at different
temperatures is given in Figure 36(b). An extended delay before start of the ester formation
as reported by Stevens [83, 110] was not observed. Besides, it can be seen that ester
formation approaches a limiting value at long curing times, which is in agreement with the
anhydride and epoxy consumption behaviour.

a) b)
100 4
24
22 4 3
20 80
Extent of reaction [%]
Raman intensity [a. u.]

18 3 2
16 60
1
14
40
12
2
10
20
8
6 1
0
4 0 200 400 600 800
1770 1760 1750 1740 1730 1720 1710 1700 1690
-1
Time [min]
Raman shift [cm ]

c)
Raman intensity [a.u.]
Cu
rin

400
gt
im

200
e[
mi
n]

0
3200

3150

3100

3050

3000

2950

2900

2850

2800

-1
Wavenumber [cm ]

Figure 36: (a) FT-Raman spectra of aromatic ester peak [1732 cm-1] at (1) 0, (2) 30, (3) 60,
(4) 120 min of cure at 75°C; (b) Plot of extent of ester group reaction (β) as a function of
cure time at (1) 55°C, (2) 65°C, (3) 75°C, (4) 100°C; (c) FT-Raman spectra as a three-
dimensional plot of the intensity versus time for polymerization temperature of 65°C
 Epoxy-anhydride crosslinking reaction 49

Side reactions:
The disappearance of the characteristic C=C vibrations from maleic anhydride at 3111 cm-1
[ν (=C-H)], 1590 cm-1 [ν (C=C)], and 636 cm-1 [γ (=CH-H)] is clearly visible in Figure 32.
Detail of the region from 1670 cm-1 to 1570 cm-1 is shown in Fig. 9(a) and the decrease in
the C=C band intensity of MA at 1590 cm-1 and an increase in a broad band that first
appears as a shoulder around 1640 cm-1 are shown in Fig. 9(b) and 9(c) respectively. This
could suggest that the unsaturation of maleic anhydride is being consumed at low
temperatures. If so, then it could provide ample evidence for the reaction of maleic
anhydride by anionic initiated homopolymerization through charge-transfer reactive
intermediates zwitterions as reported by Zweifel et al. [111] and Ricca et al. [59] and shown
in Figure 37 (reaction XVI to VXIII). Maleic anhydride is difficult to homopolymerize by a
conventional free radical or anionic initiator, but it can be easily initiated by nitrogen bases to
yield poly(maleic anhydride). The monomer is an excellent organic electron acceptor and
hence it forms zwitterions with a suitable base. In the investigated system the zwitterions is
spontaneously formed in the presence of nitrogen bases (TGDDM monomer) at room
temperature and appear to be the true initiator for poly(maleic anhydride) as shown in Figure
37. Additional indication for the zwitterions formation is given by the immediate color change
of the mixture, from light yellow to dark red, as soon as maleic anhydride was added. This
amine activated color change of maleic anhydride has been observed earlier for maleic
functional copolymers [53, 112] and was related to the formation of zwitterion intermediates.

This image is not available online. Please consult the hardcopy thesis
available from the QUT Library.

Figure 37: Polymerization mechanisms of maleic anhydride in the presence of tertiary amine
as proposed by Zweifel et al. [111].
50 Chapter 2

However, a more detailed analysis of this spectral region suggests that the changes may be
linked to the primary esterification reaction and result from changes to conjugation following
ring opening of MA. The evidence for this may be seen from an analysis of the extended
spectral region shown in Figure 38.

1.0 0min
20min
60min
0.8 120min
Intensity [a.u.]

0.6

0.4

0.2

0.0

1500 1600 1700 1800 1900 2000

-1
Wavenumber [cm ]

Figure 38: Expanded region of FT-Raman spectrum during the cure of MY721 with
MA/HHPA at 75°C showing isosbestic point at 1760 cm-1.

This is evidenced by the isosbestic point in the spectrum at 1760 cm-1 that supports the
smooth conversion of anhydride (1780 cm-1) to ester (1734 cm-1) that parallels the changes
in the MA band at 1590 cm-1 and the growth of the broad band near 1640 cm-1. This is
shown in Figure 39 in which the intensities of the four bands are plotted during cure. They
appear to all follow the same kinetic relationship so that the change in the band at
1590 cm-1 is most probably a result of the ring opening reaction of the MA producing a
change in the conjugation of the C=C rather than its actual consumption. The broadening of
the band at around 1640 cm-1 most probably reflects the range of environments that the
C=C encounters after ring opening as cure takes place.
 Epoxy-anhydride crosslinking reaction 51

Normalised Band Intensities of various deconvolved bands

1.2

Normalised Band Intensity

1.0

0.8
A1640
0.6 A1734
A1850
0.4
A1590
A1260
0.2

0.0

-0.2
0 50 100 150 200 250
Time (min)

Figure 39: Plot of band areas from indicating the similarity in the rates of disappearance of
the bands at 1850, 1590 and 1260 cm-1 and the rate of appearance of the bands at 1640
and 1730 cm-1.

The absence of a broad band at 1080 cm-1 [ν(C-O)] shows that no ether groups were
formed, even at curing temperatures up to 150°C, supporting the statement that
polyetherification, as illustrated in Figure 5 (reaction IV), is effectively suppressed. This is in
accordance with earlier results obtained by Tanaka and Kakiuchi [20], Fischer [25], and
Luston et al. [30].
Additionally, we were unable to observe changes of the peak position and intensity at
1615 cm-1 [aromatic quadrant stretch] of FT-Raman spectra that supports the hypothesis
that a “back-biting” cyclization of chlorohydrin impurities in MY721 at higher curing
temperatures as reported by St. John et al. [106, 111] does not occur.

2.3.4.2 Rate curves and kinetics analysis

Empirical rate curves were generated for the different reaction species by applying a
weighted smoothing function using a minimum number of iterations to obtain a curve. This
procedure ensured that the instantaneous reaction rates obtained from the concentration
profiles were not affected by single data point scattering.
To study this epoxy/anhydride reaction the empirical rate curves were calculated by taking
the first derivative of the epoxy concentration versus time data (Figure 33) as shown in
Figure 40(a). The obtained rate curves have a pronounced maximum which, in the reaction
scheme, may represent the development of the steady state concentration of the reactive
intermediates. This behaviour is more pronounced at lower isothermal curing temperatures.
Assuming a non-catalyzed mechanism this induction period is reflecting the generation of
intermediate COOH and OH-groups. While, considering a catalyzed reaction mechanism the
52 Chapter 2

presence of an induction period can be attributed to a rather slow initiation step when
polymerizing at low temperatures.
An alternative explanation for the maximum in the rate curve, Figure 40(a), is that it indicates
the reaction is autocatalytic in nature. Autocatalysis is observed when the tertiary amine
structure of the TGDDM-monomer acts as catalyst in epoxy/anhydride curing reaction. A
further possible source of any autocatalytic contribution could be reaction products between
MA and the aminoglycidyl monomer, as discussed before, and schematically shown in
Figure 37. These reaction products, especially the proposed formation of carboxylate
anions, could act as an additional initiator in the curing reaction.
Despite, the catalyzed reaction mechanism has obviously an autocatalytic nature; isothermal
kinetic runs have been analyzed with an excellent agreement by first-order kinetics,
particularly under isothermal curing conditions [66]. To account for this observation, Riccardi
and co-workers [118] developed and experimentally validated a mechanistic model that
predicts a chainwise polymerization proceeding by an anionic mechanism
(epoxy/anhydride/tertiary amine) an induction period followed by first-order kinetics when
using low isothermal cure conditions. Therefore, it was considered here that it is not
necessary to invoke autocatalysis in order to explain the observed concentration profiles
obtained from spectral data.
The measured data for the consumption of the epoxy groups are shown in a linearized first-
order kinetic plot in Figure 40(b). It can be seen that the reaction obeys first order behaviour
[(1-α)1)]. For an extent of conversion of around 70%, a significant drop in the experimental
reaction rate is observed due to diffusional restrictions indicating the change of the reaction
mechanism from kinetic to diffusional control. Therefore, the apparent first-order rate
constants were calculated as a function of temperature from the initial slopes and are listed
in Table 5.
Assuming the Arrhenius dependency of temperature on the rate constant, as depicted in
Figure 40(c), an activation energy of 71.01kJ/mol and a frequency factor of 19.85 min-1 was
calculated, respectively. This obtained activation energy is in excellent agreement with other
studies where the activation energy for base catalyzed anhydride/epoxy curing reactions is
reported in the range of 70-75 kJ/mole [66, 118].
 Epoxy-anhydride crosslinking reaction 53

a) 1.6

1.4

1.2

dα/dt x 10 [sec ]
-1
1.0

0.8
4

0.6

0.4

0.2

0.0

0 5000 10000 15000 20000

Time [sec]

b)
1.4
71%
1.2

1.0 70%

0.8
-ln(1-α)

0.6 67%

0.4

0.2

0.0

4 4 4 4 4 4 4
0 1x10 2x10 3x10 4x10 5x10 6x10 7x10
Time [sec]

c)

-7

-8
ln (1/tdev) [sec ]
-1

-9

Y = 15.76 -71.01 X
-10 2
R =0.9949

0.32 0.33 0.34 0.35 0.36 0.37

-1
1/(TR)*1000 [K ]

Figure 40: (a) Empirical epoxide reaction rate –d[EP]/dt as a function of curing time; (b) first
order kinetic regression plot of epoxide reaction, for 55°C (1), 65°C(), 75°C (+) as
measured by FT-Raman; (c) Arrhenius plot of the apparent first order epoxy reaction rate
constant, k2
54 Chapter 2

Table 5: First order rate constants and time of deviation from first order kinetic (tdev) as a
function of curing temperature for the MY721/MA/HHPA-system (r=0.8) as determined by
FT-Raman spectroscopy.
Temperature Anhydride Epoxide tdeviation
* -1
(k 1) [sec ] (k*2) -1
[sec ] [sec]
55°C 3.5e-5 2.9e-5 27360
65°C 5.7e-5 4.5e-5 15120
75°C 26.1e-5 19e-5 5880
100°C 3.5e-3 2.7e-3 1260

2.3.4.3 Cure mechanisms

The overall curing pattern clearly indicates that the network mainly consists of ester linkages
build by polyesterification. The observed reaction rates for the studied curing temperatures
are significantly higher than values reported for the non-catalyzed curing reaction [15]. This
observation provides an indication that the thermal cure of the investigated
TGDDM/MA/HHPA-system is probably catalyzed by the tertiary amine structure of the
TGDDM-monomer used. Hence, the curing reaction favors an anionic alternating chain-
polymerization as shown in Figure 6. However, the experimentally obtained FT-Raman data
cannot unambiguously used to clearly differentiate between a non-catalyzed and a catalyzed
mechanism, because the expected new vibrational bands are overlap with conjugated
species from the MA (cf. section 2.3.4.1). By plotting the empirical data in an epoxy-
anhydride conversion-conversion diagram, it is possible to establish the stoichiometry of the
reaction and to detect possible changes of the mechanism with conversion. A plot of this
type is shown in Figure 41 for the investigated system cured at 65°C. It can be seen that all
data points can be exactly accommodated on a single master curve. In particular, over the
complete isothermal low temperature curing, the data fit to an iso-conversion line with a
slope equal to one. This indicates a mechanism whereby, in the presence of sufficient
concentration of hardener, a single epoxy group reacts with one anhydride group in an
alternating co-polyesterification. However, from the data presented here it is not possible to
distinguish between the two anhydrides, hence it is not known if both anhydrides were
consumed at an equal rate. Moreover, the good fit to the iso-conversion line indicates that
the dominance of polyesterification excludes the potential polyetherification side reaction,
e.g. homopolymerization, in the investigated temperature range. However, the curing
proceeds not only by the polyesterification reaction between anhydride and epoxy.
Especially at higher curing temperatures a significant amount of TGDDM (tertiary amine)
initiated homopolymerization of maleic anhydride has been identified.
 Epoxy-anhydride crosslinking reaction 55

curing time [min]

0 25 50 75 100 125 150 175 200
4.0
0
3.5

reduced anhydride conversion [%]

3.0 20

(C0-Ct)epoxy [mol*kg ]
-1
2.5 isoconversion line
40
2.0
60
1.5

1.0 80

0.5
100
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
-1
(C0-Ct) anhydride [mol*kg ]

Figure 41: Absolute conversion of epoxy groups (C0-Ct) versus absolute conversion of
anhydride groups (), and the reduced anhydride conversion (Ct/Cmax) as function of curing
time () obtained from isothermal FT-Raman experiments at 75°C. The dotted line
represents the iso-conversion line. Molar anhydride to epoxy ratio of the MY721/MA/HHPA-
system was r=0.8.

2.3.5 Time-Temperature-Transformation (TTT)

The solidification behaviour during cure defines the processing window and additionally
determines the properties of the final material. Curing of epoxies is usually characterized by
different independent phenomena.
In order to visualize the sequences of different occurring events for different curing
temperatures, generalized time-temperature-transformation (TTT) graphs, representing the
specific times to gelation and vitrification as a function of cure temperature have been
developed [10, 88, 113, 114]. The TTT-cure diagrams enable the description of complex
processes that occur during the cure of a thermoset under isothermal cure conditions [88]
and have been established for a wide variety of different thermoset materials.
In these diagrams Tg0 characterizes the glass transition temperature of the initial mixture
containing the unreacted precursor monomers. Below Tg0 the crosslinking reactions are
frozen-in and the reaction rates are extremely slow. At temperatures above Tg0 the
crosslinking reaction results in the formation of small aggregates. With the progressing
curing reaction these species grow and become interconnected until the point of gelation is
reached. At gelation, the liquid to rubber transition corresponds to a non-reversible incipient
formation of a network with an infinite weight-average molecular weight. The gelation step is
of great technological interest because after it no material flow is possible [91, 115]. In
general gelation occurs at a fixed extent of conversion (αgel), as long as the reaction
mechanism does not change with temperature [88, 99].
56 Chapter 2

The glass transition temperature, Tg, of the system rises simultaneously with the
advancement in reaction. The moment when the Tg becomes equal to the curing
temperature is called vitrification. Vitrification involves a gradual, thermo-reversible process
describing the transformation from a liquid or rubbery state to a glassy state [88, 116, 117].
Therefore, the detection of vitrification processes varies with the analysis technique
employed. After vitrification the reaction rate is slowed considerably and the reaction
becomes diffusion controlled. The temperature at which gelation and vitrification occur
simultaneously is called gelTg. Above gelTg the gelation occurs before vitrification, whereas no
gelation occurs below gelTg. The further critical temperature, Tg∞, corresponds to the ultimate
glass transition temperature of the fully cured network.
The times to gelation (tgel) can be obtained by setting α=αgel. Integrating the eqn (2.11)
between α=0 and α0, and taking the natural logarithms yields:
 α 0 dα 
ln 

∫  = ln k( T ) + ln( t α 0 )
0 f (α ) 

(2.26)

Assuming further the Arrhenius form of k the left-hand side of eqn (2.26) being temperature
independent:
Ea
ln( t α 0 ) = C + (2.27)
RT
From experimental data, linear regression when plotting ln(tα0) versus 1/T provides the
constant, C, and thus it is possible to trace the iso-conversion lines in the kinetically
controlled regime, as shown in Figure 39.
Moreover, the connection between Tg and conversion is given by eqn (2.17), and
consequently it is possible to convert all the iso-conversion curves to iso-Tg curves.
Vitrification curves can then be theoretically predicted using the intersection of the iso-Tg
curve and Tg=Tcure, which is the fundamental criterion for the point of vitrification. The
calculated S-shape of the vitrification curve is influenced by the function of the kinetics as
well as the physical parameters Tg0, gelTg, Tg∞, and the relationship between the glass
transition temperature and the conversion (α) at Tg. Wisanrakkit and Gillham [85] have
confirmed the S-shape of the vitrification line by direct experiments.
The TTT-diagram for the MY721/MA/HHPA-system (r=0.8) using the physical constants
from Table 6 is shown Figure 42. It can be seen that the theoretical curves agree fairly well
with the experimental data. The obtained deviations depend on the determination and the
high sensitivity of the involved parameters. For example, a variation of only 1% in the
apparent activation energy leads to a ∼20% deviation of the calculated time to gelation at a
curing temperature of 100°C.
 Epoxy-anhydride crosslinking reaction 57

It was observed that the time to deviation obtained from the first order kinetic analysis of FT-
Raman measurement (see Table 5) agrees very well with the time to gelation data obtained
from rheological experiment as shown in Figure 42.

Table 6: Physical constants and other parameters used for TTT-diagram calculations
Parameter Symbol Value Method

Tg of the uncured resin [°C] Tg0 -21 Non-isothermal DSC

Tg of the fully cured resin [°C] Tg∞ 216 Non-isothermal DSC
Ratio of segmental mobility λ 0.193 DSC
(crosslinked/monomer)
Extent of conversion at gelation αgel 0.68 Rheology/Raman spectroscopy
Kinetic activation energy [kJ/mole] Ea 70.1 Raman spectroscopy
-1 6
Kinetic Arrhenius factor [sec ] A 2.2 *10 Isothermal DSC

α = 0.1, 0.2 ... 0.9

250
ultimate Tg
225
200 gelled
175 rubber
150 fully cured (α=0.99)
Temperature [°C]

125
gelled glass
100
75 liquid

50 gel Tg
25
0
-25 Tg0
ungelled glass
-50
-1 0 1 2 3 4 5
10 10 10 10 10 10 10
Time [min]

Figure 42: Time-temperature-transformation (TTT) diagram for the MY721/MA/HHPA-

system (r=0.8). Dotted lines represent iso-conversion contours. Solid lines represent the
behaviour of the time to gelation (tgel) and the time to vitrification (tvit). The (■) symbols
represent the time to gelation (tgel) experimentally determined by dynamic rheology; and the
() symbols denote the time to deviation (tdev) from n-th order kinetics experimentally
obtained by FT-Raman analysis.

Figure 43 shows the extent of conversion at vitrification. It is a fit of eqn (2.17) to the extent
of conversion at vitrification as a function of the isothermal curing temperature. Below the
curve the investigated MY721/MA/HHPA-system is liquid or gelled rubber, whereas above
the curve the material is in the state of an ungelled or gelled glass. The graphical
determination of gelTg yields 53°C.
58 Chapter 2

1.0

gelled glass
gelled rubber
0.8

ungelled

Conversion
0.6 α gel = 0.68
glass

0.4
liquid

0.2
gel Tg = 53°C

0.0
0 25 50 75 100 125 150 175 200 225 250
Temperature [°C]

Figure 43: Extent of conversion at vitrification as a function of temperature for the

MY721/MA/HHPA-system (r=0.8). The solid line was calculated by eqn (2.17).

2.4 Concluding remarks

Kinetics analysis of the thermal cure process of the MY721/MA/HHPA-system has been
studied by means of calorimetry, rheometry, and FT-Raman spectroscopy under isothermal
and non-isothermal conditions.
It was found that the results obtained from dynamic and isothermal DSC analysis yielded a
serious disagreement between these two methods. In particular the application of single
heating methods, e.g. the Barrett and the Carroll-Freeman relation, revealed generally
higher values for the calculated kinetic parameters than those calculated from non-
isothermal measurements. The relevance of the reaction order (n) in an autocatalytic
environment at normal processing temperatures as well as the apparent activation energy
(Ea) seems to be highly questionable. A more detailed mechanistic analysis prior to
computation of the reaction order (n) is necessary if dynamic methods should be employed.
More meaningful kinetic parameters were obtained by multiple-heating methods (Kissinger
and Ozawa). However, these model-free methods are independent of the reaction order and
hence not accurate for more complex reactions (n≠1).
Kinetic parameters obtained from isothermal methods (DSC, rheometry and Raman
spectroscopy) are consistent in the studied curing temperature range. However, the quality
of the DSC-kinetic analysis is greatly influenced by the total reaction enthalpy (∆Htot). For the
presently investigated system the total reaction heat is significantly superimposed by
competitive exothermal events, e.g. decarboxylation reaction of MA, which outlines the
principal limitations of the underlying DSC assumptions.
 Epoxy-anhydride crosslinking reaction 59

Isothermal FT-Raman spectroscopy was identified as the most powerful method to study the
curing kinetics. Quantitative analysis of in situ FT-Raman spectra showed that the total
anhydride and epoxy consumption, as well as the ester formation follows pseudo-first order
kinetics up to the point where the reaction becomes diffusion controlled (α ~70%) in
accordance to the proposed mechanistic reaction scheme. The obtained kinetic parameters
for different methods and models are summarized in Table 7.
The observed deviations between isothermal and non-isothermal kinetic calculations can be
related to mainly two facts. Firstly, the fit of non-isothermal data to hypothetical and
mechanistically free models results in a highly uncertain set of single Arrhenius parameters
that cannot be meaningfully compared to isothermal data. Secondly, the isothermal and non-
isothermal experiments are conducted at different temperature ranges. Since measurement
conditions in all applied isothermal methods are more similar to the processing environment
in many technical applications, these methods should be favored for detailed process
optimizations.

Table 7: Calculated kinetic parameters for the MY721/MA/HHPA-system (r=0.8) cure under
isothermal and non-isothermal cure conditions
Analytical method Model Activation energy Frequency factor Reaction
-1
(Ea) [kJ/mole] (lnA) [sec ] order
Dynamic DSC Freeman-Carrol 113.2 n.a. n= 2.0
1)
Dynamic DSC Barrett 112.6 27.4 n= 2.0
Dynamic DSC Kissinger 63.5 n.a. n.a
Dynamic DSC Ozawa 67.3 n.a. n.a.
2)
Isothermal DSC Autocatalysis E1=51.2 14.2 m=0.40
E2= 0.04 -7.7 n=1.042)
Isothermal Rheometry Isoconversion 58.7 11.5 n.a.
Isothermal FT-Raman n-th order kinetic 70.0 15.8 n.a.
1)
Reaction order value obtained from the Freeman-Carroll relation
2)
Arithmetic mean of reaction orders (m and n) obtained at curing temperatures of 100, 110, 120 and 140°C

In addition, FT-Raman spectroscopy was used to monitor the anhydride/epoxy curing

reaction in order to investigate the curing mechanism. Therefore, real-time quantitative
concentrations of epoxy, anhydride and ester groups were measured during the cure of
MY721/MA/HHPA-system at temperatures between 55°C to 100°C. Conversion data of the
functional groups were determined and used to evaluate the reaction mechanisms. It was
shown that the investigated anhydride-cured epoxy system reacts mainly by consecutive-
step polyesterification. FT-Raman and gas chromatography support the hypothesis that the
tertiary amine structure of the investigated amino-glycidyl resin is able to initiate
homopolymerization of maleic anhydride, which is followed by decarboxylation and the
60 Chapter 2

release of water and carbon dioxide. The formation of these reaction products could
contribute to an autocatalytic curing reaction but the kinetic analysis does not indicate
autocatalysis is significant under the conditions of low-temperature cure.
The use of temperature modulated DSC (TMDSC) technique enables differentiation of the
MY721/MA and MY721/HHPA reaction rates by monitoring the non-reversing heat flow
signal. Additionally, TMDSC experiments provide evidence of a heterogeneous sub-network
structure for the co-anhydride cured epoxy network.
A generalized time-temperature-transformation-diagram was developed in order to predict
the complex material transformations (e.g. gelation and vitrification) occurring during the
entire isothermal cure process.

2.5 References

[1] Barton J.M., Adv. Polym. Sci., 72, 111 (1985).

[2] Rogers R.N., Smith L.C., Anal. Chem., 39, 1024 (1967).
[3] Grenier-Loustalot M., Grenier P., Brit. Polym. J., 22, 303 (1990).
[4] St John N.A., George G.A., Polymer, 33, 2679 (1992).
[5] Morgan R.J., Mones E.T., J. Appl. Polym. Sci., 33, 999 (1987).
[6] Dusi K., Maximovich M., Galeos J.K., J. Appl. Polym. Sci., 28, 2567 (1983).
[7] Tajima Y.A., Polym. Compos., 3, 162 (1982).
[8] Hedvik P., Dielectric Spectroscopy of Polymers, Wiley & Sons, New York, (1971).
[9] Ellis B., Chemistry and Technology of Epoxy Resins, 1, Blakie Academic
Professional, Glasgow, (1993).
[10] May C.A., Epoxy Resins, Marcel Dekker, New York, (1988).
[11] Lee H., Neville K., Handbook of Epoxy Resins, McGraw-Hill, New York, (1967).
[12] Hagnauer G.L., Pearce P.J., ACS Symp. Ser., 221, 193 (1992).
[13] Barton J.M., Greenfield D.C.L., Brit. Polym. J., 18, 51 (1986).
[14] Morgan R.T. In: Advances in Polymer Science (ed. Dusek K.), vol. 72, Springer-
Verlag, Berlin, pp. 1, (1985).
[15] Steinmann B., J. Appl. Polym. Sci., 37, 1753 (1989).
[16] Fisch W., Hofmann W., Kaskikallio J., J. Appl. Chem., 6, 429 (1956).
[17] O'Neill L.A., Cole C.P., J. Appl. Chem., 6, 356 (1956).
[18] Dearborn E.C., J. Polym. Sci., 16, 201 (1955).
[19] Wegler R., Angew. Makromol. Chem., 67, 587 (1955).
[20] Tanaka Y., Kakiuchi J., J. Appl. Polym. Sci., 7, 1063 (1963).
[21] Fisch W., Hofmann W., J. Polym. Sci., 12, 497 (1954).
[22] Fisch W., Hofmann W., Macromol. Chem., 44-46, 8 (1961).
[23] Matejka L., Lovy J., Pokorny S., Bouchal K., Dusek K., J. Polym. Sci., 21, 2873
(1983).
[24] Antoon M.K., Koenig J.L., J. Polym. Sci., 19, 549 (1981).
[25] Fischer R.F., J. Polym. Sci., 44, 155 (1960).
[26] Trappe V., Burchard W., Steinmann B., Makromol. Chem., Macromol. Symp., 45, 63
(1991).
[27] Dusek K., Adv. Polym. Sci., 78, 1 (1986).
[28] Fedtke M., Domaratius F., Polymer Bull. (Berlin), 15, 13 (1986).
[29] Tanaka Y., Kakiuchi J., J. Macromol. Chem., 1, 307 (1966).
[30] Luston J., Manasek Z., Kulikova M., J. Macromol. Sci. Chem., A12, 995 (1978).
 Epoxy-anhydride crosslinking reaction 61

[31] Lutson J., Manasek Z., J. Macromol. Sci. Chem., 12, 983 (1978).
[32] Antoon M.K., PhD-Thesis, Case Western Reserve University, Cleveland (OH),
(1980)
[33] Tanaka Y., Haung C.M., Makromol. Chem., 120, 1 (1968).
[34] Matejka L., Pokorny S., Dusek K., Macromol. Chem., 186, 2025 (1985).
[35] Dusek K., Lunak S., Matejka L., Polym. Bull., 7, 145 (1982).
[36] Leukel J., Burchard W., Krüger R.P., Munch H., Schulz G., Macromol. Rapid
Commun., 17, 359 (1996).
[37] Steinmann B., J. Appl. Polym. Sci., 39, 2005 (1990).
[38] King J.J., Castonguary R.N., Ziggi J.P., Proc. 13th Nat. SAMPE Tech. Conf., Mount
Pocono, Pennsylvania, 13, (1981).
[39] Thoseby M.R., Dobinson B., Bull C.H., Brit. Polym. J., 18, 286 (1986).
[40] Sauerbrunn S.R., Growe B.S., Reading M., Proc. 21st NATAS Conf., Atlanta, GA,
21, 137, (1992).
[41] Van Assche G., Van Hemelrijck A., Rahier H., Van Mele B., Thermochim. Acta ,
304/305, 317 (1997).
[42] Barrett K.E.J., J. Appl. Polym. Sci, 11, 1617 (1967).
[43] Barton J.M., J. Macromol. Sci. Chem., 8, 25 (1974).
[44] Kissinger H.E., Anal. Chem., 29, 1702 (1957).
[45] Freeman E.S., Carroll B., J. Phys. Chem., 62, 394 (1958).
[46] Ozawa T., J. Therm. Anal., 2, 301 (1970).
[47] Patel R.D., Patel R.G., Patel V.S., J. Therm. Anal., 34, 1283 (1988).
[48] Patel S.P., Patel R.G., Thermochim. Acta , 202, 97 (1992).
[49] Prime R.B. Thermal Characterization of Polymeric Materials, In: (ed. Turi E.A.),
Academic Press, New York, pp. 435, (1981).
[50] Riccardi C.C., Adabbo H.E., Williams R.J.J., J. Appl. Polym. Sci, 29, 2481 (1984).
[51] Rogers R.N., Smith L.C., Thermochim. Acta , 1, 1 (1970).
[52] Martz J.T., O'Dwyer S.J., Thomas S.J., Claar J.A., EP 0316874; (1989).
[53] McCormick C.L., Chang Y., Macromolecules, 27, 2151 (1994).
[54] Panarin E.F., Sevetlove I.I., Vysokomol. Soedin. Ser. B, 15, 880 (1973).
[55] Fava R.A., Polymer, 9, 137 (1968).
[56] Peyser P., Bascom W.D., J. Appl. Polym. Sci., 21, 2359 (1977).
[57] Boey F.Y.C., Qiang W., Polymer, 41, 2081 (2000).
[58] Hong S.G., Wang T.C., J. Appl. Polym. Sci, 52, 1339 (1994).
[59] Ricca G., Severini F., Polymer, 29, 880 (1988).
[60] Bhadani S.N., Sahu S.U., Kundu S., Ind. J. Tech., 20, 118 (1982).
[61] Davie W.R., Chem. Eng. News, 42, 41 (1964).
[62] Ghaemy M., Riahy M.H., Eur. Polym. J., 32, 1207 (1996).
[63] Horie K., Hiura H., Sawada M., Mita I., Kambe H., J. Polym. Sci. Part A, 8, 1357
(1970).
[64] Jagadeesh K.S., Siddaramaiah A., Kalpagam V., J. Appl. Polym. Sci, 40, 1281
(1990).
[65] Kamal M.R., Polym. Eng. Sci., 14, 231 (1974).
[66] Mauri A.N., Galego N., Riccardi C.C., Williams R.J.J., Macromolecules, 30, 1616
(1997).
[67] Kenny J.M., Trivisano A., Nicolais L., Adv. Chem. Ser., 254, 539 (1993).
[68] Khanna U., Chanda M., J. Appl. Polym. Sci., 50, 1635 (1993).
[69] Mezzenga R., Boogh L., Manson J.A., Pettersson B., Macromolecules, 33, 4373
(2000).
[70] Sourour S., Kamal M.R., Thermochim. Acta , 14, 41 (1976).
[71] Woo E.M., Seferis J.C., J. Appl. Polym. Sci, 40, 1237 (1990).
[72] Kenny J.M., Maffezzoli A., Nicolais L., Compos. Sci. Technol., 38, 339 (1990).
[73] Kenny J.M., Trivisano A., Polym. Eng. Sci., 31, 1427 (1991).
[74] Dell´Erba R., Martuscelli E., Musto P., Ragosta G., Leonardi M., Polym. Networks
Blends, 7, 1 (1997).
62 Chapter 2

[75] Park W.H., Lee J.K., J. Appl. Polym. Sci., 67, 1101 (1998).
[76] Eom Y.S., PhD, EPFL, Lausanne, (1999)
[77] Cole K.C., Hechler J.J., Noel D., Macromolecules, 24, 3098 (1991).
[78] Ishida H., Rodriguez Y., 35th IUPAC International Symposium on Macromolecules,
Akron (OH), 634, (1994).
[79] Malavasic T., Moze A., Vizovisek I., Lapanje S., Angew. Makromol. Chem., 44, 89
(1975).
[80] Tanaka Y., Kakiuchi J., J. Appl. Polym. Sci. A, 2, 3405 (1964).
[81] Montserrat S., Flaque C., Page P., Malek J., J. Appl. Polym. Sci, 56, 1413 (1995).
[82] Montserrat S., Malék J., Thermochim. Acta , 228, 47 (1993).
[83] Stevens G.C., J. Appl. Polym. Sci, 26, 4279 (1981).
[84] Simon S.L., Gillham J.K., J. Appl. Polym. Sci., 47, 461 (1993).
[85] Wisanrakkit G., Gillham J.K., J. Appl. Polym. Sci, 41, 2885 (1990).
[86] DiBenedetto A.T., Nielsen L.E., J. Macromol. Sci. Macromol. Chem., C3, 69 (1969).
[87] Nielsen L.E., J. Macromol. Sci. Revs. Macromol. Chem., C3, 69 (1969).
[88] Enns J.B., Gillham J.K., J. Appl. Polym. Sci, 28, 2567 (1983).
[89] Venditti R.A., Gillham J.K., J. Appl. Polym. Sci., 64, 3 (1997).
[90] Wisanrakkit G., Gillham J.K., Enns J.B., J. Appl. Polym. Sci., 41, 1895 (1990).
[91] Tung C.M., Dynes J.P., J. Appl. Polym. Sci, 27, 569 (1982).
[92] St John N.A., George G.A., Cole-Clark P.A., Mackay M.E., Halley P.J., High Perform.
Polym., 5, 21 (1993).
[93] Winter H.H., Polym. Eng. Sci., 27, 1698 (1987).
[94] Winter H.H., Chambon F., J. Rheol., 30, 367 (1986).
[95] Feve M., Macromol. Chem., Macromol. Symp., 30, 95 (1989).
[96] Matejka L., Polym. Bull., 26, 109 (1991).
[97] Lairez D., Adam M., Emery J.R., Durand D., Macromolecules, 25, 286 (1992).
[98] Corcuera M.A., Mondragon I., Riccardi C.C., Williams R.J.J., J. Appl. Polym. Sci, 64,
157 (1997).
[99] Flory P.J., Principles of Polymer Chemistry, Cornell University Press, Ithaca, (1953).
[100] Stockmayer J., J. Chem. Phys., 11, 45 (1943).
[101] Macosko C.W., Miller D.R., Macromolecules, 9, 199 (1976).
[102] Smith B., Infrared Spectral Interpretation - A Systematic Approach, 1 ed., CRC
Press, New York, (1999).
[103] Dyke S.F., Floyd A.J., Sainsbury M., Theobald R.S., Organic Spectroscopy - An
Introduction, Longman Publsiher, London,
[104] Pretsch E., Bühlmann P., Affolter C., Structure Determination of Organic
Compounds, Springer Publisher, Berlin, (2000).
[105] Bellamy L.J., The Infrared Spectra of Complex Molecules, Chapman and Hall, New
York, (1975).
[106] De Bakker C.J., George G.A., St John N.A., Fredericks, P.M., Spectrochim. Acta,
49A, 739 (1993).
[107] Walton J.R., Williams K.P.J., Vibr. Spectrosc., 1, 339 (1991).
[108] Rocks J., Halter M., George G., Vohwinkel F., Polym. Int., 52, 1749 (2003).
[109] Agbenyega J.K., Ellis G., Hendra P.J., Maddams W.F., Passingham C., Willis H.A.,
Chalmers J., Spectrochim. Acta, 46A, 197 (1990).
[110] Stevens G.C., J. Appl. Polym. Sci, 26, 4259 (1981).
[111] Zweifel H., Völker T., Makromol. Chem., 170, 141 (1973).
[112] Simms J.A., Corcoran P.H., Prog. Org. Coatings, 26, 217 (1995).
[113] Gillham J.K., Polym. Eng. Sci., 19, 676 (1979).
[114] Chan L.C., Gillham J.K., Kinloch A.J., Shaw S.J., Adv. Chem. Ser., 208, 235 (1984).
[115] Eloundou J.P., Gerard J.F., Harran D., Pasault J.P., Macromolecules, 29, 6917
(1996).
[116] Babayevsky P.G., Gillham J.K., J. Appl. Polym. Sci, 17, 2067 (1973).
[117] Gillham J.K., Polym. Eng. Sci., 26, 1429 (1986).
[118] Riccardi C.C., Dupuy J., Williams R.J.J., J. Polym. Sci Part B, 37, 2799 (1999)
 Deformation and fracture behaviour 63

Chapter 3

3 Deformation and fracture behaviour

One common feature of epoxy thermosets is their inherent brittleness which is a direct
consequence of the highly crosslinked network structure. Upon deformation these dense
networks exhibit only minor elastic deformation and low elongation at break. Thus, this
intrinsically brittle behaviour often limits the use of epoxy materials in numerous industrial
applications. For this reason, the knowledge of the deformation and the connected fracture
behaviour is one of the most important to enable design of structural parts out of
thermosets materials.
At the beginning of this chapter a general introduction to the linear fracture mechanics is
given, followed by an extensive review of the different toughening approaches and related
toughening mechanisms. The deformation behaviour of the co-anhydride cured TGDDM-
formulation under tensile and compressive load was studied, and the influence of test
temperature and strain rates on the basic yield and deformation behaviour was evaluated.
Microscopy analysis was used to investigate the microstructure in order to identify the
deformation mechanisms.
The fracture behaviour of the highly crosslinked epoxy was examined and toughening
attempts with inorganic hard particles were carried out. The effects of external test
parameters such as temperature and strain rate, as well as the influence of compositional
variations like filler fraction, filler size and filler surface-treatment have been studied.
Moreover, the potential of using reactive, polydisperse and highly branched polymers
(hyperbranched polymer; HBP) as novel tougheners as well as the HBP application in
hybrid-compositions (silica+HBP) has been extensively examined. In order to obtain a
fundamental understanding of the involved micro-deformation and fracture mechanisms
this study focused primarily on the structure-property relationship, which is a result of the
curing kinetics, thermo-mechanical properties, and morphology of epoxy-silica and epoxy-
HBP blends.

3.1 Introduction to fracture toughness

The term fracture toughness, in a broader sense, is a measure of a material’s resistance to

failure. Fracture toughness is usually measured as either the stress or the energy required
for failure of a specimen under a specific load condition. Therefore, toughness can be
64 Chapter 3

defined more specifically as (1) the area under the tensile stress-strain curve, (2) the
Izod/Charpy impact strength, (3) the critical stress intensity factor (KΙc) under plane strain
conditions, or (4) the critical strain energy release rate (GΙc) under plane strain conditions.
Although the definition of toughness varies from application to application, it is nowadays
in generally agreed that, either KΙc or GΙc –values should be considered for the “worst case
scenario”.

3.2 Linear elastic fracture mechanics (LEFM)

Linear elastic fracture mechanics describes the behaviour of sharp cracks in linear,
perfectly elastic materials. From the comprehensive literature on this subject only a few
works deal with the deformation and fracture of polymers [1-4]. The foundation of the linear
fracture mechanics was originally laid by Griffith [5], who recognized for the first time in
1920, that the strength of crystalline or glassy solids was much lower than those based on
theoretical considerations. Theoretical strengths can be calculated from the force-
displacement curves of interatomic bonds, since the strength of a solid is limited
ultimatively by the strength of its atomic bonds [6]. Typical values are of the order of E/10,
where the E is the Young’s modulus. For neat epoxy resins with typical values of Young’s
modulus in order of 3 GPa the calculated theoretical failure strength is in the range of 300
MPa. In general, however, experimental evidence clearly indicated that the failure strength
is orders of magnitude below the theoretical value [7]. This clearly indicates that the
distribution of local stresses between individual bonds, even in an apparently
homogeneous material under uniform stress conditions, is not uniform. Griffith explained
this effect by the presence of small defects and/or flaws in the material, which lead to a
significant stress enhancement on a microscopic scale. The material is assumed to obey
Hooke’s law so that the stress is directly proportional to the infinitesimal strains. For any
loaded body, therefore, a linear load-deflection relationship will result.
Nowadays, the linear fracture mechanics is based on two independent fracture conditions,
the energy analysis (G-approach) and the localized stress field analysis (K-approach). A
more detailed explanation of these two methods will be given in the following sections.

3.2.1 Strain energy release rate criterion (G-Approach)

Griffith’s explanation of the deviation between the theoretical calculated strength and the
experimentally observed values was based on the presence of micro cracks in the
material. Moreover, Griffith considered that the elastically stored energy is not distributed
 Deformation and fracture behaviour 65

uniformly throughout the sample but is locally concentrated near the flaw. He used an
energy balance approach and calculated the energy released by introducing an elliptical
crack of a length (l=2a) and width (b) into the plate under a constant applied stress (see
Figure 44), and related this to the energy required to create two new fracture surfaces.
P
y

b
a x

P
Figure 44: Generalized loading of a cracked body which exhibits bulk linear elastic
behaviour

Thus the total energy of the cracked plate (Utot) can be written as eqn (3.28):
Utot = U0 + Uγ + Ua − P (3.28)
where
U0 = elastic energy of the loaded uncracked plate (a constant)
Uγ = change in the elastic energy due to the formation of two crack surfaces
(elastic surface energy)
Ua = the additional change in the elastic energy in the plate due to the
presence of the crack
P = work performed by external forces

If it is assumed that the material behaves in a perfectly linear elastic manner, the
corresponding change in the elastic energy due to the formation of the two crack surfaces,
Uγ, is simply equal to the product of elastic surface energy of the material, γ, and the new
surface area of the crack:
U γ = 2(2aγ ) = 4aγ (3.29)

To determine the change in the elastic energy in the plate due to the presence of the crack
(Ua), Griffith’s calculation requires integration over the stress field around the crack using
the mathematical solution as proposed by Inglis [8] yielding to eqn (3.30):
π σ2 a2 (3.30)
Ua =
E
where E represents the Young’s modulus of the material comprising the plate, and σ
denotes the macroscopic stress.
Assuming a “fixed grip” condition (P=0) and the change in elastic energy (Ua) is negative,
the total energy of the cracked plate is:
66 Chapter 3

π σ2 a2 (3.31)
Utot = U0 + U γ + Ua = U0 − + 4a γ
E
Consequently, the equilibrium condition for crack extension is obtained by taking the
derivative with respect to crack length (dU/da) and setting this sum equal to zero:
 − π σ2 a2  (3.32)
d + 4aγ 
 E  = 0
da

Eqn (3.32) states that when the elastic strain energy released due to a potential increment
of crack growth (da) outweighs the demand for surface energy for the same crack growth,
then the crack grows in an unstable manner, causing fracture. The results of this
equilibrium condition can be written as:
2 π σ2 a (3.33)
≥ 4γ
E
This equation can be rearranged to the general fracture criterion, as given by eqn (3.34):

(3.34)
2Eγ
σf =
πa
where σf is the fracture stress.
Eqn (3.34) is the common form in which the Griffith criterion is expressed, but it is only
strictly applicable for thin plate and when plane stress conditions prevail. For a thick plate
and under plane strain conditions it becomes eqn (3.35):

2Eγ (3.35)
σf =
(
π 1− ν2 a )
where ν is the Poisson’s ratio.
The specific values for the fracture surface energy (γ) can be calculated using the Griffith
eqn (3.35). However, for epoxy resins these values are calculated to be between 50 and
200 Jm-1 [9], which are much higher than the theoretical surface energy of epoxy resins,
which are thought to be of the order of 1 Jm-1. This discrepancy is due to Griffith’s
underlying assumption of only pure elastic behaviour of the material and the absence of
any plastic deformation.
Irwin [10] has developed a theory for fracture of slightly ductile materials, which undergo a
small amount of plastic deformation in a plastic zone in front of the crack tip. The energy
absorbed during plastic deformation is much higher than the surface energy and this is
reflected in the measured values of γ for most materials, which are much larger than the
theoretical ones. For this reason the strain energy release rate, G, instead of the 2γ, has
been introduced and this represents the total work of fracture and not only the true surface
energy. The Griffith eqns (3.34) and (3.35) for plane strain and plane stress conditions,
 Deformation and fracture behaviour 67

respectively, can be modified to eqns (3.36) and (3.37). Fracture occurs when G reaches
the critical value Gc:

GcE (3.36)
σf =
πa

Gc E (3.37)
σf =
(
π 1− ν2 a )

3.2.2 Stress intensity factor (K-Approach)

An alternative approach was proposed by Irwin [11], that considers the stress field near an
idealized crack of 2a length, as shown in Figure 45.
σyy σzz

y σxy
σxx

θ
x

Figure 45: The stress field near an idealized crack of length 2a

A more accurate stress analysis has been proposed by Westergaard [12]. In his model the
stresses at any point, with the distance (r) from the crack tip and an angle (θ) to the crack
propagation direction, are given by:
68 Chapter 3

cos (θ / 2)[1 − sin(θ / 2) sin(3θ / 2)]

KΙ
σ xx = (3.38)
(2πr )1/ 2

(3.39)
cos (θ / 2)[1 + sin(θ / 2) sin(3θ / 2)]
KΙ
σ yy =
(2πr )1/ 2

(
σ zz = ν σ xx + σ yy ) for plane strain conditions
(3.40)

σ zz = 0 for plane stress conditions (3.41)

KΙ
σ xy = cos (θ/2 )sin (θ/2) cos(3θ / 2)
(2πr )1/2 (3.42)

σ yz = σ zx = 0
(3.43)
The value of Irwin’s approach is that the stress field around crack is identical for all types
of loading situations normal to the crack. If the magnitude of the stresses (i.e. their
intensity) is constant for a given load and geometry, than KI is called the stress intensity
factor. The subscript Ι denotes to the loading mode of the crack.
For an infinite sheet with a central crack of length 2a, it was shown by Irwin that when σ
reaches the fracture stress σf, KI has a critical value that is termed the critical stress
intensity factor KIc as given by:

K Ic = σ f πa (3.44)

Consequently, a mathematical link with the previously proposed Griffith criterion is given
by a transformation of eqn (3.44) to eqn (3.45), which is identical in form to eqn (3.34).
2
K Ic (3.45)
σf =
πa

3.2.3 Fracture toughness test modes

For the determination of the material properties KIc and GIc three experimental methods
(Figure 46) of controlled crack propagation are predominantly used:

- Mode Ι, crack opening for tensile mode

- Mode ΙΙ, sliding or in-plane shear mode
- Mode ΙΙΙ, tearing or anti-plane shear mode

More details on these methods, their evaluation, and possible influence of sample
geometry on the resulting material values are given in [13-16]. A general case of cracking
 Deformation and fracture behaviour 69

may include superposition of all three modes. Nonetheless, mode Ι is the most critical
mode, hence it is the only one that will receive detailed consideration in this work.

Figure 46: Experimentally used modes of controlled crack propagation: (left) the crack
opening or tensile mode Ι, (middle) the sliding or in plane-shear mode ΙΙ, and (right) the
tearing or anti–plane shear mode ΙΙΙ

3.3 Toughening of epoxy thermosets

Common ways to toughen brittle epoxies, which typically have GIc-values less than
200 J/m2, include (1) chemical modifications of a given rigid epoxy backbone to a more
flexible backbone structure, (2) increase the molecular weight of the epoxy monomer, (3)
lowering of the crosslinking density of the cured resin by mixtures of high monomer
molecular epoxy weight or low functional curing agents, and (4) incorporation of dispersed
toughener phase(s) in the cured epoxy matrix.
Toughening via dispersed phase(s) morphology has proved to be the most effective
without a significant reduction of crosslink density (Tg) as occurs with the first three
methods. Taking this into account, fracture toughness of brittle epoxies can be most
effectively improved by using at least one of three following major methods, which all
based on the addition of second phase particles. The differences between them are the
types of particles employed; hence these modified-epoxies have different designations.
When the particles used are soft rubber, the material is called rubber-modified epoxy.
When the particles consist of hard inorganic particles, the material is known as particulate-
filled epoxy. When the particles are composed of rigid thermoplastic resin the modified
material is termed thermoplastic modified epoxy. In the following sections of this work the
different modifications are reviewed with respect to their toughening mechanisms and
applications.

3.3.1 Epoxy toughening with soft particles (rubber toughening)

The technology of rubber-modified epoxies is at least thirty years old and has been
extensively studied in traditional diglycidyl ether of bisphenol-A (DGEBA) and diglycidyl
70 Chapter 3

ether of bisphenol-F (DGEBF) resins. The work of McGarry et al. [17-19] is among the first
to show that significant improvements in toughness are obtainable by adding certain liquid
rubbers to epoxy formulations. Since their exploratory work, a tremendous amount of
research on the nature of toughening mechanisms has been carried out [20-47]. This
resulted in a wide range of different toughened epoxy materials. The type of rubber to be
added to toughen the resin is mainly determined by two factors [6]; (1) the compatibility
requirement: the rubber must be totally miscible in the resin prior to cure, but separates as
a second phase during the cure of the resin. (2) The chemical requirement: the rubber
should react with the epoxy group in order to form chemical linkages between the modifier
and the matrix.
Different types of acrylonitrile-based copolymers generally fulfill both requirements and
therefore they are widely established as toughening modifiers. The effects of the rubber
concentration, acrylonitrile content, molecular weight, solubility parameter, and end-group
functionality on the fracture toughness of epoxy/rubber blends have been extensively
studied. Carboxyl-terminated butadiene-acrylonitrile (CTBN) rubber is the most widely
investigated toughening modifier currently used [24, 25, 39, 45, 48, 49]. To a lesser extent
amino terminated butadiene-acrylonitrile (ATBN) [35, 50-54], epoxy-terminated butadiene-
acrylonitrile (ETBN) [50, 55-57] and hydroxyl-terminated butadiene-acrylonitrile (HTBN)
[58-61] have also been considered to toughen epoxies.
Even if the huge amount of available information is in some cases contradictory, for the
relationship between the micro-structural parameters and mechanical properties it can be
generally stated that:
Phase separated morphologies outperform homogeneous structures in terms of
improvement of fracture toughness.
The toughness of phase separated rubber modified DGEBA-thermosets increases with
volume fraction of the dispersed rubbery phase. An optimum of the mechanical properties
is obtained by a phase-separated morphology with a rubber content between 10 to 20
parts per hundred resin (phr).
The toughening is primarily attributed to the size and distribution of the precipitated rubber.
Large rubber particles (>5 µm) show only a moderate increase in fracture toughness by a
particle bridging/crack deflection mechanism. In contrast smaller particles (1-3 µm) provide
a significant increase in toughness through cavitation-induced shear banding.
However, the main deficiency in the use of the above reported oligomers is the high level
of unsaturation in their structure, which provides sites for degradation reactions in
oxidative environments [62]. Consequently, thermosetting materials toughened with
elastomeric particles are predominately used in low temperature applications [63]. To
 Deformation and fracture behaviour 71

overcome this intrinsic limitation, other elastomeric modifiers have been investigated as
well. For example, Banthia et al. [29] have used acrylic elastomer oligomers as dispersed
phase to improve fracture toughness. Wang and co-worker [64] obtained similar results
with epoxy functionalized poly(n-butylacrylate) (ETPnBA) and carboxyl functionalized
poly(n-butylacrylate) (CTPnBA) as novel modifier. Fluorocarbon elastomers [65-68] have
also been the focus of several research activities.
The incorporation of different rubbers to an inherently brittle resin generally modifies its
characteristics. For example, it significantly increases the fracture toughness and related
toughness improvements depend upon various parameters, such as particle size and
volume fraction, strain rate, temperature and other environmental factors. The major
postulated toughening mechanisms are schematically illustrated in Figure 47.
Consequently, because of such a gain in toughness for relatively loosely cross-linked
thermosets (e.g. low glass transition temperature), rubber toughened epoxies are being
successfully used in various engineering applications.

Figure 47: Schematic representation of several toughening mechanisms proposed by

Pearson [69] for rubber-modified epoxies: 1) rubber tearing; 2) particle bridging; 3) crack
path deflection, 4) particle cavitation; 5) cavitational induced shear bending; 6) micro-
cracking

On the other hand the rubber toughening approach causes the disadvantages of a
reduction in stiffness, lowers Tg, plasticizes the matrix (if the rubber is not fully separated
from the solution in the resin), reduces yield strength and increases the linear thermal
expansion coefficient, and hence adversely affects the resin performance in hot/wet
conditions in comparison to the unmodified system. These limitations are often
unacceptable for many structural and long-term applications. A further limitation is the fact
that the rubber toughening method becomes progressively less effective as the
crosslinking density of the matrix increases (multifunctional epoxy systems), as
summarized in Figure 48.
72 Chapter 3

The effect of the matrix crosslink density on toughness has been extensively analyzed [44,
72, 73]. Rubber toughening of thermosetting resins depend primarily upon micro-
deformation in the resin matrix, which is usually accompanied by cavitation within the
rubber particles, and consequent stress whitening of the yield zone. As the resin matrix is
crosslinked more densely, its capability to plastically deform by shear yielding is reduced
[74]. In view of this inherent problem in causing yielding in high glass transition
temperature resins, it appears that successful toughening can be achieved only by
introducing energy absorption processes that do not depend upon matrix ductility.
Following the need to increase the toughness by maintaining a high glass transition
temperature, hard particles and/or high performance ductile thermoplastics can be used
instead of rubber to modify dense epoxy resins. These two toughening concepts will be
further reviewed and discussed in the following sections.

1.8
C H3 C H3
1.6 HO O C H 2C 2
C R n
C C H2 2
COO H
C H3 C H3

1.4
R = C H2 CH CH C H2 a
C H2 CH b
C H2 CH c
1.2 CH CN
KIc [MPam ]
0.5

C H2

1.0

0.8

0.6

0.4

0.2
0 5 10 15 20 25 30
Concentration [wt% CTBN]

Figure 48: Critical stress intensity factor (KIc) as a function of carboxyl-terminated

butadiene-acrylonitrile co-polymers (CTBN) fraction for different highly crosslinked epoxy-
systems (0) ref. [41], () ref. [70], () ref. [70], ref. () [70], (^) ref. [71]. All values were
measured at room temperature.

3.3.2 Epoxy toughening with hard inorganic particles

The technology of reinforcing epoxies with inorganic hard particles is at least as old as the
rubber modification approach. Mechanisms proposed for particle-filled epoxies include (1)
an increase in fracture surface due to the irregular path of the crack, (2) plastic
deformation of the matrix around the particles; and (3) the crack pinning and debonding
and/or micro-cracking mechanism. These mechanisms are schematically illustrated in
Figure 49. As described above, toughening through the use of inorganic particles has
some rather complicated aspects and it is improbable that a simple mechanism can
 Deformation and fracture behaviour 73

explain the complex fracture behaviour. Despite this reservation, today the crack pinning
(crack bowing) mechanism is considered as the most relevant toughening mechanism for
inorganic particle toughening.

This image is not available online. Please consult the

hardcopy thesis available from the QUT Library.

Figure 49: Schematic toughening mechanisms for glass-filled epoxies: (1) crack pinning,
(2) particle bridging, (3) crack path deflection, (4) debonding, (5) crack branching and (6)
micro-cracking [69].

3.3.2.1 Crack pinning mechanism of hard particles

Lange has proposed the development of the crack pinning mechanism for a two-phase
brittle material in 1968 [75, 76]. He explained the toughening effect of rigid particles in
brittle matrices with the “line tension concept” that has been originally developed for
dislocation motion in ceramic materials. The obstacles (rigid filler particle) create an
obstruction for the propagation of the primary linear crack front, and cause an increase in
toughness by increasing the crack length compared with the unfilled system (compare L(u)
and L(f) in Figure 50).
unfilled filled

a)

b) L(u)

L(f)

Figure 50: Crack-pinning mechanism as proposed by Lange [77]. (a) side view, (b) top
view. L(u) is the line length of the crack front in an unfilled resin. In a filled resin the line is
bowed and thus L(f) is longer than L(u).
74 Chapter 3

A simple mathematical description based on this line tension concept and related to the
increase in fracture energy, G0, due to pinning, was derived:
G0 = G Ι c (matrix ) + 2TL / D s (3.46)

Where TL is the line energy per unit length, and Ds is the interparticle distance. This eqn
(3.46) presumes that the crack front breaks away from the pinning positions when it attains
a radius of Ds/2. In addition, the interparticle distance can be obtained from eqn (3.47),
[78]:
2d (1 − Vf )
Ds = (3.47)
3 Vf
where, d is the average particle diameter and Vf is the volume fraction of rigid filler.
For a penny-shaped crack, Lange [79] showed that the line energy TL predicts an inversely
proportional relationship between the increase in fracture energy, the filler size, and the
filler volume fraction. Nonetheless, Lange’s equation has been found to be inadequate to
describe the crack pinning mechanism in practice [75, 76, 78] except for the case of a
glass filled alumina [79]. In particular, the initial slopes of plots of GΙc against Ds-1 are often
a function of volume fraction [80] and thus the value of TL is not invariant with the particle
size as suggested in (3.46). Consequently, more satisfactory theoretical models have been
proposed.
Evans [81], refined the original concept developed by Lange and calculated the increase in
strain energy necessary to bow out the crack between the particles by a detailed
calculation of TL. Evans assumed the existence of semi-elliptical crack fronts between
particles, and demonstrated that the line energy contribution to the fracture energy
depends on both the particle size and spacing as given by eqn (3.48):
 0 .5
    
0.5
K c  π   π  
(3.48)
=  tan 
K 0  2 + dp / D s    2 + dp / D s  
 
Where Kc and K0 is the critical stress intensity factor of the modified and the unmodified
epoxy, respectively, Ds is the surface-to-surface obstacle spacing, and dp is the particle
diameter. Evans also calculated the ratio of GIc of the composite to that of the resin or the
ratio of the tensile strength of the composite to that of the resin, assuming that the flaw
size is constant. Evans model predicts a non-linear increase in fracture toughness with
increasing filler fraction and the prediction appears to give a better fit to the data of Lange
and Radford [78]. Although, in most analyses performed by Evans, particles were
considered as impenetrable obstacles, he mentioned the importance of the impenetrability
of particles. He proposed the use of an impenetrability factor (ε0), but did not study the
effects of impenetrability, which was instead performed later by Green et al. [82, 83].
 Deformation and fracture behaviour 75

Green and co-workers extended the work of Evans and observed a bowed secondary
crack front using ultrasonic fractography. They have examined, both experimentally and
theoretically, the effects of particle shape, particle fraction, fracture toughness of particles,
particle thermal expansion and particle/matrix adhesion on fracture in model composites of
nickel spheres in a brittle glass matrix. At lower filler contents, the theoretical values are in
very good agreement, however, at higher volume fractions there is a deviation from the
theoretical prediction [84]. This probably arises because the crack front bows out in an
elliptical rather than in a semi-circular manner [83, 85].
Even if the discussed crack pinning theories still exhibits some uncertainties, such as the
different geometries of crack front bowing have never actually been experimentally
confirmed, the different models can be satisfactorily used to explain the fracture toughness
changes with varying volume fraction and size of inorganic particles.

3.3.3 Epoxy toughening with thermoplastics

Since the initial work in the eighties, toughening with engineering thermoplastics has
received much attention in the last decade. Many studies using a variety of different
thermoplastic grades have been published [69, 86-91].
In order to avoid deterioration of the stiffness, the strength and the glass transition
temperature of the cured resins an engineering thermoplastic has to be blended. In
particular polyethersulphone (PES) and polyetherimide (PEI) have been widely used to
produce blends consisting of ductile and tough thermoplastic particles dispersed in highly
crosslinked epoxy formulations. PES and PEI have high glass transition temperatures (210
to 225°C), possess an excellent chemical resistance and are stable at temperatures
<200°C. Because of the high molecular weight of PES and PEI, the amount which can be
dissolved in epoxy resins is rather limited for industrial processes. Hence both
thermoplastic modifiers have to be mixed together with an appropriate solvent and the
solvent has to evaporate before adding the curing agent to the mixture.
Raghava [91] has examined the enhancement in fracture toughness of PES modified
TGDDM-epoxies cured in the presence of aromatic anhydrides. Especially, he studied the
influence of morphology on toughness and observed a two-phase structure and a bimodal
particle size distribution by dynamic mechanical analysis and microscopy. However, the
obtained increase in fracture toughness was only negligible. As an explanation for this
observation the poor interfacial adhesion between the separated PES-particles and the
epoxy matrix was outlined. To overcome this limitation Raghava [89] has blended a low
molecular weight polyethersulphone containing hydroxyl end groups (H-PES) with a
TGDDM/DDS matrix, expecting that the reactive end groups of H-PES can participate in
76 Chapter 3

the chemical reactions during curing of the epoxy and achieve in this way a high level of
interfacial adhesion. The increase in toughness that has been achieved with these two-
phase epoxy-polyethersulphone systems, however, was only marginal. For a 15 phr
H-PES modified network the critical stress intensity factor (KIc) determined at room
temperature increased only by about 0.1 MPam0.5. The obtained results were attributed to
the fact that the highly crosslinked TGDDM-matrix remains brittle, even after inclusion of
low concentration of H-PES, and thus formation of shear bands, the primary energy
dissipation mechanism, is not possible.
Bucknall and Partridge [90] have confirmed the results reported from Raghava by using a
commercial PES (Victrex 100P) to toughen various blends of tetrafunctional (TGDDM) and
trifunctional (TGpAP) epoxy resins cured with diaminodiphenyl-sulphone (DDS) and
dicyandiamide (DICY). They also reported that the PES-toughened TGDDM-resin cured
with both type of hardeners did not show phase separation. In contrast, phase separation
was observed in the case of trifunctional TGpAP epoxy resin. For all studied PES-modified
systems the toughness increase remains a secondary effect.
Based on the available literature it can be concluded that the blends of epoxy resins with
polyethersulphone (PES) up to a critical concentration of 15 wt.% show only modest
improvements in the fracture toughness properties [86, 91-95] compared with the
unmodified systems as summarized in Figure 51. However, if the PES concentration
increases above 15 to 20 phr an increase in fracture toughness can be obtained.

1.8

1.6

1.4

1.2
KIc [MPam ]
0.5

1.0

0.8

0.6
O
0.4 O S
O n

0.2
0 5 10 15 20 25 30
Concentration [wt% PES]

Figure 51: Critical stress intensity factor (KIc) as a function of polyethersulphone (PES)
weight fraction for various highly crosslinked epoxy-systems () ref. [94, 95], () ref. [92],
(%) ref. [92], (1) ref. [94, 95]. All values were measured at room temperature.
 Deformation and fracture behaviour 77

Early toughening studies from Diamont and Moulton [70] showed that thermoplastic
loading of 10 wt% polyetherimide (PEI) can be used as a effective modifier in a TGDDM-
resin, provided that phase separation combined with good interfacial adhesion is obtained.
They observed that a mixture of TGDDM/DDS and PEI built a two-phase structure and,
hence, the fracture toughness was significantly improved (∆KIc=0.24 MPam0.5) compared
with the reference system.
Hourston, Lane and McBeath [96] reported that a rapid increase in fracture toughness
occurs for a TGDDM/DDS-system blended with PEI when the morphology changed from
particulate to co-continuous. Their results demonstrated that a significant increase in KIc by
almost 100% was obtained with a 15 phr addition of PEI.
Chen et al. nitrated a PEI thermoplastic to study the effects of interfacial adhesion of a
TGpAP/DDS system by comparing the toughness of nitrated and un-nitrated PEI at
various concentrations. They observed that, for cured and post-cured samples, KIc
decreased for the nitrated compared with the unmodified PEI. Scanning electron
microscopy also revealed that nitrated PEI was substantially more compatible with the
TGpAP/DDS-system resulting in totally miscible blends. Consequently the obtained
improvements in fracture toughness remain of secondary influence.
Studies on polyetherimide (PEI) modifications of high performance epoxies have reached
similar conclusions than PES modified multifunctional epoxy systems. However,
significantly higher improvements in toughness were obtained in the case of epoxy/PEI
blends. An overview of results from different studies on highly crosslinked epoxy/PEI
blends is given in Figure 52. From comparison of Figure 51 and Figure 52 it can be clearly
seen that PEI modifications exhibited a superior improvement in toughness over PES.

1.6

1.4

1.2
KIc [MPam ]
0.5

1.0

0.8

O CH 3 O
0.6 C O C O C
N N
C CH 3 C
n
O O

0.4
0 5 10 15 20 25 30
Concentration [wt% PEI]

Figure 52: Critical stress intensity factor (KIc) as a function of polyetherimide (PEI) weight
fraction for various highly crosslinked epoxy-systems () ref. [96] () ref. [94, 95], (1) ref.
[70], (^) ref. [94, 95], (<) ref. [97], (i) Ref. [98], (u) ref. All values were measured at
room temperature.
78 Chapter 3

More recently, Boogh and co-workers [99] demonstrated a novel thermoplastic toughening
procedure. They used a low viscosity polyester based hyperbranched polymer (HBP) with
reactive epoxy-endgroups as toughening modifier for bisphenol F resin (DGEBF) cured
with isophorondiamine. By tailoring the epoxy functionality of the hyperbranched modifier,
a chemical induced phase separation (CIPS) was demonstrated resulting in a two-phase
morphology of the cured system. Remarkably, they did not report a significant decrease in
stiffness and glass transition temperature, which they attribute to an almost perfect phase
separation. An outstanding performance of the multifunctional epoxidized hyperbranched
polymers yielded an increased fracture toughness of around 140%.
In contradiction, Wu et al. [100] showed only a moderate increase in fracture toughness for
a HBP modified two-phase epoxy system. They added a commercial hydroxyl-
functionalized hyperbranched polymer to a diglycidyl ether of bisphenol-A (DGEBA) matrix
cured with diaminodiphenylsulfone (DDS) and reported an increase in KΙc of 60%. They
additionally reported, that an increased molecular weight of the HBP (higher generations)
enhanced the effectiveness of the modifier by decreasing the reduction of the glass
transition temperature of the phase-separated composite. The poor interfacial adhesion
between the HBP particles and the matrix, due to the absence of chemical reactions
between the matrix and the HBP-shell, probably explained the differences between these
two studies. However, by a cross-comparison to linear polyesters as impact modifiers, they
stated that the linear counterparts performed essentially as well as the hyperbranched
polymer modifiers at identical molecular weights.
Further studies on thermo-mechanical properties and morphology on third generation,
hydroxyl-functionalized hyperbranched polymers in an epoxy resin have been performed
by Ratna and Simon [101]. Depending on the epoxy monomer used, they observed either
a phase separated morphology for liquid DGEBA, or a homogeneous structure in the
presence of a trifunctional aminoglycidyl (TGpAP) resin. For both matrices the addition of
hyperbranched polymers accelerates the epoxy-amine curing reaction due to the catalytic
effect of the HBP hydroxy end groups. A significant improvement in impact strength was
observed at higher concentrations (>10 wt.%) for both examined epoxy formulations. The
enhancement of toughness was mainly attributed to elastification of the matrix, having a
drawback of lowering the network density and finally reducing the mechanical properties.
Despite the large number of experimental studies, the number of theoretical models that
are able to predict improvements in toughness in thermoplastics-modified epoxies in a
quantitative way is rather limited. The main mechanisms are illustrated in Figure 53.
Enhancement in toughness is often explained in terms of crack pinning [79, 81, 83] or
particle bridging [38, 102] in a manner very similar to that explained for hard particle filled
epoxy. However, in contrast to hard particles, thermoplastic inclusions are able to yield
 Deformation and fracture behaviour 79

and this yielding can provide an additional source of toughness. Even though no model
has been developed so far for predicting the phenomenon, shear-yielding events seems to
be mostly involved in thermoplastic epoxy toughening.

This image is not available online. Please consult the hardcopy thesis
available from the QUT Library.

Figure 53: Mechanisms for toughening of thermoplastic-modified epoxies; (1) crack

pinning, (2) particle bridging, (3) crack path deflection, (4) particle yielding, (5) particle
yielding induced shear banding and (6) micro-cracking [69].

3.4 Experimental

3.4.1 Materials

For details on the used epoxy monomer and hardeners as well as the system composition
the reader it is referred to section 2.2.1.

3.4.1.1 Hyperbranched polymers (HBP) modifier

The hyperbranched polymers used in the present study are commercially available under
the trade name Boltorn (Perstorp, Sweden). The HBP’s molecules used are synthesized
from a tetrafunctional ethoxylated pentaerythritol core (C5H12O4), with an average of five
ethoxy units per pentaerythritol [103], and multiple 2,2-diethylol propionic acid (C5H10O4)
chain extenders or repeating units. An idealized reaction scheme for the acid catalyzed
condensation reaction is shown in Figure 54 and a more comprehensive description of the
synthesis can be found elsewhere [104-107]. The HBP’s are primarily produced by a
divergent synthesis, but to some extent, also through a convergent reaction [108]. The
growth is controlled by stoichiometric ratios of core to chain extenders under carefully
selected processing conditions to avoid side reactions. The resulting polydisperse and
highly branched polymer has a main core shell function and a minor fraction consisting of
COOH-functional branches. A degree of branching is typically about 45% [105].
80 Chapter 3

This image is not available online. Please consult the hardcopy thesis
available

Figure 54: Left: Idealized reaction mechanism of ethoxylated pentaerythritol and 2,2-
diethylol propionic acid (Bis-MPA) for the production of hydroxyl functionalized polyester-
based hyperbranched polymer (e.g. third generation BoltornH-30). Right: A simulated
BoltornH-30 hyperbranched polymer with 32 end groups [109]

The unmodified shell consists mainly of hydroxyl functional groups but the end-groups are
generally accessible for further functionalization as described in literature [110]. In this
study three unmodified hydroxyl functional products (Boltorn®H-20, H-30, and H-40) and
two epoxy functionalized third-generation HBP products (Boltorn®E1 and E2) were utilized
as toughening additives. For the epoxy functionalization of the hydroxyl terminated
precursor product (H-30), a partial epoxidation step was carried out. In all cases, a
minimum of an average of 5%, of the theoretical 32 functional groups in the HBP shell was
remaining hydroxyl groups. The other 95% of the end groups were reacted with 10-
undecenonic-acid by an esterification reaction. Afterwards the oxirane ring was formed by
epoxidation of the conjugation by use of hydrogen peroxide. Further details of the
procedure are described in [111, 112]. A typical property profile of the used aliphatic
hyperbranched polyesters is given in Table 8.
 Deformation and fracture behaviour 81

Table 8: Characteristic properties of hydroxyl- and epoxy-functional hyperbranched

polymers used in this study [113, 114]

This image is not available online. Please consult the hardcopy thesis
available from the QUT Library.

3.4.1.2 Inorganic filler

All investigations on particulate filled systems were carried out with a commercial non-
surface treated white crystalline silica flour of type Millsil-W10 (Quarzwerke, Germany).
The specific physical properties and corresponding scanning electron micrographs are
given in Table 9 and Figure 55. To study the influence of the filler size, different filler
fractions were separated in a sieving tower. The particle size distribution of the silica fillers
were measured by a non-contact liquid sedimentation method using a Horiba CAPA-700
particle size analyzer. Data were obtained from the change in particle concentrations on
the basis of light transmission. For the experimental particle size distribution
measurements, the silica fillers were thoroughly mixed in ethylene glycol (carrier medium)
and sodium hexametaphosphate (dispersion agent) in an ultrasonic bath for 45 minutes.
The concentrations of silica and sodium hexametaphosphate were 0.01 and 0.1 wt%
respectively. Particle size distribution curves were obtained as average curve from at least
five independent measurements.
82 Chapter 3

Table 9: Specific properties of the silica filler Millsil-W10 used [115]

This table is not available online. Please consult the hardcopy thesis
available from the QUT Library.

Figure 55: Scanning electron micrographs of the silica filler Millsil-W10

3.4.2 Sample preparation

The viscosity of the commercial MY721-epoxy monomer was reduced by pre-heating to

65°C, before the premixed hardener-mixture was added and dissolved with a high shear
mixer at 13.500 rpm for 10 minutes to obtain a homogeneous mixture. Afterwards the
mixture was degassed for about 5 minutes at pressures below 100 mbar in order to
remove trapped air. For blend compositions containing silica filler and/or hyperbranched
polymer components, the appropriate weight fractions of the components were added
slowly and mixed under vacuum for an additional 15 minutes.
The final mixture was poured into steel moulds (4 mm thick plates and tensile dumbbells)
that had been preheated at 60°C. Curing was carried out at 60°C for 4 hours in an air-
circulating oven. Then the specimens were demoulded and post-curing was continued at
200°C for 3 hours. After the post-curing period, the oven was switched off and the samples
were allowed to cool back to room temperature slowly in order to minimize internal
stresses.
To study the effects of different filler surface treatments, the silica filler W10 was pre-
treated with γ-glycidoxy-propyl-trimethoxy-silane (Silquest A187, Witco) to promote
adhesion and a commercial siloxane based mold release agent (Aquathan, Block-
 Deformation and fracture behaviour 83

Chemie) to reduce the adhesion to the epoxy resin. The molecular structure of the used
silane adhesion promoter is shown in Figure 56.
The filler particles were treated using the procedure based on that suggested by Trachte
and DiBendetto [116]. 200 g of filler were immersed either in a 5% solution of the Silquest
A187-silane in n-heptane or directly in Aquathan. The suspensions were stirred for 8
hours at 40°C and then filtered, rinsing with n-heptane. The treated filler was then dried at
80°C under vacuum for at least 16 hours.
O
H2C CH CH2 O CH2 3
Si O CH3 3

Figure 56: Silquest A-187 adhesion promoter used

3.4.3 Double torsion test (DT)

The double torsion test (DT) for the evaluation of fracture properties, such as the critical
stress intensity factor and the critical energy release rate, has been increasingly used for
testing brittle materials, including polymers and metals [117-123].
The basic geometry of the double torsion specimen is illustrated in Figure 57. It consists of
a plate in which a machined V-groove runs a length of the test piece and is centered at
w/2. The V-groove provides a route for crack propagation through the centre of the
sample. Each of these beams is loaded in torsion, usually by a four-point loading at the
ends, causing the crack to propagate down the centre of the specimen leading to mode-Ι
fracture. The plate is normally notched, or side grooved, on one side to help constrain the
propagating crack to the centre of the specimen. In this study sharp cracks were
introduced by applying controlled force to a very sharp razor blade, which induced a sharp
natural crack ahead of the razor cut.

Figure 57: Schematic representation of the Double-Torsion (DT) test configuration.

84 Chapter 3

The crack does not propagate directly across the thickness of the sample, but extends
further along the tensile side of the plate to form a curved crack front as shown in Figure
58.

Figure 58: Typical curved front crack profile for the Double Torsion test specimen.

The critical stress intensity factor (KIC) and the critical strain energy release rate (GIC) of the
materials were experimentally determined using the double torsion (DT) geometry (cf.
Figure 57) on a Z100-Zwick universal testing machine. The specimens were loaded in
compression, with various rates of crosshead displacements ranging from 0.01 to
10 mm/min at several temperatures between –40 and 200°C. For all performed
experiments the test specimens were isothermal conditioned for one hour at the
appropriate test temperature and KIC was calculated for plane strain conditions according
to eqn (3.49) as proposed in [124]:
0.5
 (1 + ν ) ⋅ 3  (3.49)
K Ic = Fmax Wm  
1000 ⋅ w ⋅ t ⋅ a 
3

where Fmax denotes the fracture load; Wm is the moment arm (13 mm); w is the specimen
width (∼34 mm); t is the specimen thickness (∼4 mm); a represents the specimen
thickness in the plane of the crack (∼3 mm).
In addition, GIC was calculated from the stress intensity values using the relationship from
Mations and Small [125] as given in eqn (3.50):
K 2Ιc (3.50)
G Ιc =
E(1 − ν )
where ν is the Poisson’s ratio (taken as 0.4 for unfilled resins and 0.3 for filled resins), and
E is the Young’s modulus (determined from tensile tests).
In all figures the given results were based on the arithmetic mean of at least 5 measured
values and the error bar indicates the standard deviation. The 4 mm plate thickness of the
DT-specimens was chosen to be well within the requirements of ASTM E399-83 [16] for
plane-strain conditions.
Depending upon composition of the material and the external test conditions, such as test
temperature, strain rate etc. two principal modes of crack propagation can be distinguished
by the profile of the related DT-load/displacement trace:
 Deformation and fracture behaviour 85

a) stable, brittle crack growth, which may occur at a constant load, with the rate of crack
propagation being dependent upon the rate of crosshead displacement. A typical load
curve is shown in Figure 59(a).
b) unstable, brittle crack propagation may occur in a “stick-slip” manner as shown in Figure
59(b). For unstable (slip-stick) crack propagation, maximum and minimum values of KIC
and GIC, corresponding to initiation and arrest of crack propagation, are given.

Initiation of crack a) Initiation of crack b)

extension extension
Load

crack
arrest

Displacement
Figure 59: Typical Double Torsion-load/displacement curves for (a) stable continuous
crack extension; (b) unstable stick-slip crack extension

3.4.4 Tensile test

The tensile strength (σrm) and the Young’s modulus (E) of the unfilled and filled samples
were determined in a tensile test according to standard DIN-EN-ISO 527-1 [126] on a
mechanical Zwick universal testing machine type Z100. It was equipped with a
temperature chamber capable of temperatures ranging from –40 to 200°C. For all
performed experiments the test specimens (dumbbell type-A) were isothermally
conditioned for one hour at the appropriate test temperature. Tensile tests were performed
at strain rates varying between 0,1 to 50 mm/min. In all figures the given value is the
arithmetic mean and the error bar represents the standard deviation of at least 5 measured
samples.

3.4.5 Compression test

Compression tests were carried out on a mechanical Zwick universal testing machine type
Z100 according to the standard DIN-EN-ISO 604 [127]. For all performed experiments the
cylindrical test specimens (∅=20 mm, height=40 mm) were isothermally conditioned for
one hour at the appropriate test temperature. Compression tests were performed at strain
rates of 0,1 and 10 mm/min and temperatures at 50, 100, 150 and 200°C. A lathe was
used to ensure a perpendicular cut and the specimens were lubricated with silicone grease
before they were compressed between parallel plates. In all figures the given value is the
86 Chapter 3

arithmetic mean and the error bar represents the standard deviation of at least 5 measured
samples.

3.4.6 Dynamic mechanical analysis (DMA)

The dynamic mechanical analysis was performed using a Perkin-Elmer DMA-7e analyzer
equipped with a three point bending assembly (15 mm span length). Bending-bars of
18 x 4 x 2 mm3 were cast into a steel mould and cured at room temperature. The DMA
was operated in the temperature-scanning mode with a constant displacement amplitude
and frequency. The frequency was fixed at f=1 Hz using a strain amplitude of ±10 µm. The
temperature ramp covered a range from 30 to 280°C with a continuous heating rate of
3°C/min under helium atmosphere.

3.4.7 Microscopy
Microscopic characterization of the inorganic particle size and distribution, as well as the
fracture surface of a large number of specimens was examined using scanning electron
microscopy (SEM). All SEM-experiments were carried out on a CAMSCAN microscope.
Fractured samples were mounted on an aluminum sample holder using a conductive paint.
The specimens were sputtered with a thin Au-Pd layer (∼20 nm) to reduce surface
charging. SEM-micrographs were obtained under conventional secondary electron
imaging conditions and an accelerating voltage of 20 kV.
Optical microscopy (OM) was mainly used to examine sub-surface damage in compressed
samples. Therefore, thin-microtomes with a thickness from 20 to 100 µm were prepared
using a polishing and sectioning technique as proposed by Smith [77]. Afterwards, all
samples were examined in transmitted mode using a Zeiss-Axiolab microscope in
polarized or cross-polarized light. In some cases a quarter wave plate was used to
visualize the contrast of the birefringent regions.
 Deformation and fracture behaviour 87

3.5 Results & discussion

3.5.1 Effects of strain rate and temperature on the tensile deformation

The changes of the tensile strength (σrm) as a function of temperature and strain rate for
the unfilled and silica-filled MY721/MA/HHPA-system are shown in Figure 60. It can be
seen that the tensile strength for both systems tends to increase with increasing strain rate
and decreasing test temperature, providing evidence that the results for different
temperatures could be superimposed with the applied strain rates. Furthermore, it can be
seen that the silica filled system exhibits slightly higher absolute tensile strength values.

100
a)
0.1 mm/min
1 mm/min
80
10 mm/min
Tensile strength [MPa]

50 mm/min

60

40

20

0
0 25 50 75 100 125 150 175 200 225
Temperature [°C]

100
b) 0,1 mm/min
1 mm/min
10 mm/min
80 50 mm/min
Tensile strength [MPa]

60

40

20

0
0 25 50 75 100 125 150 175 200 225
Temperature [°C]

Figure 60: Tensile strength (σrm) as a function of strain rate and temperature for (a) unfilled
MY721/MA/HHPA-system; (b) 60% w/w silica filled MY721/MA/HHPA-system.
88 Chapter 3

The influence of the test temperature and loading rate on the initiation and development of
the fracture process under tension load was studied more in detail for the neat
MY721/MA/HHPA-system. A sub-critical and smooth zone, as shown on the fracture
surface of the tensile bars in Figure 61 and 62, increases with higher temperatures and
lower strain rates. These observed effects agree well with earlier observations by Cantwell
et al. [128].

a) b)

c) d)
Figure 61: Optical micrographs of the fracture plane as a function test temperature for the
neat MY721/MA/HHPA-system tested with a constant strain rate of 0.1 mm/min under
tension load; (a) 23°C, (b) 100°C, (c) 150°C, (d) 200°C.

a) b)

c) d)
Figure 62: Optical micrographs of the fracture plane as a function of strain rate for the neat
MY721/MA/HHPA-system tested at 150°C under tension load; (a) 50 mm/min, (b) 10
mm/min, (c) 1 mm/min, (d) 0.1 mm/min

A more detailed examination of the fracture surface highlighted three distinct zones or
regions as shown schematically in Figure 63. The irreversible damage initiates under
normal circumstances at some internal defects or from regions corresponding to an area of
increased stress concentration, i.e. the edge of the specimen. After the crack initiation at
an internal defect, the crack followed slow crack propagation in the sub-critical zone,
 Deformation and fracture behaviour 89

before it accelerates into the smooth zone and finally reaches the highest crack
propagation velocity in the rough zone [6].

Figure 63: Typical fracture surface of silica filled epoxy tensile bar.

Analysis of the microstructure by means of optical microscopy techniques revealed that

none of the tested specimen yielded in tension and all samples fractured in brittle manner.
The absence of any yielding can be theoretically explained by the fact that the specific
fracture stress is lower than the intrinsic material’s yield stress under the investigated
temperature/strain rate conditions. Furthermore, Griffith’s defect theory [5] assumes that
many more defects are active under loading in tension than in compression mode, which
explains that yielding is generally much more difficult to observe in tension mode.

3.5.2 Effect of the filler fraction on the tensile deformation

Inorganic fillers are commonly used to modify various properties of epoxy polymers, or
simply as extender filler for lowering the price/volume ratio of the material. In any case, it is
essential to know the dependence of the material composition on the various properties. In
particular the tensile strength (σrm), Young’s modulus (E), and yield stress (σy) are of
primary importance for technical applications.
Figure 64(a) shows the tensile strength (σrm) as a function of the silica-filler fraction. It can
be seen that the MY721/MA/HHPA-system does not exhibit any significant dependency of
the ultimate tensile strength on the filler content between 0 to 50 vol.%, which is consistent
with previously performed work on different epoxy systems [129]. Therefore, it can be
assumed that the critical debonding strength or strain of the investigated system remains
almost constant and independent of the filler content.
Additionally, the obtained σrm-data fit reasonably well within the upper boundary of a
simple mathematical prediction as proposed by Nicolais and Nicodemo [130], which is
given by eqns (3.51) and (3.52) and also depicted in Figure 64(a).
For the upper boundary condition:
90 Chapter 3

σ rm = σp (3.51)

For the lower boundary condition:

(
σ rm = σ p 1− 1.21 Vf
2/3
) (3.52)

where σp is tensile strength of the polymer matrix, and Vf denotes the volume fraction of
inorganic filler.
In agreement with several other studies [131-133], the Young’s modulus, as obtained from
the initial slope of the tensile stress-strain curves, gradually increases with increasing
inorganic particle content as shown in Figure 64(b). Many different theoretical predictions
have been proposed to determine the elastic modulus of particulate filled composites in
the past. One of the most widely accepted model is the theoretical approach proposed by
Ishai and Cohen [134]. Ishai and Cohen’s equations originally derived from a two-phase
model of a cubic matrix containing spherical inclusions and assuming either a uniform
stress eqn (3.53) or a uniform strain eqn (3.54) condition (i.e. comparable to a series and
parallel model in fiber composite theories).

1 + (m − 1)Vf
2/ 3
Ec (3.53)
φf = =
(
Em 1 + (m − 1) Vf 2 / 3 − Vf )
Vf
φf = 1 + (3.54)
m / (m − 1) − Vf
1/ 3

where
m = Ef/Em
Ef = Young’s modulus of the filler
Em = Young’s modulus of the matrix
Ec = Young’s modulus of the composite
Vf = Volume fraction of the filler

In Figure 64(b), the theoretical predictions from eqns (3.53) and (3.54) are compared to
the experimentally obtained values from the tensile test. A fairly good agreement between
the theoretical predictions assuming uniform strain conditions in the material and the
experimental data can be seen.
 Deformation and fracture behaviour 91

a)
100

Upper bound
Lower bound
80

Tensile Strength [MPa] 60

40

20

0
0 10 20 30 40 50 60

Volume Fraction [%]

b) 20

uniform stress
Young's modulus [GPa]

15

10

uniform displacement
5

0
0 10 20 30 40 50 60
Silica volume fraction [%]

Figure 64: Influence of the silica volume fraction in the MY721/MA/HHPA-system on (a)
tensile strength (σrm) and (b) Young’s modulus (E). All tensile measurements were
performed at 23°C with a constant strain rate of 1 mm/min

3.5.3 Effect of strain rate and temperature on the compressive deformation

The behaviour of the ultimate compression strength as a function of temperature and
strain rate for the unfilled and silica filled MY721/MA/HHPA-system is shown in Figure 65.
92 Chapter 3

a) 200

180 0.1mm/min

Compression strength [MPa]

10mm/min
160

140

120

100

80

60

40

20

0
0 50 100 150 200 250
Temperature [°C]

b) 300

0.1mm/min
Compression strength [MPa]

250 10mm/min

200

150

100

50

0 50 100 150 200

Temperature [°C]

Figure 65: Compression strength (σrm) as a function of temperature and strain rate for (a)
unfilled and (b) 60% w/w silica-filled MY721/MA/HHPA-system.

In contrast to experiments performed under tension load, where only brittle fracture was
observed, the load displacement curves in compression mode conditions were non-linear.
This indicates that a substantial plasticization process occurred in the investigated
temperature range between 23°C to 200°C. The curves for both the unfilled and the silica
filled-system exhibited a peak stress (yield stress, σy) and a post yield-softening
phenomenon before fracture occurred in a brittle manner (cf. small stress-strain curves in
Figure 65). Moreover, both investigated systems show a decrease of the elastic Young’s
modulus and the maximum compression strength with increasing temperature. This is a
consequence of an increased molecular mobility on the scale of molecular chains and
local groups at elevated temperatures; hence, the resistance to deformation becomes
 Deformation and fracture behaviour 93

weaker. In particular, for the neat resin system the elongation at break is significantly
increasing with increased temperature. On the other hand, the elongation at break remains
nearly unaltered in the studied temperature range for the silica filled system.
Based on optical microscopy investigations under polarized light, it was possible to identify
the primary inelastic micro-deformation mechanisms. It was primarily established that
shear processes, which are increasing with increased temperatures, are responsible for
relatively large strains in the glassy epoxy polymers. As expected the strain softening
behaviour promoted the localized shear yielding in the form of shear bands preferably at
stress inhomogeneities, e.g. the silica obstacles. These shear processes include both
diffuse shear yielding and localized shear band formations, and occur without loss of
intermolecular cohesion in the polymer, so that they finally produce only little, if any
change in volume. In agreement with other studies, cavitation processes, e.g. crazing,
have not been observed [6, 135-137]. It is hypothesized that the large number of
crosslinking points and the high entanglement density can effectively restrain crazing in
the studied MY721/MA/HHPA-system.
The observed bulk shear bands produce the classic shear offset at the edges of the
unfilled specimen. A typical shear band formation of the neat MY721/MA/HHPA
compression specimen is shown in Figure 66. It is obvious that the coarse band is
approximately at an angle of 45° to the applied stress. Conducting similar tests with the
filled epoxy was more difficult due to the non-transparent nature of the system. However,
some evidence was found from the thin sectioning and microscopic analysis to propose
that shear yielding is also the primary deformation mechanism in the filled resin.

a) b)
Figure 66: Transmission optical micrograph of a thin-section (25 µm) of a neat
MY721/MA/HHPA specimen compressed at 200°C (a) non-polarized, transmitted light, (b)
same specimen viewed between two crossed polarizers. The compression direction is
horizontal as indicated by the arrows.

For the filled system the first visual indication of irreversible damage occurs at the yield
point (σy) where the stress-strain curve shows a maximum. Beyond this characteristic
stress point, the highly filled material begins to stress-whiten reflecting the initiation of an
irreversible damage processes. The stress-whitening phenomenon occurs in a diffuse
94 Chapter 3

band that runs diagonally across the specimen, i.e. along the direction of maximum shear
stress angle of 45°. At higher strains (∼3%) the band becomes visually denser (whiter) and
wider. It was observed that all compression specimens were severely whitened,
independent of the test temperature and the rate of deformation. Typical examples of
stress whitened and fractured compression specimens are shown in Figure 67. In
particular the sample in Figure 67(a), where a specimen was loaded to 65% of the ultimate
load at 150°C, two intersecting stress-whitened bands extending at approximately 45° can
be clearly seen. The macroscopic stress-whitening effect is mainly provoked by the
presence of new surfaces or voids, which scatter light within the solid. Considerations on
the microscopic scale reveal that primarily particle-matrix debonding with minor
contributions due to particle fracture and inelastic matrix deformation occurred in the
particulate filled systems.

a) b) c) d)

Figure 67: Silica-filled (60% w/w) MY721/MA/HHPA samples of uniaxially compressed

specimen at 150°C. The different deformation stages are; (a) 65% ultimate load deformed
specimen showing stress-whitening before failure; (b-d) specimen deformed to failure
showing fracture at the dead-zone deformation-zone interface

3.5.4 Characterization of the fracture toughness behaviour

3.5.4.1 Effect of the test temperature

Glassy polymers such as epoxies are generally considered to be inherently brittle
materials when tested at temperatures well below their glass transition temperature. In this
section a series of fracture toughness tests at different temperatures were undertaken in
order to determine the effect of the test temperature on the fracture toughness behaviour
of the MY721/MA/HHPA-system. In the examined temperature range between
 Deformation and fracture behaviour 95

-40°C to 200°C all samples failed predominantly in an unstable crack propagation manner
as shown by the corresponding DT-load/displacement curves in Figure 68. However, it can
be seen that with decreasing test temperature the tendency to a mixed crack propagation
appears, see Figure 68.

200

-40°C 25°C
150

100

50
Load [N]

150
100°C 200°C

100

50

0
0.00 0.25 0.50 0.75 0.00 0.25 0.50 0.75 1.00

Displacement [mm]

Figure 68: DT-load/displacement curves of the silica-filled (60% w/w) MY721/MA/HHPA-

system (r=0.8) as a function of test temperature. All DT-tests were performed with a fixed
strain rate of 0.1 mm/min.

Moreover, variations of the test temperature significantly affect the characteristic fracture
toughness values KIc and GIc as shown in Figure 69. Both, the neat and the silica filled
formulation, exhibited an almost identical KIc- and GIc-temperature behaviour, i.e. with
increasing temperature the material becomes more brittle resulting in a reduction of the KIc
and GIc-values. In particular, between 60 and 100°C a substantial decrease in the fracture
toughness, KIc, was observed.

a) 2.0 b) 400
1.8 SiO2 filled SiO2 filled
350
neat resin neat resin
1.6
300
1.4
KIC [MPa m ]
0.5

1.2 250
GIC [J/m ]
2

1.0 200
0.8
150
0.6
100
0.4
0.2 50

0.0 0
-40 -20 0 20 40 60 80 100 120 140 160 180 200 220 -40 -20 0 20 40 60 80 100 120 140 160 180 200 220
Temperature [°C] Temperatur [°C]

Figure 69: Variations of (a) critical stress intensity factor (KIc), and (b) critical strain energy
release rate (GIc) as a function of test temperature. The silica fraction of the filled
MY721/MA/HHPA-system (r=0.8) was fixed to 60% w/w. All DT-tests were performed at a
strain rate of 0.1 mm/min.
96 Chapter 3

This observation is very remarkable because a continuous increase in the fracture

toughness up to the ultimate glass transition temperature was generally expected and
previously reported for different TGDDM-based epoxy formulations [71]. Moreover, the
Ludwig-Davidenkov-Orowan (LDO) theory predicts an increase in the fracture toughness
properties with increasing test temperature [138, 139], because the yielding stress
decreases with increasing temperature and hence micro-deformation processes will be
favored and consequently can occur before fracture. This finally leads to an increase of
fracture toughness with increasing temperature. However, for the studied
MY721/MA/HHPAsystem the opposite behaviour, that means with increasing yield stress
the material becomes tougher, was observed (Figure 70). As explanation for the observed
behaviour, it is postulated that at elevated temperature the fundamental principles of the
linear elastic fracture mechanics (LEFM) are not any longer valid, because the LEFM can
only be applied at temperatures clearly below the glass transition temperature where plain
strain condition can be applied. As soon as the test temperature increases above the glass
transition temperature of the material the viscoelastic material properties become more
dominant and thus a sharp decrease of the fracture toughness, i.e. KIc, appears. Moreover
a broadening of the Tg region may occur due to an inhomogeneous network structure.

1.6

1.4

1.2
KIC [MPa m ]
0.5

1.0

0.8

0.6

0.4

0.2
80 100 120 140 160 180 200 220 240
σy [MPa]

Figure 70: Critical stress intensity factor (KIc) as function of compressive yield stress (σy)
for the silica-filled (60% w/w) MY721/MA/HHPA (r=0.8).

Additionally performed experiments on a same set of samples verified that the observed
reduction of the KIc-values (~50%) between room temperature and 200°C is a reversible
process. This fact provides evidence that irreversible material changes, e.g. post-curing
processes at higher temperature, can be unambiguously excluded. Therefore, the
experimentally observed embrittlement with increasing temperature can be most probably
explained by the presence of an inhomogeneous multiphase-network consisting of an
independent MY721/MA and a MY721/HHPA sub-network. Moreover, it is proposed that
 Deformation and fracture behaviour 97

most probably the MY721/HPPA sub-network shows an underlying primary relaxation, i.e.
a glass transition temperature, between 60 to 100°C.
To obtain more evidence for the hypothesis of an inhomogeneous network structure
further fracture tests have been performed. The MY721-resin was cured with only pure MA
(48 phr) and pure HHPA (32 phr), reflecting the equivalent hardener weight fraction as
used in the co-anhydride mixture. As shown in Figure 71(a), the KIc of the MA-cured
system is clearly independent of temperature as long as the material is in its glassy state.
This results in an almost constant KIc-value until the test temperature finally reached the
glass transition at around 180°C. On the other hand, the MY721/HHPA-system (Figure
71(b)) exhibited a continuous increase of the fracture toughness with increasing
temperature until the glass transition temperature at 64°C. At temperatures above this
transition area a sharp decrease of experimental KIc-data was induced by a change of the
material from the glassy state to a highly viscoelastic material.
Assuming that both used anhydrides reacted independently from each other, a simple
mixing rule that considered the molar co-anhydride proportions of the MA/HHPA mixture
can be applied as shown in Figure 71(c). It can be seen that the experimental KIc-data
showed a fairly good agreement with the trend predicted by this simple mixing rule. These
observations provide some additional indications that confirm the existence of a multi-
phase network structure.

0.8

0.6

0.4

0.2 Tg

a)
0.0

0.6
KIC [MPa m ]
0.5

0.4

0.2 Tg
b)
0.0

0.6
Tg
0.4

0.2
c)
0.0
-50 0 50 100 150 200
Temperature [°C]

Figure 71: Variation of the critical stress intensity factor (KIc) as a function of temperature
for (a) MY721/MA (100/48) (b) MY721/HHPA (100/32) and (c) MY721/MA/HHPA
(100/48/32). The dotted line represents the trend as predicted by a mixing rule considering
the molar ratio of MA/HHPA=(0.7/0.3). All DT-tests were performed at a constant strain
rate of 0.1 mm/min.
98 Chapter 3

3.5.4.2 Effect of the strain rate

Since polymeric materials are generally viscoelastic by nature, the effects induced by
increasing test temperature are often analogous to those produced by decreasing the
strain rate. To determine the effect of strain variations on the fracture behaviour of the neat
and silica filled MY721/MA/HHPA-system Double Torsion (DT) fracture tests have been
performed at room temperature.
The obtained DT-load/displacement curves exhibited a “saw-tooth” appearance for both
studied systems up to a strain rate of less than 10 mm/min. This also called “slip-stick”
crack mode is the unambiguous indication of unstable crack propagation and allows the
determination of the critical stress intensity factor for crack initiation (KIci) and crack arrest
(KIca) [140-144]. Increasing the strain rate to 10 mm/min results in a different DT-
load/displacement curve appearance, which corresponds to a change of the intrinsic crack
propagation mode. The crack propagation under higher strain rate conditions occurs
almost at a constant load until the final fracture of the sample. This type of continuous
crack growth is generally described as stable, brittle crack mode and is favored by higher
displacement rates [136, 145-147].
In Figure 72 the effects of a variable strain rate on the critical stress intensity factor is
shown. It can be seen that an almost 10% increase in fracture toughness was obtained
over a range of approximately three decades of strain rates. These results are in good
agreement with the previously discussed mechanical tensile and compression behaviour
that exhibited similar behaviour (cf. chapter 3.5.1and 3.5.3). The determined KIci-values
slightly decreased with increasing strain rate. On the other hand the corresponding KIca
values remain on an almost constant level and thus strain rate independent.
The first qualitative explanation for the obtained “slip-stick” crack propagation has been
given by Phillips and co-workers [146, 147]. They explained the observed “slip-stick”
phenomenon in terms of crack tip blunting during re-loading after crack arrest due to the
ability of the material to deform plastically in the vicinity of the crack tip. At a certain
increased load a new sharp crack initiates and extends rapidly until the load corresponds
to the stress intensity value for crack arrest. Later on scanning electron microscopy studies
[148] and also in-situ video recording of scanning electron microscope experiments [149]
have confirmed this theory.
 Deformation and fracture behaviour 99

1.6

KIC [MPa m ]
0.5
1.4 SiO2 filled initiation
SiO2 filled arrest
neat resin initiation
neat resin arrest
0.6

0.4
0.01 0.1 1 10
Strain rate [mm/min]

Figure 72: Variation of the critical stress intensity factor for crack initiation (KIci) and crack
arrest (KIca) as a function of the strain rate. The silica content of the filled
MY721/MA/HHPA-system was fixed to 60% w/w. All DT-tests were performed at room
temperature.

3.5.4.3 Effect of the inorganic filler fraction

Unmodified epoxy resins in the uncured state have relatively high viscosities when
compared with other thermosets, i.e. polyesters. Thus, for most technical systems the
maximum volume fraction of fillers, which can be added to the resin, is limited to 50% w/w
without impeding the flow characteristics of the composite. Moreover, it becomes
increasingly difficult to degas the mixture to obtain a void free system. On the other hand
at volume fractions below 30% w/w sedimentation effects could occur resulting in an
inhomogeneous filler distribution in the material. In order to produce homogeneous
castings, the mixture containing a filler content less than 30% w/w were cast just prior to
gelation as described elsewhere [150]. Afterwards a DT-fracture test was performed and
the fracture toughness properties were determined. From the corresponding DT-
load/displacement curves (cf. Figure 73) it can be seen that all examined samples
exhibited continuous crack propagation independent of the applied silica volume fraction.
Furthermore, a constant increase of the critical load, associate to the maximum force
before the crack started to propagate, was obtained by increasing filler fraction.
100 Chapter 3

175

150 a) b) c)

125

100

75

50

25

0
Load [N]

150 d) e) f)
125

100

75

50

25

0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 0.5

Displacement [mm]

Figure 73: DT-load/displacement curves of the MY721/MA/HHPA-system as a function of

silica weight fraction (a) unfilled, (b) 10%, (c) 20%, (d) 40%, (e) 50%, (f) 60%. All DT-tests
were performed at room temperature with a strain rate of 0.1 mm/min.

The specific fracture toughness, expressed here as the critical stress intensity factor (KIc)
and the critical energy release rate (GIc), of the silica filled epoxies are significantly
affected, see Figure 74. It can be seen that the KIc-values of the investigated system
increase gradually with increasing filler fraction. This behaviour is in agreement with
previous studies [78, 80, 85, 129, 145, 151, 152]. However, most of this research work
concentrated on the effects of volume fraction upon the fracture energy in term of GIc [78,
80, 152, 153]. This has led to a more complex analysis, since the filler content has a
strong influence on the Young’s modulus (E) of the composite as already discussed in
section 3.5.2.
In this study the critical strain energy release rate (GIc) was computed from eqn (3.50) by
using the KIc (Figure 74(a)) and the appropriate modulus data (Figure 64(a)) as determined
before. The calculated dependency of GIc as a function of the filler fraction is illustrated in
Figure 74(b). One can see that the general trend of the obtained GIc-behaviour is different
to the KIc-evolution. The initial GIc-value of fracture energy shows a maximum at a filler
weight fraction of ∼0.4. This observed peak behaviour has been a focus of a number of
discussions in the past. Some explained the observed behaviour with a change in failure
mechanism [153]. However, another reasonable explanation for the difference curve trend
between the KIc and the GIc response can be given by considering their final mutual
relationship as given by eqn (3.50). For particulate filled materials, the elastic modulus (E)
 Deformation and fracture behaviour 101

generally increases more rapidly than KIc with a comparable volume fraction of inorganic
particles [145]. Consequently, GIc has the common ability to decrease even while KIc
increases.

a) 2.0 b) 300

1.8
250
1.6

1.4
KΙC [MPa m ]
0.5

200
1.2

GIC [J/m ]
2
1.0 150

0.8
100
0.6

0.4
50
0.2

0.0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Filler fraction [wt. %] Filler fraction [wt. %]

Figure 74: (a) Critical stress intensity factor (KIc) and (b) critical strain energy release rate
(GIc) as a function of silica weight fraction for the MY721/MA/HHPA-system. All DT-tests
were performed at room temperature with a constant strain rate of 0.1mm/min.

The fracture surfaces of ruptured samples were examined using SEM-technique, see
Figure 75. From the micrographs of the neat resin system (Figure 75(a)) one can clearly
see the smooth fracture surface with cracks in different planes. This unambiguously
indicates the brittle fracture nature of the unmodified epoxy network, which accounts for its
poor toughness (KIc≈0.6 MPam0.5) and limited deformation ability. For low silica volume
fractions it can be seen that the particles are widely spaced, while with increased filler
fraction the interparticle distance decreased and the connectivity significantly increased.
Various theories concerned with the mechanisms involved in the particulate toughening
have been reviewed in the introduction section. It is proposed that crack pinning (cf.
3.3.2.1) is the dominant mechanism for the obtained increase of fracture toughness, even
when none of the filled samples exhibited a typical tail structure behind each particle, as
expected for the crack pinning mechanism. However, the absence of these tails does not
necessarily mean that no crack pinning has occurred, because the irregular filler shape
and the relatively high filler content may produce considerable overlap of the secondary
cracks. The simple fact that the surface roughness increases with increasing filler content
indicates already an extension of the crack path.
102 Chapter 3

a) b)

c) d)

e) f)
Figure 75: Scanning electron micrographs of the fracture surface topology as a function of
the silica weight fraction; (a) unfilled, (b) 10%, (c) 20%, (d) 40%, (e) 50%, (f) 60%.

A comparison of the theoretical models and the experimental values is given in Figure 76,
where the reduced critical stress intensity factor (KIc/KIc0) is plotted against the ratio
between the average particle size (dp) and the interparticle distance (Ds) as calculated
according to eqn (3.47). The different lines represent theoretically calculated data as
predicted by Green’s [83] and Evans’ [81] theories.
It can be seen that the agreement between Evans’ theory and the experimental data is
fairly good. The fact that the microscopic investigation of the fracture surfaces showed a
considerable breakdown of the particle/matrix interface underline the observation that the
crack bowing mechanism might be more difficult to occur. On the other hand this
observation gives a reasonable explanation for the fact that the measured data are
generally lower than the predicted values.
Deformation and fracture behaviour 103

4.0

3.5

3.0

2.5

KΙC / KΙC0 2.0

1.5
Green (non-interacting)
Green (interacting)
1.0
Evans (non-interacting)
Evans (interacting)
0.5 experiment

0.0
0.0 0.5 1.0 1.5 2.0
dP / DS

Figure 76: Variation of (KIc/KIco) as a function of (dp/Ds) for different silica-filler fractions.
The full lines represent the theoretical predictions of Green [83] and Evans [81] for
interacting and non-interacting elliptical cracks. Symbols denote the experimentally
obtained values of the DT-test.

3.5.4.4 Effect of the inorganic filler size

To study the influence of the silica filler size on the deformation and fracture behaviour of
filled epoxies, silica filler grades with five different particle size distributions were used.
Typical particle size distributions and the corresponding SEM-micrographs of the classified
fillers are shown in Figure 77.
Unfortunately, a generally accepted prediction of the stiffness and toughness properties as
a function of the particle size is not available hitherto, because most of the developed
theories explain the variations of the mechanical behaviour of polymeric composites
without considering the particle size and their size-distribution. However, from Figure 78(a)
it can be clearly seen that there is an increase of the tensile Young’s modulus with
increasing the average particle size. The elastic modulus of the investigated
MY721/MA/HHPA-composites increased by ∼35% with increasing the average particle
size from 2 to 160 µm. The distinctly affected stiffness of the samples with a relatively
small average particle size, i.e. 2 µm, can be most probably to the poor wetting behaviour
of smaller particles, due to the fact that the greater surface area considerably increases
the final mixture viscosity.
104 Chapter 3

a) 60 100

50
80
Fractions [weight %]

Total sum [%]

40
60

30
40
20

20
10

0 0
0 2 4 6 8 10
Particle size [microns] a)

b)

40 100

80
Fractions [weight %]

30
Total Sum [%]
60

20
40

10
20

0 0
0 5 10 15 20 25
Particle size [microns]
b)

30
c)
100

80
Fractions [weight %]

Total sum [%]

20

60

40
10

20

0 0
0 10 20 30 40 50 60 70 80
Particle size [microns]
c)

d) 60 100

50
80
Fractions [weight %]

Total sum [%]

40
60

30
40
20

20
10

0 0
0 50 100 150 200 250 300
Particle size [microns] d)

Figure 77: Particle size distribution and the corresponding SEM-micrographs of different
used silica filler grades with an average particle size of (a) 2 µm, (b) 7 µm, (c) 38 µm, (d)
160 µm
Deformation and fracture behaviour 105

Unlike the particle volume fraction, the varying average particle size does not significantly
affect the fracture toughness behaviour above a threshold value of about 10 µm, see
Figure 78(b) and (c). Similar to the range of particle sizes investigated by Spandoukis and
Young [145], the experiments performed on the silica filled MY721/MA/HHPA-systems
showed that varying the particle size between 7 and 160 µm did not appreciably change
the fracture toughness behaviour, and hence the experimental KIc and GIc-data remain
almost on a constant level. Moreover, all tested samples showed stable crack propagation
independent of the particle sizes.
However, it is clear that there is a lower and upper boundary in particle size where the
mechanical composite-properties will be adversely affected due to the higher probability of
a flaw being present within the particle or the matrix.

a) 20.0

17.5

15.0
Young's modulus [GPa]

12.5

10.0

7.5

5.0

2.5

0.0
0 10 20 30 40 140 160
Average particle size [µm]

b) c) 300
2.0

1.8
250
1.6

1.4 200
KΙC [MPa m ]
0.5

GΙ C [J/m ]
2

1.2

1.0 150

0.8
100
0.6

0.4
50
0.2

0.0 0
0 10 20 30 40 140 160 0 10 20 30 40 140 160
Average particle size [µm] Average particle size [µm]

Figure 78: (a) Tensile Young’s modulus (E), (b) critical stress intensity factor (KIc), (c)
critical energy release rate (GIc) as a function of average silica particle size. For all
samples a fixed silica filler fraction of 60% w/w and a constant strain rate of 0.1 mm/min
was used.
106 Chapter 3

3.5.4.5 Effect of the filler/matrix adhesion

In order to investigate the effects of the resin-filler adhesion on the composite mechanical
properties different filler pre-treatments were studied. The silica particles were treated with
a mould release agent Aquathan (Block-Chemie, Germany), which should effectively
reduce the filler-resin adhesion. Furthermore, a commercial silane-coupling agent
γ-glycidoxy–propyltriethoxysilane, Silquest-A187 (Witco, Switzerland) was applied to
enhance the interface adhesion between the matrix and filler. Despite the differences in
particle adhesion, as shown by the appearance of the fracture surfaces, it was observed
that the different silica pre-treatments had only a minor effect on the tensile modulus of the
composite as shown in Figure 79(a). Similarly, one can also observe that the filler-matrix
adhesion does not significantly change the fracture toughness properties of the
composites (Figure 79(b-c)). The KIc and GIc-values of the two different pre-treated fillers
varied only within the normal scatter range. Recorded DT-load/ displacement curves
revealed, that in case of poorly bonded silica-particles unstable crack propagation was
obtained, whereas stable propagation was reached for well-bonded silica particles. This
observations are consistent with previously performed studies by Spandoukis and Young
[145, 151], who used glass beads treated with a mould release agent and a silane
coupling agent and observed that the surface treatment had only a little effect on the
composite modulus and fracture behaviour. It follows that very good interfacial bonding is
not absolutely necessary to reach high fracture toughness properties.
The filler surface treatment, however, has a significant influence on the maximum
mechanical strength of the studied samples. Well-bonded filler provides composites with
notably higher tensile strength values (σrm=84.2±6.9 MPa) compared to poorly bonded
particles (σrm=28.9±8.7 MPa). This observation may be explained by the fact that weakly
bonded particles act as sources of “inherent” flaws and hence induce micro-crack initiation.

a)
14
Young's modulus [GPa]

12

10

untreated epoxy silane mold release agent

Deformation and fracture behaviour 107

b) 2.0 c) 400

1.8
350
1.6
300
KΙ C [MPa m ]
0.5

1.4

GΙ c [J/m ]
250

2
1.2

1.0 200

0.8
150
0.6
100
0.4
50
0.2

0.0 0
untreated epoxy silane mold release agent untreated epoxy silane mold release agent

Figure 79: (a) Tensile Young’s modulus (E), (b) critical stress intensity factor (KIc) and (c)
critical energy release rate (GIc) as function of different silica filler treatment. All
measurements were performed at room temperature with constant silica filler content of
60% w/w.

3.5.5 Hyperbranched polymer toughening

Hyperbranched polymers (HBPs) are characterized by dendritically branched structures

and in contrast to traditional rubber modifiers (e.g. CTBN) or thermoplastic modifier (e.g.
PES, PEI) by a large number of functional end groups. The high density of the terminal
groups on the HBP-modifier further offers the ability for tailoring their compatibility either
through conversion of the accessible end groups to chemically suitable moieties or through
in-situ reaction to form covalently bound networks. Apart from the higher functionality,
HBPs differ generally from common rubber and thermoplastic-modifiers in their rheological
property profile. The densely branched polymer backbone of the HBP contributes with less
entanglement and a more globular structure, and hence to an essentially lower viscosity
and Newtonian behaviour when comparing HBPs of a given molecular weight with
conventionally branched or linear polymers of the same molecular weight. This specific
property profile demonstrates the unique potential of HBPs as effective low viscosity
tougheners for densely crosslinked thermosets [99-101].
In this section the dependency of the final morphology and the related thermo-mechanical
properties of blends between the MY721/MA/HHPA-system and different functionalized
HBPs were examined. The aim of this investigation is to characterize and understand the
physical properties of these epoxy-HBP blends. In particular the miscibility, stiffness, glass
transition temperature and fracture toughness properties of cured epoxy-HBP blends have
been studied more in detail.

3.5.5.1 Effects of the hyperbranched polymer end-groups

The effects of the shell chemistry of three commercially available HBP grades (referred
collectively as BoltornH-type) on the solubility and morphology of blends with the
108 Chapter 3

MY721/MA/HHPA-system were studied. The used H-types are all hydroxyl functionalized
but differ in molecular weight and number of hydroxyl end groups, see Table 8.
The compatibility of the 10% w/w HBP modified epoxy systems was studied by gradually
heating the mixtures (ϕ=1°C/min) from room temperature to determine the solution
temperature (Ts=cloud point temperature) by using optical microscopy. It was observed
that the solubility of the modifier in the MY721/MA/HHPA-matrix decreased with increasing
number of terminal hydroxyl groups of the HBP, and hence the measured critical solution
temperature (Ts) showed an analogous behaviour as illustrated in Figure 80.

200
Critical solution temperature [°C]

180

160

140

120

100

80

60
0 10 20 30 40 50

Theoretical no. of hydroxyl groups per molecule

Figure 80: Critical solution temperature (Ts) as a function of terminal hydroxyl groups for a
10% w/w modified MY721/MA/HHP-system. Ts-data were experimentally determined by
optical microscopy with a heating rate of 1°C/min.

It is assumed that the polar OH-end groups form strong hydrogen bonds that tend to
contribute to the relatively high solution temperatures (T>150°C) of the modified epoxy-
systems. However, pre-curing of the utilized epoxy system cannot be performed at
temperatures above 85°C, due to the required control of the exothermal energy release in
order to avoid a thermal runaway during cure. Therefore, it was not possible to start with
an initial homogeneous blend mixture prior to cure. Consequently, a polar solvent
(acetone) was used to pre-dissolve the OH-terminated HBPs before adding the modifier to
the epoxy/anhydride-system. After pre-dissolving the HBP-modifier and mixing thoroughly
all components, the solvent was removed again in vacuum (2 hours at 45°C), followed by
curing for 4 hours at 65°C and post-curing for 2 hours at 160°C. The obtained morphology
of a 10% w/w BoltornH-40 modified MY721/MA/HHPA-system is shown in the
corresponding SEM-micrographs in Figure 81. It can be seen that the spherical H-40 rich
domains are randomly dispersed within the cured epoxy matrix. However, a significant
coarsening of the primary HBP-particles occurred during the curing reaction at elevated
temperature. Consequently, a precise control of the finally obtained morphology when
Deformation and fracture behaviour 109

applying a solvent methodology is rather limited. For this reason and the fact of very
inhomogeneous obtained samples further mechanical analyses were abandoned.

Figure 81: SEM-micrographs of a cryo-fractured plane of MY721/MA/HHA systems

modified with 10% w/w Boltron®H-40 pre-dissolved in acetone. Acetone was removed in
vacuum for 2 hrs at 45°C. The sample was cured for 3 hours at 65°C and post-cured for 3
hours at 160°C.

The solubility of the HBPs within the epoxy matrix, however, can be greatly improved by a
chemical derivatisation of a partial fraction of the polar OH end-groups. Thus not only the
number of end groups but also the chemical nature of these end groups is expected to
play an important role in the solubility of the HBP modifier. Decreasing the epoxy
equivalent weight (EEW), that is increasing the number of epoxy groups on the HBP shell
at a fixed total molecular weight, was anticipated to increase the initial miscibility of the
HBP in the epoxy resin.
For instance the practical end group capping of an average 30 of the theoretical
32-hydroxyl groups to less polar epoxy groups (e.g. HBP-E2; EEW 402 g/eq.) resulted in
an improved miscibility between the resin and the hyperbranched polymer modifier. This
modification is seen as an obvious reduction of the related critical solution temperature (Ts)
from about 160°C to a temperature below 25°C. Thus finally processing and curing of
blends consisting of the MY721/MA/HHPA matrix and end-group tailored HBPs is feasible
without any additional solvent at cure temperatures below 85°C.

3.5.5.2 Hyperbranched polymer modifier with a high epoxy functionality

In the previous section it was shown that partial end capping of the high amount of
hydroxyl end-groups of the HBP led to a significant reduction of the critical solution
temperature in the studied epoxy system. This finally enables the processing of the end-
group tailored HBP-modifiers without the additional use of a solvent. In this section the
effects of such a tailored HBP (HBP-E2) composition on the morphology of the
MY721MA/HHPA-system and the related thermo-mechanical and fracture properties were
studied in more detail. Therefore, the hyperbranched molecule HBP-E2 (3rd generation, cf.
Table 8), where theoretically thirty primary epoxy groups are grafted onto the shell, has
110 Chapter 3

been blended with the epoxy-system in weight ratios between 0 to 100%. The HBP-E2,
with the relatively low epoxy equivalent weight (EEW) of 402 g/eq. induces total miscibility
of blends up to 80% w/w modifier fraction at room temperature. Furthermore, the samples
remain absolutely miscible over the entire curing and post-curing process and no micro
phase separation was detected by SEM examinations.
The curing kinetics of the reactive HBP-E2/MA/HHPA system, as determined from
isothermal DSC measurements, was observed to be very similar to that of the neat
MY721/MA/HHPA-system. Consequently, it is proposed that phase separation process is
effectively suppressed by that fact that during the initial state of cure the HBP-molecule is
already linked to the network.
The effect of the HBP-E2 concentration on the thermo-mechanical properties as
determined by DMA is shown in Figure 82(a) and (b). The normalized DMA-thermograms
clearly indicate that the reactive HBP-E2 modifier acts as a chain extender as manifested
by the continuous drop of the Young’s modulus (E) with increasing modifier fraction (see
Figure 82(d)). This behaviour is in accordance to the continuous decrease of the glass
transition temperature with increasing HBP-E2 concentration as determined from
tanδ-peak values of the same DMA-scans as shown in Figure 82(c). However, the
experimentally determined glass transition temperatures decreased only moderately from
about 220°C for the neat system to 203°C for a 20% w/w HBP-E2 modified system. The
Fox-equation [154], which is widely used to predict the ultimate glass transition
temperature (Tg∞) of statistical co-polymers, estimates the Tg∞ of the HBP-E2/epoxy
compositions only in an approximate way. All experimentally measured Tg-data showed
significantly higher glass transition temperature than the predicted values. Thus, it is
proposed that the HBP-E2 might be partially separated in sub-micron domains that do not
necessarily lead to a pronounced decrease of the thermo-mechanical properties, e.g. the
glass transition temperature. Moreover, strong interactions between the individual
components, such as hydrogen bonding interactions [155, 156] may additionally contribute
to a reduced influence on the thermo-mechanical behaviour. Nonetheless, a confirmation
of this hypothesis by means of microscopic investigations was not possible, because the
phase contrast between potential sub-micron domains and the epoxy matrix was not
sufficient to be resolved by scanning and transmission electron microscopy. In addition,
OsO4 and RuO4-staining of fractured surfaces in order to enhance the phase contrast were
not successful.
Since no direct assessment of the stiffness of the continuous and dispersed phase is
possible in this case, an approximation using an upper and lower boundary condition was
applied. The lower boundary was estimated by considering the tensile Young’s modulus
(E) to be a linear function of composition. For the upper boundary it was assumed that the
Deformation and fracture behaviour 111

blend formed a continuous phase where the stiffness is equal to that of the ultimately
crosslinked MY721/MA/HHPA-matrix. However, experimental studies revealed, that for the
fully cured HBP-E2 modifier with a Tg below room temperature, a more complex behaviour
was obtained (Figure 82(d)). Herein a slight but almost linear decrease of the E-value was
obtained until the Tg of the blend composition approached room temperature. Moreover, it
can be see that up to 80% w/w HBP-E2 fraction the epoxy matrix preferably dominates the
elastic tensile properties of the blends. Decreasing further the Tg (Tg≤RT), the Young’s
modulus rapidly decreased to the value of the crosslinked HBP-E2/MA/HHPA value of
570 MPa. Hence the prediction of E at temperatures below room temperature leads to a
linear extrapolation towards the lower matrix boundary.

a) 0.25
b)
10%
Normalized storage modulus [-]

0%
1 80%
20% 0.20
40%

80% 20%
0.15
tan δ [−]

60%

0.10
0.1

40%
0.05
0%
50 75 100 125 150 175 200 225 250 275 50 75 100 125 150 175 200 225 250
Temperature [°C] Temperature [°C]

c) 550 d)
upper matrix boundary Composition Tg = 25°C
500
Glass transition temperature [K]

Young's modulus [GPa]

450

400

1E9
350
Fox equation
DMA 1.run
300
DMA 2.run
lower matrix boundary
DSC 5°C/min
250
0.0 0.2 0.4 0.6 0.8 1.0
0.0 0.2 0.4 0.6 0.8 1.0
HBP-E2 weight fraction [-]
HBP-E2 weight fraction [-]

Figure 82: (a) Normalized storage modulus (G’/G’RT); (b) mechanical loss factor (tan δ); (c)
glass transition temperature (Tg) as obtained from DMA measurements; (d) tensile
Young’s modulus as a function of the HBP-E2 weight fraction.

The toughening effects between the “fully” miscible HBP-E2 and MY721/MA/HHPA resin
system were determined by DT-fracture tests at room temperature. It can be seen from the
corresponding DT-load/displacement curves in Figure 83 that the crack propagation mode
changed from initially stable towards unstable behaviour with increasing fraction of the
HBP-E2 modifier.
 112 Chapter 3

80

70 a) b) c)
Load [N] 60

50

40

30

20

10

0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5 0.6

Displacement [mm]

Figure 83: DT-load/displacement curves of MY721/MA/HHPA-system as a function of the

HBP-E2 fraction (a) 0% w/w, (b) 5% w/w, (c) 10% w/w. All DT-fracture tests were
performed with a constant strain rate of 0.1 mm/min at room temperature.

The fracture toughness corresponding to the critical stress intensity factor (KIc) and the
critical energy release rate (GIc) increased almost linearly with the HBP-E2 modifier
fraction, as shown in Figure 84. For a blend with a 20% w/w fraction of HBP-E2 the
averaged critical stress intensity factor increased from initial 0.58 MPam0.5 for the neat
resin system to 0.88 MPam0.5 for the modified resin blend. Considering further the only
moderate decrease in the Young’s modulus in this composition area (cf. Figure 82(d)), the
critical strain energy release rate (GIc) yields to a ∼3.5-fold increase from initial 82 J/m2 for
the neat resin to 290 J/m2 for the 20% w/w HBP-E2 modified system.

a) 1.0 b) 350

0.9 300

250
0.8
KIc [MPam ]
0.5

200
GIc [J/m ]
2

0.7
150
0.6
100

0.5
50

0.00 0.05 0.10 0.15 0.20

0.00 0.05 0.10 0.15 0.20
HBP-E2 weight fraction [-]
HBP-E2 weight fraction [-]

Figure 84: (a) Critical stress intensity factor (KIc) and (b) critical strain energy release rate
(GIc) as a function of the HBP-E2 weight fraction on the MY721/MA/HHPA-system. All DT-
fracture tests were performed with a constant strain rate of 0.1 mm/min at room
temperature.

The obtained fracture toughness results clearly demonstrate that the use of a tailored
HBP-modifier (KIC∼0.9 MPam0.5) represents a good alternative to traditional rubber
modification with CTBN (KIC∼0.7 MPam0.5, cf. Figure 48). However, compared to
particulate filled systems (KIC∼1.6 MPam0.5, cf. Figure 74) as investigated in
Deformation and fracture behaviour 113

section 3.5.4.3 the toughness improvement remains low. Therefore, a combination of the
two different toughening modifiers (HBP and silica filler) may result in synergistic
toughening effects. This so-called hybrid-concept is known from previous studies on hard
particle/rubber-modified epoxies [7, 46, 157-159]. Moreover, the hybrid approach balances
the intrinsic drawbacks of rubber- or thermoplastic-toughening arising from pronounced
variations of the thermo-mechanical property profile.
Experimental fracture toughness data of the hybrid-composites using a fixed silica fraction
(60% w/w) and a varying HBP-E2 content are given in Figure 85. It can be seen that for
the investigated system no synergistic toughening effects between the silica-filler and the
HBP-E2 modifier were observed. All obtained KIc-values of the hybrid-composites were on
a stabilized level at ∼1.6 MPam0.5 and independent of the HBP-E2 fraction between 0 to
10% w/w of the epoxy mixture (resin and hardener). Consequently, the final toughness
properties are predominately governed by the inorganic filler fraction and the implemented
crack pinning mechanism. The small increase in the critical strain energy release rate (GIc)
is likely caused by the moderate decrease of the corresponding elastic modulus (E).
Consequently, the obtained changes cannot be really considered as an enhancement of
the fracture toughness properties due to synergistic HBP and silica toughening effects.
It is proposed that most probably the added HBP modifier interfere with the polar silica
filler surface. Thus, a diffusion of the HBP to the filler surface will additionally explain the
only marginal influence on the fracture toughness properties, because as discussed
already in section 3.5.4.5, modifications of the filler/matrix adhesion have no significant
effect on the final fracture toughness.

a) 2.0 b)
260
1.9
240
1.8
1.7 220
GIc [J/m ]
2

1.6 200
KIc [MPam ]
0.5

1.5 180
1.4
160
1.3
140
1.2
120
1.1
1.0 100
0.00 0.03 0.05 0.08 0.10 0.00 0.02 0.04 0.06 0.08 0.10
HBP-E2 weight fraction [-] HBP-E2 weight fraction [-]

Figure 85: (a) Critical stress intensity factor (KIc), and (b) critical strain energy release rate
(GIc) of a hybrid-composite (MY721/MA/HHPA+silica+HBP-E2) as a function of HBP-E2
weight fraction. The silica filler content was fixed to 60% w/w. All DT-fracture tests were
performed with a constant strain rate of 0.1 mm/min at room temperature.
114 Chapter 3

3.5.5.3 Hyperbranched polymer modifier with a low epoxy functionality

In section 3.5.5.1 it was shown that blends of epoxy and HBPs with a high number of
hydroxyl end groups resulted in compositions with a critical solution temperature above
150°C. On the other hand in the previous section it was verified that a conversion of the
terminal hydroxyl groups into more non-polar epoxy groups lead to fully homogeneous
epoxy-HBP blends without macroscopic phase separation (particles >0.5 µm) over the
entire cure and post-cure process. Therefore, the procedure of modifying the final
morphology by tailoring the modifier shell nature was shown to be sensibly larger in the
case of HBPs than with other traditional toughening modifiers.
To promote a more favourable two-phase morphology by means of a thermodynamically
controlled phase separation the use of a HBP modifier with only partially end capped
hydroxyl groups has been studied more in detail. Therefore, a third-generation epoxidized
HBP-E1 (BoltornE1) has been blended with the MY721/MA/HHPA-system in different
compositions. The shell of the HBP-E1 contains in average 21 hydroxyl groups and 11
secondary aliphatic epoxy groups yielding to an epoxy equivalent weight of 402 g/eq. (see
Table 8). The concentration of the added HBP-E1 has been varied between 0 to 95% w/w
fractions.
After mixing all investigated systems the visual mixture appearance changed
spontaneously from transparent to turbid. This observation provides evidence that there is
presumably no marked nucleation barrier for phase separation. The obtained
morphologies depend on the initial HBP-concentration. Some corresponding optical
micrographs of the blends recorded directly after mixing at room temperature are given in
Figure 86. It can be seen that up to a HBP-E1 weight fraction of 40% w/w the epoxy matrix
builds the continuous phase and the HBP-E1 phase separates as spherical particles.
Increasing further the HBP-E1 fraction above ∼40% w/w the finally obtained morphology
gradually shifted towards co-continuous morphology consisting of the epoxy and the HBP-
E1 phases. At HBP-E1 weight fractions higher than 80% w/w a phase-inverted morphology
is obtained wherein the HBP is the continuous phase and the epoxy is the dispersed
phase.
Deformation and fracture behaviour 115

20µm
a) b)

c) d)

e) f)
Figure 86: Optical micrographs of the obtained thin films morphology for HBP-E1-
MY721/MA/HHPA-blends as a function of HBP-E1 weight fraction for (a) 0%, (b) 10%,
(c) 20%, (d) 40%, (e) 80% and (f) 95%. All examinations were performed directly after
mixing at room temperature.

A simplified temperature-composition phase diagram of the epoxy/HBP-E1 system was

constructed from the related cloud point temperatures using a hot-stage in combination
with optical microscopy. As shown in Figure 87 the obtained phase diagram for the
blended systems exhibited an enhanced miscibility with increasing temperature. The
experimentally obtained temperature maximum of liquid–liquid (L-L) phase-separation line
between the homogeneous and two-phase appearance was determined at 82°C for a
5% w/w modifier fraction. Increasing further the HBP-E1 modifier content line resulted in a
gradually shifted of the L-L transition towards lower temperatures again. For a 30% w/w
HBP-E1 modifier fraction the L-L transition temperature was observed at 48°C. This
obtained reversible phase transition behaviour is corresponding to an upper critical
solution temperature behaviour (UCST) [160].
116 Chapter 3

85
80
1-phase
75
70
65
Temperature [°C]

60
55
50
2-phases a)
45
40
35
30
25
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
HBP-E1 weight fraction [-]

b)
Figure 87: Left: Experimental phase diagram of the MY721/MA/HHPA-system (r=0.8) as a
function of HBP-E1-modifier weight fraction. The (V) symbols denote to experimental
cloud point temperatures determined by optical microscopy with a heating rate of 1°C/min.
Right: Optical transmission micrograph (magnification 500x) of the uncured
MY721/MA/HHPA-system (r=0.8) modified with 10% w/w HBP-E1 modifier (a) at 85°C,
and (b) at 65°C

The general application of the Flory-Huggins lattice theory allows reconstructing phase
diagrams to express the L-L phase separation behaviour. Assuming the validity of the
ordinary Flory-Huggins equation [161] and dividing this mathematical relation by the
overall volume, the free energy of mixing per unit volume, ∆Gv can be expressed as:
Φ Φ  (3.55)
∆G v = RT  1 ln Φ1 + 2 ln Φ 2 + Φ1 Φ 2 χ12 
 1
V V 2 
where T is the absolute temperature, R is the ideal gas constant, Φ1 and Φ2 are the volume
fractions of the components 1 and 2 (resin and modifier), V1 and V2 is the molar volume of
the two components, and χ12 is the interaction parameter expressing the interactions
between the two species present. According to eqn (3.55) the free energy of mixing is a
function of temperature, the interaction parameter, and the volume fraction. The
occurrence of phase separation and the related appearance of the phase diagram depend
strongly on the interaction parameter (χ12), which is a material specific parameter.
Moreover, the interaction parameter is not a constant value, but depends on temperature,
the composition, and even the pressure. Assuming the simplification that the interaction
parameter is pre-dominantly affected by temperature, the temperature dependence can be
described, as originally proposed by Huggins [162], by:
b
χ12 = a + (3.56)
T
From eqn (3.56) two types of phase diagrams can be derived. If the value of χ12 is positive
and the increases linearly with 1/T, the resulting phase diagram will exhibits an upper
Deformation and fracture behaviour 117

critical solution temperature (UCST) behaviour. If χ12 is negative and the decreases linearly
with 1/T the system displays a lower critical solution temperature (LCST) behaviour [163].
Besides, if curing of reactive compounds is carried out, the molecular weight gradually
increases from the initial monomers towards an infinite molecular weight. Though this is a
simplified picture of the more complex reaction scheme, it can be used for a qualitative
description of the reaction induced phase separation process that is alternatively known as
chemically induced phase separation (CIPS). In terms of the temperature-composition
phase diagram, the increase of the molecular weight can be pictured as a shift of the L-L
phase transition line to higher temperatures, assuming that the system generally exhibits a
UCST behaviour. This general behaviour is schematically shown in Figure 88. At the start
of polymerization at temperature, T, the system with composition, C, is initially
homogeneous. As the polymerization proceeds, the modifier becomes less miscible and
the system is thrust into the two-phase region, eventually leading to phase separation into
a modifier-rich and modifier lean phase. Depending on the initial modifier concentration,
the modifier rich phase will become the dispersed or the continuous phase (phase
inverted) [50, 164].

Figure 88: A schematic representation of the evolution of the liquid-liquid (L-L) phase
transition line for upper critical solution temperature (UCST) phase behaviour upon
polymerization (t1<t2<t3). The point C represents the starting temperature and
concentration.

Consequently, the CIPS during cure of the epoxy resin enables a more precise control of
the phase separation mechanism and hence the resulting morphology and related
structure-property relationship. Some examples of reactive systems that exhibit an UCST
phase behaviour are found for most rubber–modified epoxy systems [17, 18, 21, 23, 30] as
118 Chapter 3

well as in PPE/epoxy-systems [165, 166]. Herein the CIPS-methodology has been widely
used for blending numerous thermosets with a number of high molecular compounds in
the past [50, 52, 167-169], because the obtained two-phase nature of the material greatly
improves crack and impact resistance of the inherently brittle epoxy matrix. An example of
the LCST behaviour for reactive systems is the polyethersulphone/epoxy system [170-
172]. Since the given examples all involve crosslinking reactions, the evolution of the
phase behaviour as depicted in Figure 88, only meaningfully describes the situation in the
pre-gel state.
Curing experiments that have been performed slightly above the liquid-liquid phase
transition line at 85°C (cf. Figure 87) started with an initially one-phase mixture. However,
none of the examined samples showed a micro-phase separation and remained totally
homogeneous upon completion of the curing reaction as investigated by scanning electron
microscopy.
The absence of the chemically induced phase separation (CIPS) can be explained in two
different ways. Firstly, the cloud point conversion (αcp) of the system is higher then the
conversion at the point of gelation (αcp>αgel). Thus, the fast network formation of the matrix
links the reactive hyperbranched molecules already in the initial stages of cure, and hence
phase separation is limited by the diffusional restrictions which take place upon gelation of
the epoxy matrix.
The second explanation can be given by thermodynamic consideration of the blend-
composition. Separating eqn (3.55) into the two specific contributions of the interaction
enthalpy of mixing (∆Hv), and the entropy of mixing (∆Sv), yields:
∆Hv = RTΦ1 Φ 2 χ12 (3.57)
Φ Φ 
∆S v = − RT  1 ln Φ1 + 2 ln Φ 2  (3.58)
 V1 V2 
So far a chemically induced phase separation of HBPs have only been successfully
demonstrated for amine-cured epoxy compositions that exhibit a distinct polarity change
upon cure [99, 173-176]. Due to significant changes of the molecular structure of the
reactive components, caused by the ring opening reaction of the epoxy group and the
subsequent formation of new hydroxyl groups, a 3.9 fold increase in polarity of an amine-
epoxy system has been reported by Boogh et al. [99]. As a result, the related enthalpy
change (∆Hv) can be considered as the major contribution for changes of the Gibb’s free
energy of mixing.
The entropic contribution (∆Sv), which is generally related to the configurational orders of
the blend, generally increases with size of the molecules. However, for polymer-polymer
systems the entropic contribution (cf. eqn (3.58)) is reported to be relatively small [177].
Deformation and fracture behaviour 119

Consequently, the major thermodynamic driving potential for a chemically induced phase
separation processes in epoxy resins can be related to the pronounced enthalpy changes
and their significant influence on the Gibb’s free energy of mixing. On the other hand the
esterification reaction during the epoxy/anhydride polymerization does not generally result
in a pronounced polarity change of the systems upon cure. Hence, the related energetic
enthalpy changes can only be stated as secondary effects. The simultaneously occurring
configurationally entropic changes (∆Sv) are not sufficient to enter the thermodynamically
meta-stable region in order to precipitate the hyperbranched molecules from the epoxy
matrix. As a consequence, if a change in polarity is the primary trigger for a chemically
induced phase separation process, this phenomenon may not apply to the blends studied
here.
Samples of a 10% w/w HBP-E1 modified epoxy were isothermally pre-cured below the L-L
phase separation line at 65°C and afterwards post-cured at 160°C. The initial two-phase
structure retained over the entire cure process yielding a continuous epoxy matrix and
spherically dispersed HBP-inclusions as co-continuous phase. Nevertheless, cryogenic-
fractured samples showed a distinct distribution gradient of the randomly dispersed HBP-
inclusions over the entire sample height, as shown in Figure 89. It can be seen that the
dispersed particle size increases with sample height due to substantial particle
coalescence of the primary HBP-E1 domains. Consequently, a tacky surface layer
consisting of merely crosslinked HBPs was detectable on all blended samples. Moreover,
the thickness of the surface layer steadily increased with the HBP-E1 concentration on the
studied systems.
The observed structural coarsening mechanism is probably favored by phase instabilities
that can be attributed to discrepancies between the specific densities of the cured HBP-E1
(ρ=1.08 g/cm3) and the crosslinked epoxy resin (ρ=1.27 g/cm3). Additionally, the long time
to gelation (tgel) and the relatively high values of the diffusion coefficient (low system
viscosities) may contribute further to the particle coalescence and thus to the
macroscopically inhomogeneous morphology. Besides, it is also possible that a critical
interfacial energy balance between the modifier and the epoxy mixture further supports the
observed segregation effects as shown in Figure 89.
120 Chapter 3

Separation line

Surface Layer

Height
Bottom
Mould

Figure 89: SEM-micrographs of different cryogenic-fractured planes of the

MY721/MA/HHA-system blended with 10% w/w HBP-E1. The sample was cured for 3
hours at 65°C and post-cured for 3 hours at 160°C.

3.6 Concluding remarks

In this chapter the fundamental understanding of the deformation and fracture behaviour of
highly crosslinked epoxy-systems has been reviewed and experimentally studied. It has
been shown that the investigated neat and silica-filled MY721/MA/HHPA-system deforms
predominantly by shear yielding processes that are sensitive to the presence of
inhomogeneities. Moreover, specific test conditions such as compressive load, high
temperature, and low strain rate additionally amplify the initiation of shear banding
processes. Under compressive load the particulate filled epoxy system exhibited extensive
stress-whitening damage that generally starts above the yield stress. Particle-matrix
debonding has been identified as a primary mechanism for this irreversible deformation
phenomenon.
The crack propagation and fracture of the neat and silica filled MY721/MA/HHPA-system
have been examined using the Double-Torsion (DT) test method. In particular the effects
of the strain rate, temperature, volume fraction, particle size and filler surface-treatment
were assessed and discussed. It has been shown that with increased strain rates, the
studied systems failed preferentially in an unstable crack propagation manner. At
Deformation and fracture behaviour 121

increased test temperatures the critical stress intensity factor (KIc) and the critical strain
energy release rate (GIc) significantly decreased. It is proposed that the observed
behaviour provides evidence of an existing sub-network structure. The two independent
MY721/MA and MY721/HHPA networks are strongly related to the co-anhydride mixture
used, and the different curing rates of the individual anhydrides.
For the particulate filled system it was observed that an increase of the silica filler fraction
leads to a continuous increase of the critical stress intensity factor (KIc) and the
corresponding critical energy release rate (GIc). Micro-structural investigation and
theoretical models have provided evidence that the toughening behaviour can be
predominately related to the crack pinning mechanism. A fairly good agreement between
the experimental data and a theoretical prediction, as previously proposed by Evans et al.,
has been obtained. It was demonstrated that different pre-treatment of the filler surface
and thus the filler-matrix adhesion has only a minor influence on the final fracture
toughness. Although there might be some exceptions, some general trends describing the
influence of filler particles on the short-term deformation and fracture behaviour of epoxy
thermosets are summarized in Table 10.

Table 10: Parameters determining the modulus, fracture toughness and mechanical
strength of particulate filled thermosets
Property Effects on
Modulus Fracture toughness Mechanical
strength
Increased filler fraction + + 0
1) 1)
Increased particle size 0, - 0, - 0, -1)
Improved filler/matrix adhesion 0 0 +
Increased strain rate + + +
+: increase; 0: constant; -: decrease; 1) above a particle size threshold level of 2 to10 µm

The effects of modifying the epoxy matrix with different hyperbranched polymer (HBP)
additives have been studied in order to assess and demonstrate their toughening
efficiency as potential low viscosity modifiers. It was confirmed that the final blend
morphology and the corresponding structure-properties strongly depend on the HBP shell
chemistry. An increased number of polar hydroxyl end groups leads to a decreased
compatibility between the modifier and the epoxy matrix. Moreover, tailoring the end-
groups functionality by conversion of the accessible hydroxyl-groups to more non-polar
epoxy-groups, e.g. HBP-E2 (EEW 402 g/eq.), leads to an enhanced solubility of the HBP
and hence provides totally miscible compounds at every concentration. Indication that the
HBP-E2 acts as internal chain extender was obtained from thermo-mechanical analysis.
However, the related ultimate glass transition temperature of the blends is not strongly
122 Chapter 3

affected. Therefore, it is proposed that the HBP-E2 possibly arrange in sub-domains that
consequently induce only minor changes in the related network structure and the
corresponding property profile. A homogeneous blend with 20% w/w HBP-E2 modifier
leads to about 50% increase in the KIc-values. Consequently, specifically tailored HBPs
can be considered as good alternatives to traditionally used rubber-modifiers, e.g.
carboxyl-terminated butadiene-acrylonitrile rubber (CTBN). However, synergistic effects
between the silica filler and HBP-modifier of envisaged hybrid formulations have not been
observed.
Epoxy/HBP blends with only partially epoxidized HBP-modifier (HBP-E1; EEW 902 g/eq.)
revealed an upper critical solution temperature (UCST) phase behaviour. However,
samples that have been cured slightly above the determined liquid-liquid phase separation
temperature (T=85°C) line remain homogeneous upon the entire cure and no chemically
induced phase separation (CIPS) was observed. It is proposed that the absence of a CIPS
could be mainly attributed to only minor polarity changes of anhydride cured epoxy
formulations upon cure. Consequently, the resulting enthalpy changes in the investigated
system, and the direct influence on the Gibb’s free energy of mixing limits the
thermodynamically induced phase separation process.

3.7 References

[1] Williams J.G., Stress Analysis of Polymers, Longman, London, (1973).

[2] Andrews E.H., Fracture of Polymers, Oliver & Boyd, Edingburgh, (1968).
[3] Brostow W., Corneliussen R.D., Failure of Plastics, Carl Hanser Publishers,
Munich, (1986).
[4] Williams J.G., Fracture Mechanics of Polymers, Ellis Horwood Ltd., Chichester,
(1984).
[5] Griffith A.A., Phil.Trans. Roy. Soc., A221, 163 (1920).
[6] Bucknall C.B., Toughened Plastics, Applied Science Publishers, London, (1977).
[7] Moloney A.C., Kausch H.H., Kaiser T., Beer H.R., J. Mater. Sci., 22, 381 (1987).
[8] Inglis G.E., Proc. Int. Naval Architect., 55, 219 (1913).
[9] Young R.J., Introduction to Polymers, Chapman & Hall, New York, (1983).
[10] Irwin G.R., American Society of Metals, Cleveland (OH), pp.147, (1949).
[11] Irwin G.R., J. Appl. Mech., 24, 361 (1957).
[12] Westergaard H.M., J. Appl. Mech., 61, A49 (1945).
[13] Hertzberg R.W., Deformation and Fracture Mechanics of Engineering Materials,
John Wiley Publishers, New York, (1976).
[14] ASTM, Fracture Toughness and its Applications, ASTM STP 381, Philadelphia,
1965
[15] BS, Methods of Test for Plane Strain Fracture Toughness of Metallic Materials, BS
5447, London, 1977
[16] ASTM, Standard Method of Test for Plane Strain Fracture Toughness of Metallic
Materials, ASTM E 399-81, Part 1, Philadelphia, 1982
[17] Sultan J.N., Laible, R.C., McGarry, F.J., Appl. Polym. Symp., 16, 127 (1971).
[18] Sultan J.N., McGarry, F.J., Polym. Eng. Sci., 13, 29 (1973).
Deformation and fracture behaviour 123

[19] McGarry F.J., Proc. Roy. Soc., London, A319, 59, (1970).
[20] Huang Y., Kinloch A.J., J. Mater. Sci. Lett., 11, 484 (1992).
[21] Huang Y., Kinloch A.J., J. Mater. Sci., 27, 2753 (1992).
[22] Huang Y., Kinloch A.J., Polymer, 33, 5338 (1992).
[23] Huang Y., Kinloch A.J., J. Mater. Sci., 27, 2763 (1992).
[24] Kinloch A.J., Shaw S.J., Tod D.A., Hunston D.L., Polymer, 24, 1341 (1983).
[25] Kinloch A.J., Shaw S.J., Hunston D.L., Polymer, 24, 1355 (1983).
[26] Kinloch A.J., Hunston D.L., J. Mater. Sci. Lett., 5, 1207 (1986).
[27] Bagheri R., Pearson, R.A., Polymer, 37, 4529 (1996).
[28] Bagheri R., Pearson, R.A., Polymer, 41, 269 (2000).
[29] Banthia A.K., Chaturvedi, P.N, Jha, V., Pendyala, V.N.S. Rubber Toughened
Plastics, In: Rubber Toughened Plastics (ed. Riew C.K.), American Chem. Soc.,
Washington, D.C., pp. 343, (1989).
[30] Chen K.C., Huang J.Y., J. Mater. Sci., 26, 5848 (1991).
[31] Chen T.K., Jan Y.H., Polym. Eng. Sci., 35, 778 (1995).
[32] Edwards D.C., J. Mater. Sci., 25, 4175 (1990).
[33] Frounchi M., Mehrabzadeh M., Parvary M., Polym. Int., 49, 163 (2000).
[34] Guild F.J., Young R.J., J. Mater. Sci., 24, 2454 (1989).
[35] Kim D.S., Kim, S.C., Polym. Eng. Sci., 34, 1598 (1994).
[36] Kang B.U., Jho J.Y., Kim J., Lee S.S., Park M., Lim S., J. Appl. Polym. Sci, 79, 38
(2000).
[37] Kim D.S., Kim, S.C., Park, C.E., Polym. Eng. Sci., 36, 755 (1996).
[38] Kunz-Douglass S., Beaumont P.W.R., Ashby M.F., J. Mater. Sci., 15, 1109 (1980).
[39] Kunz S.C., Beaumont P.W.R., J. Mater. Sci., 16, 3141 (1981).
[40] Lee B.L., Lizak C.M., Riew C.K., Moulton R.J., 12th Nat. SAMPE Tech. Conf.,
Seattle (WA), pp.1116, (1980).
[41] Pearce P.J., Morris C.E.M., Ennis B.C., J. Therm. Anal., 37, 1137 (1996).
[42] Pearson R.A., Yee A.F., J. Mater. Sci., 21, 2475 (1986).
[43] Pearson R.A., Yee A.F., J. Mater. Sci., 21, 2462 (1986).
[44] Pearson R.A., Yee A.F., J. Mater. Sci., 24, 2571 (1989).
[45] Pearson R.A., Yee A.F., J. Mater. Sci., 26, 3828 (1991).
[46] Zhang H.X., Berglund L.A., Polym. Eng. Sci., 33, 100 (1993).
[47] Yamanaka K., Inoue T., J. Mater. Sci., 25, 241 (1990).
[48] Kalfoglou N.K., Williams H.L., J. Appl. Polym. Sci, 17, 1377 (1973).
[49] Bartlet P., Pasault J.P., Sautereau H., J. Appl. Polym. Sci, 30, 2955 (1985).
[50] Hwang J.F., Manson J.A., Hertzberg R.W., Miller G.A., Sperling L.H., Polym. Eng.
Sci., 29, 1477 (1989).
[51] Kim D.S., Kim S.C., Polym. Eng. Sci., 34, 1598 (1994).
[52] Kunz S.C., Sayre J.A., Assink R.A., Polymer, 23, 1897 (1982).
[53] McGarry F.J., Rosner R.B., Adv. Chem. Ser., 233, 305 (1993).
[54] McGarry F.J., Subramaniam R., Adv. Chem. Ser., 252, 133 (1996).
[55] Montarnal S., Pascault J.P., Sautereau H., Adv. Chem. Ser., 222, 193 (1989).
[56] Vallo C.I., Hu L., Frontini P.M., Williams J.J., J. Mater. Sci., 29, 2481 (1994).
[57] Grillet A.C., Galy J., Pascault J.P., Polymer, 33, 34 (1992).
[58] Han X., Yun Z., Guo F., Chin. Chem. Lett., 2, 739 (1991).
[59] Han X., Li S., Zhang Q., Chin. J. Polym. Sci., 8, 335 (1990).
[60] Sankaran S., Chanda M., J. Appl. Polym. Sci., 39, 1459 (1990).
[61] Han X., Yun Z., Guo F., Mater. Res. Soc. Symp. Proc., 274, 11 (1992).
[62] Okamoto Y., Polym. Eng. Sci., 23, 222 (1983).
[63] Lee W.H. Toughened Polymers, In: Polymer Blends and Alloys (ed. Folkes M.J.,
Hope, P.S.), Chapman & Hall, London, pp. 163, (1993).
[64] Wang H., Li S., Ye J., J. Appl. Polym. Sci., 44, 789 (1992).
[65] Bergstrom D.F., Burns G.T., Decker G.T., Durall R.L., Fryrear D., Gornowicz G.A.,
Tokunoh M., Odagiri N., Proc. 37th Int. SAMPE Symp. Exhib., Anaheim, 37, 278,
(1992).
124 Chapter 3

[66] Decker G.T., Tobukuro K., Gornowicz G.A., EP475611; (1992).

[67] Martuscelli E., Musto P., Ragosta G., Riva F., Mascia L., J. Mater. Sci., 35, 3719
(2000).
[68] Ochi M., Katayama T., Mimura K., Yamana H., Kobunshi Ronbunshu, 50, 621
(1993).
[69] Pearson R.A., Adv. Chem. Ser., 233, 405 (1993).
[70] Diamant J., Moulton R.J., Proc. 28th Intl. SAMPE Conf., Washington, 16, 13,
(1984).
[71] Lee W.H., PhD-Thesis, Brunel University, Middlesex (UK), (1986)
[72] Levita G. In: Rubber-Toughened Plastics (ed. Riew C.K.), ACS, Washington (DC),
pp. 94, (1989).
[73] Van der Sanden M.C.M., Meijer H.E.H., Polymer, 34, 5063 (1993).
[74] Yee A.F., Pearson R.A., J. Mater. Sci., 21, 2462 (1986).
[75] Lange F.F., Radford K.C., J. Mater. Sci., 6, 129 (1968).
[76] Lange F.F. Fracture of Brittle Matrix, Particulate Composites, Composite Materials,
In: Fracture and Fatigue (ed. Broutman L.J.), vol. 5, Academic Press, New York,
pp. (1974).
[77] Smith J.W., PhD-Thesis, École Polytechnique Fédérale de Lausanne, Lausanne
(CH), (1989)
[78] Lange F.F., Radford K.C., J. Mater. Sci., 6, 1197 (1971).
[79] Lange F.F., Phil. Mag., 22, 983 (1971).
[80] Mallick P.K., Broutman L.J., Mater. Sci. Eng., 8, 98 (1975).
[81] Evans A.G., Phil. Mag., 26, 1327 (1972).
[82] Green D.J., P.S. N., Embury J.D., J. Mater. Sci., 14, 1413 (1979).
[83] Green D.J., Nicholson P.S., Embury J.D., J. Mater. Sci., 14, 1657 (1979).
[84] Moloney A.C., Kausch H.H., Stieger H.R., J. Mater. Sci., 1984, 1125 (1984).
[85] Moloney A.C., Kausch H.H., Stieger H.R., J. Mater. Sci., 18, 208 (1983).
[86] Hedrick J.C., Patel N.M., McGrath J.E., Adv. Chem. Ser., 233, 293 (1993).
[87] Hodgkin J.H., Simon G.P., Varley R.J., Polym. Adv. Technol., 9, 3 (1998).
[88] Muelhaupt R., Chimica, 44, 43 (1990).
[89] Raghava R.S., Proc. 28th Nat. SAMPE Symp., Corvino (CA), 367, (1983).
[90] Bucknall C.B., Partridge I.K., Polym. Eng. Sci., 26, 54 (1986).
[91] Raghava R.S., J. Polym. Sci. Part B, 25, 1017 (1987).
[92] Bucknall C.B., Gilbert A.H., Polymer, 24, 639 (1983).
[93] Raghava R.S., J. Polym. Sci. Part B, 26, 65 (1988).
[94] Stenzenberger H.D., Composite Structures, 24, 219 (1993).
[95] Bauer R.S., Stenzenberger H.D., Römer W., Proc. 35th Int. SAMPE Symp.,
Corvino (CA), 395, (1990).
[96] Hourston D.J., Lane J.M., MacBeath N.A., Polym. Int., 26, 17 (1991).
[97] Chen M.C., Hourston D.J., Schafer F.U., Huckerby T.N., Polymer, 36, 3287 (1995).
[98] Bucknall C.B., Gilbert A.H., Polymer, 30, 213 (1989).
[99] Boogh L., Pettersson, B., Manson, J.A., Polymer, 40, 2249 (1999).
[100] Wu H., Xu J., Liu Y., Heiden P., J. Appl. Polym. Sci., 72, 151 (1999).
[101] Ratna D., Simon G.P., Polymer, 42, 8833 (2001).
[102] Rose L.R.F., Mech. Mater., 8, 11 (1987).
[103] Johansson M., Malmström E., Hult A., J. Polym. Sci. Part A, 31, 619 (1993).
[104] Malmström E., PhD-Thesis, Royal Institute of Technology, Stockholm (SE), (1996)
[105] Malmström E., Trollsas M., Hawker C.J., Johansson M., Hult A., Polym. Mater.Sci.
Eng., 77, 151 (1997).
[106] Patent Application WO9900439; Annby U., Malmberg M., Pettersson B., Rehnberg
N., (1999).
[107] Patent Application WO96/12754; Pettersson B., Soerensen K., (1997).
[108] Newkome G.R., Moorefield C.N., Vögtle F., Dendritic Molecules, VCH Publishers
Inc., New York, (1996).
Deformation and fracture behaviour 125

[109] Boogh L., Jannerfeldt G., Pettersson B., Björnberg H., Manson J.A., Proc. 12th
ICCM Conf., Paris, France, (1999).
[110] Malmström E., Johansson M., Hult A., Macromol. Chem. Phys., 197, 3199 (1996).
[111] Patent Application WO97/23538; Pettersson B., (1997).
[112] Patent Application WO97/45474, Manson J.A., Boogh L., Pettersson B.,
Soerensen K., (1997).
[113] Product data sheet, Boltorn H20 - H40, Perstorp Specialty Chemicals AB, 2000
[114] Product data sheet, Boltorn E1 - E2, Perstorp Specialty Chemicals AB, 2000
[115] Product data sheet, Millsil W10, Quarzwerke Frechen, 2000
[116] Trachte K.L., DiBenedetto A.T., Int. J. Polym. Mater., 1, 75 (1971).
[117] Outwater J.O., Murphy M.C., Proc. 26th Society of Plastic Industry Conf., 10,
(1971).
[118] Kies J.A., Clark A.B., Proc. 2nd Int. Conf. on Fracture, Brighton, paper 42, (1969).
[119] Evans A.G., J. Mater. Sci., 7, 1134 (1972).
[120] Evans A.G., International Journal of Fracture, 9, 267 (1973).
[121] Williams D.P., Evans A.G., Journal of Testing and Evaluation, 1, 264 (1973).
[122] McKinney K.R., Smith H.L., Journal of American Ceramic Society, 56, 30 (1973).
[123] Weidmann G.W., Holloway D.G., Physics and Chemistry of Glasses, 15, 116
(1974).
[124] Irwin G.R., Handbuch der Physik, Springer Verlag, Berlin, (1958).
[125] Matonis V.A., Small N.C., Polym. Eng. Sci., 9, 90 (1969).
[126] DIN, Zugversuch an Kunststoffen, DIN EN ISO 727-1, Berlin, 1996
[127] DIN, Druckversuch an Kunststoffen, DIN EN ISO 604, Berlin, 1997
[128] Cantwell W.J., Smith J.W., Kausch H.H., Kaiser T., J. Mater. Sci., 25, 633 (1990).
[129] Lee J., PhD-Thesis, University of Michigan, Michigan (USA), (1998)
[130] Nicolais L., Nicodemo L., Polym. Eng. Sci., 13, 460 (1973).
[131] Young R.J., Beaumont P.W.R., J. Mater. Sci., 12, 684 (1977).
[132] Amdouni N., Sautereau H., Gerard J.F., J. Appl. Polym. Sci., 46, 1723 (1992).
[133] Rubiolo G.R., Marconi J.D., Goyanes S.N., Polym. Int., 51, 1290 (2002).
[134] Ishai O., Cohen L.J., Int. J. Mech. Sci., 9, 539 (1967).
[135] Kramer E.J., Advances in Polymer Science, Springer-Verlag, Berlin, (1983).
[136] Yamini S., Young R.J., J. Mater. Sci., 13, 1814 (1980).
[137] Ward I.M., Hadley D.W. An Introduction of Mechanical Properties of Solid
Polymers, In: (ed. John Wiley & Sons, Chichester, pp. (1993).
[138] Ward I.M., The Mechanical Properties of Solid Polymers, Wiley Interscience,
London, (1971).
[139] Orowan E., Prept. Prog. Phys., 12, 185 (1949).
[140] LeMay J.D., Kelly F.N. Stucture and Ultimate Properties of Epoxy Resins, In: Adv.
Chem. Ser. (ed. Dusek K.), 80, Springer, Berlin, pp. 115, (1986).
[141] Kinloch A.J., Young R.J., Fracture Behaviour of Polymers, Applied Science
Publishers, London, (1983).
[142] Cantwell W.J., Roulin-Moloney A.C., Kaiser T., J. Mater. Sci., 23, 1615 (1988).
[143] Kinloch A.J. Mechanics and Mechanisms of Fracture of Thermosetting Epoxy
Polymers, In: Adv. Chem. Ser. (ed. Dusek K.), 72, Springer, Berlin, pp. 45, (1985).
[144] Cherry B.W., Thomson K.W., J. Mater. Sci., 16, 1913 (1981).
[145] Spanoudakis J., Young R.J., J. Mater. Sci., 19, 473 (1984).
[146] Kinloch A.J., Williams J.G., J. Mater. Sci., 15, 987 (1980).
[147] Philips D.S., Scott J.M., Jones M., J. Mater. Sci., 13, 311 (1978).
[148] Chakachery E.A., Bradley W.L., Polym. Eng. Sci., 27, 33 (1987).
[149] Moloney A.C., Cudré-Mauroux N., Kausch H.H., Polym. Compos., 8, 324 (1987).
[150] Feltham S.J., Yates B., Martin R.J., J. Mater. Sci., 17, 2309 (1982).
[151] Spanoudakis J., Young R.J., J. Mater. Sci., 19, 487 (1984).
[152] Broutman L.J., Sahu S.U., Mater. Sci. Eng., 8, 98 (1971).
[153] Brown S.K., Brit. Polym. J., 24 (1980).
[154] Fox T.G., Bull. Am. Phys. Soc., 1, 123 (1956).
126 Chapter 3

[155] Rodriguez-Parada J.M., Percec V., Macromolecules, 19, 55 (1986).

[156] Pearce E.M., Kwei T.K., Min B.Y., J. Macromol. Sci. Chem., A21, 1181 (1984).
[157] Kinloch A.J., Maxwell D.L., Young R.J., J. Mat. Sci. Lett., 4, 1276 (1985).
[158] Maazouz A., Sautereau H., Gerard J.F., J. Appl. Polym. Sci., 50, 615 (1993).
[159] Azimi H.R., Pearson R.A., Hertzberg R.W., Polym. Eng. Sci., 36, 2352 (1996).
[160] Verchere D., Sautereau H., Pascault J.P., Moschiar S.M., Riccardi C.C., Williams
R.J.J., Polymer, 30, 107 (1989).
[161] Flory P.J., Principles of Polymer Chemistry, Cornell University Press, Ithaca,
(1953).
[162] Huggins M.L., J. Am. Chem. Soc., 64, 1712 (1942).
[163] Balsara N.P. Thermodynamics in Polymer Blends, In: Physical Properties of
Polymers Handbook (ed. Mark E.J.), AIP Press, New York, pp. 257, (1996).
[164] Hedrick J.L., Yilgor I., Jurek M., Hedrick J.C., Wilkens G.L., McGrath J.E., Polymer,
32, 2020 (1991).
[165] Venderbosch R.W., PhD-Thesis, Eindhovens University of Technology, Eindhoven
(NL), (1995)
[166] Meijer H.E.H., Venderbosch R.W., Goossens J.G.P., Lemstra P.J., High Perform.
Polym., 8, 133 (1996).
[167] Manzione L.T., Gillham J.K., McPherson C.A., J. Appl. Polym. Sci, 26, 889 (1981).
[168] Manzione L.T., Gillham J.K., McPherson C.A., J. Appl. Polym. Sci., 26, 907 (1981).
[169] Wu S., Polymer, 26, 1855 (1985).
[170] Kim B.S., Chiba T., Inoue T., Polymer, 34, 2809 (1993).
[171] Bucknall C.B., Gomez C.M., Quintard I., Polymer, 35, 353 (1994).
[172] De Graf L.A., PhD-Thesis, University of Twente, Twente (NL), (1994)
[173] Mezzenga R., Luciani A., Manson J.A., Polym. Eng. Sci., 42, 249 (2002).
[174] Mezzenga R., Plummer C.J.G., Boogh L., Manson J.A., Polymer, 42, 305 (2001).
[175] Mezzenga R., Boogh L., Manson J.-A.E., J. Polym. Sci. Part B, 38, 1883 (2000).
[176] Mezzenga R., Boogh L., Manson J.-A.E., J. Polym. Sci. Part B, 38, 1893 (2000).
[177] Munk P., Introduction to Macromolecules, John Wiley & Sons, Austin, (1989).
 Thermo-oxidative degradation behaviour 127

Chapter 4

4 Thermo-oxidative degradation behaviour

4.1 Introduction to thermo-oxidative degradation

Polymer degradation generally involves changes in the chemical structure of the materials
and can include oxidation, chain-scission or crosslinking (post-curing). Several authors have
intensively reviewed these processes in the past [1-3]. Struik [4] firstly separated polymer
degradation processes into two different categories based on chemical and physical
mechanisms that will be reviewed in the next sections. Before giving more details about
these two fundamental aspects of degradation, it should be pointed out that a strong inter-
relationship exists between these two modes. For example, since it is difficult to obtain a
chemically pure polymeric composite, it is not easy to separate thermal and thermo-
chemical degradation. Indeed, impurities present in the material might react with the
polymeric matrix.
Although the above-described phenomena are important for practical reasons, the focus in
the current work will be on the chemical and the thermo-oxidative degradation, which are the
main technical interests.

4.1.1 Physical ageing of epoxy-networks

In contrast to the chemical degradation, which is caused by mutual reactions with various
chemicals (e.g. acids, bases, solvents and reactive gases i.e. oxygen), the phenomenon of
physical ageing involves changes in the physical network structure, e.g. embrittlement,
increase in tensile modulus, decrease in loss modulus, increase in shrinkage and
densification of the polymer matrix, when continuously exposed to elevated temperatures for
prolonged periods. Many conventional methods have been applied to measure physical
ageing in epoxy resins, which include volumetric [5, 6], free volume [7] and enthalpy
techniques [5, 8, 9]. Besides, the phenomenon has been studied by changes in deformation
properties such as viscoelasticity [5, 10-14], yielding [8, 9, 11, 15] and fracture behaviour [5,
11, 15]. In general physical ageing is associated with the slow approach to equilibrium of a
system quenched to below its glass transition temperature [4, 16-18]. The quenching
process freezes into the system disorder (entropy), which reflects the thermal history of the
sample. It has been postulated [4] that physical ageing can be provoked by molecular
128 Chapter 4

segmental rearrangement causing a decrease in free volume, densification of the structure

and the tendency of the material to rearrange towards the thermodynamic equilibrium state.

4.1.2 Thermo-oxidative degradation of epoxy-networks

A great number of scientific publications on thermo-oxidative degradation of epoxy
thermosets and degradation effects on mechanical [19-21], viscoelastic [13, 14, 22], and
optical properties [23] were published in the past. Nevertheless, postulated degradation
mechanisms for anhydride-cured epoxy networks remain rather limited.
Lee [24, 25] has studied degradation of a variety of epoxy resins; both uncured and cured
with amine and anhydride hardeners. He showed that in all cases large quantities of
phenolic products were formed and proposed several reaction schemes. Cured and uncured
resins were not treated as two different systems because the results showed that phenolic
products were generated in both cases.
Keenan and Smith [26] used a hot filament technique to study the degradation of crosslinked
networks of a diglycidylether of bisphenol-A (DGEBA) and
p,p’-diaminodiphenylmethane (DDM). The volatile products were chromatographed utilizing
thin layer chromatography. In addition, infrared spectroscopy was carried out and an attempt
was made to characterize the residues remaining on the pyrolysis head after degradation.
The hypothesized reaction schemes supplemented those as presented by Lee [24, 25].
Leisegang and co-workers [27] studied the thermo-oxidative ageing of stoichiometric
mixtures of DGEBA cured with DDM at degradation temperatures below 305°C. The
condensable and gaseous decomposition products were collected and the major
components were separated as pure compounds and analyzed by chromatography, mass
spectroscopy and infrared spectroscopy. They proposed that at relatively low temperatures
(220°C to 240°C) a large number of chain scissions of the least stable bonds yield radicals
that may recombine, or abstract hydrogen as illustrated in Figure 90 and Figure 91. At
higher temperatures (>250°C) the dehydration reaction predominates (see Figure 92), and
consequently significantly weakens the network structure for further degradation processes.
At temperatures above 280°C, the weakened structure breaks down and small fragments on
the network appear as phenolic, benzofuryl and bisphenol-A derivatives.
 Thermo-oxidative degradation behaviour 129

This image is not available online. Please consult the

hardcopy thesis available from the QUT Library.

Figure 90: Thermally initiated aryl-methylene scissions as proposed by Leisegang et al. [27]

This image is not available online. Please consult the

hardcopy thesis available from the QUT Library.

Figure 91: Thermally initiated allyl-nitrogen scissions as proposed by Leisegang et al. [27].

-H2O
N
N
CH2 CH CH2 O
CH CH CH2 O

OH

Figure 92: Dehydration mechanism of unsaturated structures

Levchik and co-workers [28] have investigated the mechanism of thermal decomposition of
commercial TGDDM and DGEBA resins cured with Bis(m-aminophenyl) methylphosphine-
oxide (BAMPO) or 3,3’,5,5’-tetraethyl-4,4’-diaminodiphenylmethane (M-DEA) under nitrogen
and vacuum atmospheres. They observed that the decomposition consists of three different
stages. The most essential feature in the first step (<320°C) was the elimination of water by
dehydration of secondary alcohol groups leading to an unsaturation as shown in Figure 92.
This observation is in agreement with studies where a decrease in alcohol concentration
was also observed [27, 29-31]. The formed unsaturation weakens the aliphatic C-O and C-N
bonds in the β-position and consequently bond scissions lead to amine terminal functions.
When the C-N and C-O scissions occur through a molecular or free-radical mechanism with
mutual disproportionation various cyclic chain end structures, especially the quinoline-type
chain ends (Figure 93) were reported as typical degradation products of TGDDM resins [32].
130 Chapter 4

This image is not available online. Please consult the hardcopy thesis available
from the QUT Library.

Figure 93: Quinoline-type chain end formation as proposed by Levchik et al. [28]

The second (320-420°C) and the third stage of degradation (>420°C) was mainly attributed
to cleavage of remaining C-N and C-O bonds and aromatic substituted carbon bonds, which
show relatively low thermal stability in epoxy resins [27]. At the same time, carbonization of
the earlier generated conjugated structures may occur as a competition reaction.
Thermal oxidation in an intermediate temperature range (140 to 200°C) of
hexahydrophthalic anhydride (HHPA) or methyl-tetrahydrophthalic anhydride (MTHPA)
cured epoxy resins have been studied by Le Huy and co-workers [33-35]. They have
identified oxidation as a dominating ageing process. Moreover, vibrational spectroscopic
measurements revealed the reformation of anhydride structures from presumably monoacid
dangling chains by oxidation or hydrolysis as shown in Figure 94 and 95. Moreover, they
reported that the unsaturation of MTHPA is bound to the network by radical addition, yielding
to a reduction in weight-loss and obvious increase in the glass transition temperature in
contrast to HHPA cured epoxy systems.

This image is not available online. Please consult the hardcopy thesis available
from the QUT Library.

Figure 94: Anhydride monomer reformation through oxidation as proposed by Le Huy et al.
[34, 35]
 Thermo-oxidative degradation behaviour 131

This image is not available online. Please consult the hardcopy thesis
available from the QUT Library.

Figure 95: Anhydride monomer reformation through hydrolysis as proposed by Le Huy et al.
[34, 35]

4.2 Principals of dynamic mechanical analysis (DMA)

Although the Cole-Cole plot is an approved tool to study dielectric relaxation data [36-38] the
use of this method to analyse mechanical relaxation phenomena is rather limited [39]. In
general a Cole-Cole plot visualizes the linear variation of the storage modulus (E’) as a
function of the loss modulus (E’’) with varying temperature or frequency. This type of
diagram presents in a simple visual way the individual or collective nature of the microscopic
mechanisms controlling the macromolecular relaxation processes. Assuming Maxwell
behaviour, the complex modulus (E*) is defined by eqn (4.59).
E * = E ' + iE' ' (4.59)

A solution of this basic differential Maxwell equation reveals that the two frequency
dependent components, the in-phase modulus component or storage modulus (E’) and the
out-of-phase component or the loss modulus (E’’) can be separated and stated by the so-
called dispersion relations as given by eqn (4.60):
ω2 τ 2 (4.60)
E ' = E0
1 + ω2 τ 2

ωτ
E '' = E 0
1 + ω2 τ 2

where E0 is the relaxed modulus referring to the α-relaxation, ω the angular frequency, and τ
the appropriate relaxation time. Considering further the sum of the square moduli, eqn (4.60)
can be rearranged to:

[E ] + [E ]
' 2 '' 2
[ ] + [E ] = [1 + ω τ ][. E ] = [1 + ω τ ].
= ω2 τ 2 E ''
2 '' 2 2 2 '' 2 2 2 E 02 ω2 τ 2 (4.61)
[1 + ω τ ]
2 2 2

ω2 τ 2
= E 02 . = E0 E'
1+ ω τ
2 2

And, this finally leads to eqn (4.62):

132 Chapter 4

[ ]
2 2
 ' E0  '' 2 E0  (4.62)
E − 2  + E =  2 
   
Typically, if the viscoelasticity exhibits only one relaxation time, as in the case of a Maxwell
material, the Cole-Cole plot is described by a semi-circular shape, centered on the
horizontal E‘-axis at E0/2 with a radius of E0/2 (see Figure 96). On the other hand, this ideal
case does not describe the viscoelastic behaviour of more complex polymer systems where
several discrepencies can appear. For instance, when the viscoelasticity is governed by
several relaxational events, the Cole-Cole diagram will show a distortion of the semi-circle
plot appearance and exhibits more a composite curve with a set of semi-circles connected
by transition zones.

E‘‘

~Q

κ’ ·π/2 κ’ ·π/2

Eo E∞ E‘
Figure 96: Complex-plane diagram of the storage and loss modulus (Cole-Cole plot) for a
material showing an ideal Maxwell behaviour

It should be emphasized that the classical Cole-Cole diagram is two-dimensional and hence
the temperature is not directly displayed. However, it is implicitly included in the evolution of
the moduli. Therefore, it might be useful to see the evolution of the moduli with temperature.
This is shown in the Figure 97 where a three-dimensional Cole-Cole diagram giving
temperature in z-axis and projections of the moduli in the temperature-E‘ and temperature-
E‘‘ planes.
 Thermo-oxidative degradation behaviour 133

280
260
240
220

Temperature (°C)
200
180
160
140
120
100
80
60
40 8
20 2.0x10
0.0 1.5x10
8
8
5.0x10 9 8
1.0x10 1.0x10
9 )
E' (
1.5x10 7 (Pa
5.0x10 E''
9
Pa) 2.0x10
9
2.5x10 0.0

Figure 97: 3-D Cole-Cole diagram for the viscoelastic behaviour of the MY721/MA/HHPA-
system (r=0.8).

Different empirical modifications of eqn (4.62) have been proposed in attempt to represent
variations in the complex modulus (E*) [40, 41]. From a thermodynamical analysis of
deformation Perez has introduced a theoretical model that describes the linear and non-
linear stress-strain behaviour of various amorphous thermoplastics and thermosets [39, 41-
44]. Moreover, the model based on a physically consistent theory that allows interpreting the
results in terms of structural changes, when two underlying physical assumptions are
considered to be valid:
(I) The source of mobility in amorphous condensed polymers is the nanofluctuation of
density or quasi-punctual defects (qpd) with a concentration Cd. These defects can be
+
characterized by regions with positive fluctuations of packing densities (d dqp, or anti-free
-
volume) or negative fluctuations (d dqp, or free volume) as shown in Figure 98. Both
fluctuations are forming more disordered sites with less cohesive energy.

d-
d+

Figure 98: Representation of the quasi-punctual defects in amorphous polymers according

to Perez
134 Chapter 4

(II) The translational molecular mobility is described in terms of a serial process or a

hierarchical correlation leading to the characteristic mean time (τα) for a movement of a
molecular unit over a distance comparable to its size. The theory yields:
1
 τ κ (4.63)
τ α = t 0  1 
 t0 
where t0 is the scaling time parameter and τ1 corresponds to the characteristic time of an
elemental molecular movement. κ denotes the correlation parameter which is related to the
disorder and expresses of how much hierarchically constrained are the correlated molecular
entities during the deformation process. The value of κ can vary between zero and one
(0<κ<1). A value of zero corresponds to a fully constrained situation with an infinite value for
τα. In contrast a κ-value of unity reflects a constraint-free situation with τα=τ1.
Finally, Perez’s model assumes that the response to an applied stress can be described in
terms of thermo-mechanical relaxation, growth and coalescence of micro shear-domains,
resulting in a elastic and viscous deformation. As a consequence the complex modulus can
be expressed as [43]:
E∞ − E0
E * = E0 + (4.64)
1 + (iωτ α ) + Q (iωτ α )
−κ − κ'

where E0 is the unrelaxed modulus referring to the α-relaxation (glassy modulus). E∝

denotes the relaxed elastic modulus, which corresponds to the value of the modulus at the
rubbery plateau. Q is a material constant that depends on temperature and defect
concentration and is close to unity. The κ’ parameter (0<κ <κ’<1) takes into consideration
the fact that segmental mobility (i.e. κ) of polymers is sensitive by the strain above a
threshold, due to the fact that the chains are locally stretched. The maximal value of κ’∼1 is
given for molecular glass and decreases with the length of chain segments between
chemical crosslink nodes.
Besides, the Perez coefficients κ, and κ’ can additionally give valuable insight into the effect
of the architecture of the polymer network on molecular motions in the solid state. In
physical terms it means when the distance between crosslink points is small, the local
percentage extension is large and κ becomes more time dependent, which in turn causes κ’
to become small. On the other hand κ decreases when the number of obstacles (crosslink
points) is maximal, i.e. when these induced a more hierarchical correlation effect.
Consequently, eqn (4.64) describes the complex modulus (E*) of amorphous polymers in the
region of the α-relaxation by five parameters (E0, E∞, κ, κ‘, Q). Even though the expression
seems to be complicated since it includes five parameters, each parameter can be
independently deduced from the characteristic of a Cole-Cole diagram. The parameters E∞
 Thermo-oxidative degradation behaviour 135

and Eo are given by the low and high temperature limits of E*, respectively. The Perez
parameters κ and κ’ can be determined by the slopes of the G’’/G’ curve on the low and high
temperature side. Q is proportional to the maximum E’’-value on the Cole-Cole curve and
corresponds to the defect concentration Cd.

4.3 Experimental

4.3.1 Materials

A commercial multifunctional amino-glycidyl-resin, AralditeMY721 (Vantico, Switzerland)

mainly based on N,N,N’,N’-tetraglycidyl-4,4’-diamino-diphenylmethane (TGDDM) was used
as epoxy monomer for all investigations. The epoxy monomer was cured with either maleic
anhydride (MA), hexahydrophthalic anhydride (HHPA), or a hardener mixture consisting of
MA and HHPA. In addition, for some long-term experiments, the well-established high
performance epoxy system [45-48] consisting of MY721 and 4,4’-diaminodiphenylsulfone
(DDS) was selected in order to benchmark the different epoxy/anhydride-systems. All used
anhydrides and amine hardeners were supplied by Fluka and used to cure the TGDDM resin
without further purification. The selected mixing ratios of the components are given in Table
11, all reflecting a molar anhydride to epoxy ratio (r) of 0.8. For the structure of the used
materials the reader is referred to Table 2 in section 2.2.1.

Table 11: Material compositions

Chemical name Composition (phr)
/Tradename MY721/MA/HHPA MY721/MA MY721/HHPA MY721/DDS

Araldite MY721 100 100 100 100
Maleic anhydride 48 70 - -
Hexahydrophthalic 32 - 110 -
anhydride
Diamino- - - - 44
diphenylsulfone

For details on the epoxy/anhydride sample preparation the reader is referred to section
2.2.2.
The epoxy/amine-system (MY721/DDS), DDS was dissolved in the MY721 resin by
thoroughly stirring at 100°C until a homogeneous mixture was obtained. The mixture was
then degassed at p<10 mbar for 15 minutes before casting into 100°C preheated steel
molds. Afterwards curing was carried out in an air-circulating oven for 3 hours at 100°C
followed by a post-curing treatment for 8 hours at 160°C and additional 8 hours at 200°C.
136 Chapter 4

4.3.2 Isothermal ageing experiments

Samples for isothermal ageing were prepared from 4 mm thick cast plates. The specimens
were aged below the glass transition temperature in air circulating ovens at 150, 170, and
200°C under atmospheric pressure. The steady state temperature was controlled with an
accuracy of ±2.0°C. The samples were periodically removed from the oven and allowed to
cool down to room temperature in a desiccator prior to analysis.

4.3.3 Dynamic mechanical analysis (DMA)

The dynamic mechanical measurements were performed using a Perkin-Elmer DMA-7e
analyser working in three point bending. The selected span length was 15 mm. Rectangular
specimens of 18 x 4 x 2 mm3 (length x width x thickness) were cut from cured 4 mm thick
plates. All experiments were carried out at a frequency of 1 Hz and a continuous heating
rate of ϕ=3°C/min in a temperature range between 30°C to 300°C under inert helium
atmosphere. The glass transition temperature (Tg) was determined by using the maximum of
the loss factor (tanδ) as a criterion.

4.3.4 Thermal gravimetric analysis (TGA)

The dynamic thermo-gravimetric measurements were carried out on a Perkin-Elmer TGA 7
thermo-gravimetric analyzer. Non-isothermal experiments were performed at a heating rate
of 10°C/min in a temperature range between 30°C to 800°C. In all cases, samples of about
10-12 mg were cut from 4 mm cured plates, and measured either under constant nitrogen or
air gas flow of 60 ml/min.
For long-term isothermal thermo-gravimetric studies the samples (70 x 70 x 4 mm3) were
dried to a constant weight under vacuum at 70°C before ageing. The gravimetric
experiments were performed by simple periodic weight measurements of samples aged at
150, 170 and 200°C for various periods of time. Samples were allowed to cool down to room
temperature in a desiccator before weight experiments on an analytical scale with a relative
accuracy of 10-5 g were carried out. After the measurement the samples were returned to
their respective ageing environments. Results presented in this work were obtained as the
arithmetic mean of at least 3 measured samples. Quantitative reproducibility variations were
less than 2% between samples.
 Thermo-oxidative degradation behaviour 137

4.3.5 Mechanical bending test

The flexural tests were made in a three-point bending configuration according to DIN-EN-
ISO 178 [49] on a mechanical Zwick universal testing machine Z100. All experiments were
performed on test specimens (80 x 10 x 4 mm3) at room temperature. Bending tests were
performed at strain rates 2 mm/min using a bending support length of l0=60 mm. In all
figures the given value represents the arithmetic mean of at least 5 measured samples.

4.3.6 FT-IR micro attenuated total reflection (ATR) spectroscopy

The IR spectra in this study were collected using a Nicolet Nexus-870 FT-IR spectrometer
including a ContinuumTM IR microscope. The spectrometer was operating with a modulation
frequency of 20 kHz and a high sensitivity mercury cadmium telluride photoconductive cell
(MCT-detector) at 77K. The ATR lens incorporated a silicon (Si) internal reflection element
(IRE). All samples were polished before examination in order to reach a flat contact area.
The contact area was circular with an approximate diameter of 100 micron. Reproducible
contact between the surface of the epoxy sample and the ATR-crystal was achieved by use
of an electronic sensor plate attached to the microscope. Typically, 128 co-added scans
were recorded at 4 cm-1 resolutions in a range of 600 to 4000 cm-1. Background spectra
were obtained through the ATR element when it was not in contact with the sample. For line
maps the stage was moved under computer control with a 20 micron step size.
The control of the penetration depth (dp), the depth at which the electrical field amplitude is
reduced to ∼37% (factor e-1 of its initial value), is defined by equation (4.65) as first proposed
by Harrick [50].
λ
dp = (4.65)
2π (sin Θ − n 21
n12 2
)
0.5

where λ is the wavelength of the radiation beam and Θ is its incident angle (∼45°); n1 is the
refractive index of the Si-IRE (n1= 3.4) and n21 is n2/n1, where n2 is the refractive index of the
organic matrix. Given a sample refraction index of 1.5, which is appropriate for many
polymers [51], then the range of penetration depth that can be achieved is approximately
varying between 14.6 micron at 600 cm-1 to 2.2 micron at 4000 cm-1, respectively.
138 Chapter 4

4.4 Results and discussion

4.4.1 Mechanical endurance during thermo-oxidative ageing

4.4.1.1 Dynamic mechanical behaviour of the virgin state

To characterize the viscoelastic properties of the virgin systems, dynamic mechanical
analysis (DMA) measurements were performed at 1 Hz. The storage modulus (E’) and loss
factor (tanδ) curves for the different systems are given in Figure 99. In each case, a
continuous decrease in the storage modulus (E’) with increasing temperature between 35 to
200°C was obtained; see Figure 99(a). This behaviour can be explained by thermal
expansion that induces higher free volume. At temperatures above 200°C the investigated
systems reached their glass to rubber transition region. This transition is assigned to the α-
relaxation, which is manifested by a sharp decrease of the elastic modulus over nearly two
orders of magnitude, before leveling off on the rubber modulus plateau. The α-relaxation
phenomenon can be attributed to collective chain movements and hence it is associated
with the macroscopic glass transition temperature (Tg) of the studied networks. A more
pronounced indication of the α-relaxation mode could be seen in the tanδ-curve as shown in
Figure 99(b). The values of the tanδ-maxima for the different investigated systems are listed
in Table 12. The HHPA cured system revealed the highest α-relaxation (243°C) followed by
the MA-cured (240°C) and the MA/HHPA-cured system (215°C). However, an unambiguous
assignment of Tg in the case of the MY721/MA-system is difficult because the tanδ curve
exhibits a broad peak appearance and hence no pronounced maximum temperature. The
fact that the amplitude of the loss factor is maximal and the width of the α-peak (the glassy
dispersion) is much sharper for the MY721/HHPA-system may be explained by a higher
homogeneity of the MY721/HHPA network. On the other hand, the presence of MA tends to
initiate a more heterogeneous network structure, which is reflected in a broad α-peak. It is
hypothesized that the more inhomogeneous network presumably can be related to the
inherent dangling chains (mono-acids, e.g. -CH=CHOOH) in the presence of MA (see
section 2.3.4).
 Thermo-oxidative degradation behaviour 139

Storage Modulus [Pa]

1E9

75 100 125 150 175 200 225 250 275

Temperature [°C]

a)

0.35
α

0.30

0.25
α
tan δ [-]

0.20
α

0.15

ω ω
0.10

0.05

75 100 125 150 175 200 225 250 275

Temperature [°C]

b)
Figure 99: Viscoelastic behaviour as a function of temperature for (a) storage modulus (E’),
and (b) loss factor (tanδ), as experimentally determined by DMA of virgin samples (ˆ)
MY721/MA, (‰) MY721/HHPA, (Ž) MY721/MA/HHPA with a molar anhydride/epoxy ratio,
r=0.8.

Table 12: Characteristic DMA values of the different investigated epoxy networks.
Systems Tg (°C); (1 Hz) (tanδ)max Tcure/Tg
MY721/HHPA/MA 215 0.21 0.64
MY721/HHPA 243 0.33 0.81
*)
MY721/MA 240 0.10 0.60
*)
Precise determination difficult due to very broad tanδ-peak

Additionally, the co-existence of an additional sub-glass relaxation for MA-cured systems in

an intermediate temperature range is noticeable from the behaviour of tanδ (Figure 99 (b)),
even when the recorded E’-curve shows no significant change and remains on the same
140 Chapter 4

order of magnitude over the entire transition range. This intermediate temperature relaxation
(50-100°C for MY721/MA-system and 115-150°C for the MY721/MA/HHPA-system), is often
recognized as ω-relaxation, and has been the subject of numerous studies in the past [52-
54]. Unfortunately, in spite of all these efforts a fully confirmed explanation of the ω-
relaxation in terms of molecular mobility remains a controversial matter. While some
researchers attributed the secondary ω-relaxation to unreacted molecular segments, chain
ends and inhomogeneities in the network [55, 56], other authors assigned the ω-relaxation
to motions of the p-phenylene groups. In addition, it has been observed that the ω-relaxation
is very sensitive to either ‘free’ or ‘bound’ water in the network and thus some considered
this intermediate relaxation as rising from micro-structural rearrangement in combination
with moisture absorption [53, 57, 58]. Although the explanation of the ω-relaxation is not
unambiguous here, we assume that the ω-transition can be mainly assigned to the presence
of dangling chain-ends within the maleic anhydride-cured epoxy network-structure.
Classical Cole-Cole plots of all initial systems are shown in Figure 100. As expected, the
Cole-Cole plots illustrate different thermo-mechanical behaviour depending on the chemical
nature of the hardeners used; hence each epoxy system exhibits a different viscoelastic
spectrum. The MY721/HHPA-system displays mainly a high-temperature relaxation,
associated with the primary α-transition that leads to an almost symmetric Cole-Cole plot
appearance. On the other hand the MY721/MA-system shows a more complex curve
consisting of partially overlapping α and ω-relaxations. It is surprising that the intermediate
ω-dissipation (∼100°C), reveals a higher E’’-value than the primary
α-relaxation (∼220°C). In a similar way, the co-anhydride MA/HHPA cured system also
displays a non-symmetric Cole-Cole plot. In this case, however, the transition zone between
the two different relaxational events was significantly decreased due to a reduced MA
content. The Perez parameters of the different systems were graphically determined and the
obtained values are summarized in Table 14. Since the theoretical boundary criterion of
1>κ‘>κ>0 (see section 4.2) is fulfilled for all examined systems the molecular model can be
applied to explain the experimental results. Despite this, the calculated values are in a good
agreement with reported results from similar studies [39, 41].
 Thermo-oxidative degradation behaviour 141

250

200

E'' (MPa)
150

100 c)
b)
a)
50

0
0 500 1000 1500 2000 2500 3000
E' (MPa)

Figure 100: Cole-Cole plot for virgin epoxy-networks: (a) MY721/HHPA; (b) MY721/
MA/HHPA; (c) MY721/MA with a molar anhydride/epoxy ratio, r=0.8

Table 13: Numerical Perez parameters in the glass transition region for the investigated
epoxy networks in the virgin state
System κ‘ κ E0 E∞ E∞/E0
(MPa) (MPa)
*)
MY721/MA 0.308 0.221 310 2350 7.7
MY721/HHPA 0.516 0.192 95 1960 21
MY721/HHPA/MA 0.455 0.066 105 4300 41
* Interference with the ω-relaxation
)

4.4.1.2 Effects of thermo-oxidative ageing on the dynamic mechanical behaviour

A comparison of changes in the relaxational spectra during thermo-oxidative ageing can
help to understand the micro-mechanisms involved in the ageing process. Therefore,
periodical DMA runs were performed over an ageing period of up to 1000 hours at 200°C.
Considering the fact that the DMA apparatus is relatively accurate in phase measurements
but less precise in determining the absolute components of the complex modulus, the
storage- and loss-moduli values were normalized with respect to their value at room
temperature. The normalized storage modulus for all investigated systems as a function of
ageing temperature and time is shown in Figure 101.
For the pure MA-cured system a continuous increase of the glassy storage modulus and the
α-relaxation was observed over the entire ageing period, see Figure 101(a). This is a
common behaviour of various thermosets during the early ageing period [59]. It can be
generally attributed to the “under-cured” state of the industrial materials, which can be only
rarely avoided. As a result these materials undergo further post-crosslinking as soon as they
are subjected to higher temperatures. However, a common post-curing process over
142 Chapter 4

1000 hours of ageing can only hardly explain the steady shift of the modulus-cure towards
higher temperature. Therefore, it is proposed that in addition the conjugation of maleic
anhydride tends to react by radical addition at higher temperature, consequently yielding an
increased network density of the MA-cured networks.

a) b)
1
1000hrs 0 hrs
Storage modulus E/E0 [-]

1 20hrs

Storage modulus E/E0 [-]

0 hrs
475hrs 0.1
717hrs
475hrs

717hrs
0.01

1000hrs

0.1
50 100 150 200 250 300
50 100 150 200 250 300
Temperature [°C]
Temperature [°C]

c)
1

90hrs
Storage modulus, E/E0 [-]

0.1 0 hrs

475hrs

186hrs

1000hrs

50 100 150 200 250 300

Temperature [°C]

Figure 101: Normalized storage modulus (E/E0) from DMA-measurements as a function of

ageing time during thermo-oxidative ageing at 200°C for (a) MY721/MA, (b) MY721/HHPA,
and (c) MY721/MA/HHPA with a molar anhydride/epoxy ratio, r=0.8.

On the other hand the pure HHPA-cured and the co-anhydride-cured network behaved in a
completely different way. The analysis of the experimental dynamic mechanical data clearly
separated the thermo-oxidative ageing processes into two co-existent stages. In the early
period of exposure a post-curing step is predominant (~20 hours for the MY721/HHPA and
~90 hours for the MY721/MA/HHPA-system, respectively), which results in a shift of the
sharp modulus drop (α-relaxation) towards higher temperatures. The obtained increase in
the crosslink density coincides with decrease in rigidity in the glassy state (T<Tα). However,
with further increasing ageing time the modulus curve tends to shift again towards lower
temperatures. This behaviour is correlated to an increase of the glassy modulus. The
observed decrease in the glass transition temperature, caused mainly by irreversible
structural network changes can be considered as a decrease in network density. Moreover,
 Thermo-oxidative degradation behaviour 143

the increase in the glassy modulus with decreasing crosslink density is known as
antiplasticization phenomena [59, 60].
The continuous network degradation has a profound influence on the evolution of the glass
transition temperature as shown in Figure 102. The MY721/MA-system exhibits an increase
of the Tg-value over the entire ageing period. This observation provides further evidence that
the intrinsic conjugation of MA, tends to react during thermo-oxidative ageing as discussed
before. In addition, this behaviour is consistent with a study by Le Huy et al. who reported
comparable thermo-oxidative ageing behaviour for epoxy resin cured with unsaturated
methyl-tetrahydrophthalic anhydride (MTHPA) [35].

280

260
Glass transition temperature [°C]

240

220

200

180
ageing temperature
160

140

120

0 200 400 600 800 1000

Ageing time [h]

Figure 102: Behaviour of the glass transition temperature (Tg), determined by the maximum
of tanδ, as a function of ageing time at 200°C for the (0) MY721/MA, () MY721/HHPA,
() MY721/MA/HHPA-system. A molar anhydride to epoxy ratio, r=0.8, was used for all
samples.

In both HHPA-containing systems the competition between the post-curing and degradation
mechanisms is noticeable by a distinct maximum in the glass transition temperature in
Figure 102. The observed behaviour of Tg for the MY721/HHPA- and the MY721/HHPA/MA-
system revealed fundamental similarities between the two systems. Both Tg’s decreased
below the ageing temperature of 200°C after approximately 400 hours and 700 hours for the
MY721/HHPA and the MY721/MA/HHPA-system, respectively. The observed behaviour
indicates that the presence of HHPA weakens the network structure in terms of thermo-
mechanical properties. Consequently, HHPA-cured networks are more susceptible to
thermo-oxidative depolymerization processes, e.g. by chain–scission, as proposed by Le
Huy and Levchik [29, 35]. The experimental observations tend to support the hypothesis that
the MY721/MA/HHPA-system has a multi-phase network structure, which comprises
144 Chapter 4

independent MY721/HHPA and the MY721/MA sub-networks. This conclusion is consistent

with experiments performed in section 2.3.2.3 and 3.5.4.1.
The dynamic evolution of the Cole-Cole plot as a function of thermo-oxidative ageing is
shown in Figure 103. In particular for the MY721/MA-system (Figure 103(a)) the two Perez
coefficients (κ and κ’) only slightly decreased (see Table 14). This corresponds to an almost
unchanged network density, which is in agreement with the obtained Tg-evolution.
Supporting evidence for the disappearance of the intermediate ω-relaxation with ongoing
exposure was provided by the Cole-Cole analysis. It can be seen that the influence of the ω-
relaxation on the Cole-Cole plot diminishes continuously and the relaxational spectrum for
the MY721/MA-system becomes mono-relaxational after about 700 hours.

a) 260 600
1000hrs b)
240 550
0 hrs 1000hrs
220 500
200 450
180
475hrs 400
160 717hrs
E'' [MPa]

E'' [MPa] 350

140
120 300
100 250
80 717hrs 20hrs
200
60 150
40 0 hrs
100
20
50
0
0
0 500 1000 1500 2000 2500 3000 3500 0 500 1000 1500 2000 2500 3000
E' [MPa] E' [MPa]

c) 320 1000hrs

280
717hrs 0 hrs
240
E'' [MPa]

200

160

120
475hrs
80

40

0
0 500 1000 1500 2000 2500 3000
E' [MPa]

Figure 103: Dynamic Cole-Cole plot as a function of different ageing times for (a)
MY721/MA, (b) MY721/HHPA, (c) MY721/MA/HHPA-system during thermo-oxidative ageing
at 200°C. A molar anhydride to epoxy ratio, r=0.8, was used for all samples.

For the MY721/HHPA-system (Figure 103(b)) the peak maximum in the Cole-Cole diagram
continuously shifted towards higher E’-values. This thermal behaviour is consistent with a
decrease in the measured Tg and coincides with the fact that after about 700 hours of
ageing some of the samples were completely destroyed. Information on the network
 Thermo-oxidative degradation behaviour 145

architecture on molecular motions can be obtained from the κ and κ‘ parameters as shown
in Figure 104. In physical terms the observed decrease in the κ’ parameter in the early
period of exposure can be related to a post-curing reaction (increase of crosslink density).
The consequently followed monotonous increase in the κ’-value with increasing ageing time
can be correlated with the increased distance between crosslinks points (decrease of
crosslink density) due to irreversible network degradation. Moreover, this behaviour is
consistent with the behaviour of κ as function of ageing time, where κ increases when the
crosslink density decreases. The fact that for both HHPA-containing samples the
parameters κ and κ’ converge to unity with increasing ageing time reflects the sample
conversion towards molecular glass (κ’=1) as proposed by Cavaillé [61].

1.0
b)
0.9

0.8

0.7
κ' 0.6

0.5 c)

0.4

0.3 a)

0.2

0 200 400 600 800 1000

Time [h]

Figure 104: Behaviour of the Perez parameter (κ’) as a function of ageing time at 200°C for
(a) MY721/MA, (b) MY721/HHPA, (c) MY721/MA/HHPA. A molar anhydride to epoxy ratio,
r=0.8, was used for all samples.

In case of the MY721/MA/HHPA-system the Cole-Cole plot, as shown in Figure 103(c),

shows a composite curve of the superimposed mono-hardener systems. The dynamic
disappearance of the ω-relaxation with progressing ageing becomes visible in the
development of a mono-relaxational Cole-Cole plot, which is consistent with the behaviour of
the pure MA cured sample. In comparison to the pure HHPA system, the composite system
exhibited a delayed onset of thermal degradation. This effect can be attributed to the
partially substituted HHPA content with MA and hence to the higher thermo-oxidative
stability of the MA sub-network structure.
146 Chapter 4

Table 14: Numerical Perez parameters in the glass transition region for different investigated
networks as a function of thermo-oxidative ageing time at 200°C
System Ageing time [h] κ‘ κ E0 E∞ E∞/E0
(MPa) (MPa)
*
MY721/MA 0 0.308 0.221 310 2350 7.7
*
90 0.264 0.140 90 2050 23
475 0.261 0.213 18 2350 130
717 0.264 0.208 17.5 1750 100
1000 0.268 0.188 17 2900 171
MY721/HHPA 0 0.516 0.192 95 1960 21
20 0.504 0.172 75 2200 29
186 0.456 0.181 100 3500 35
717 0.840 0.263 40 2800 70
1000 0.971 0.500 3 2750 917
MY721/HHPA/MA 0 0.455 0.066 105 4300 41
20 0.189 0.070 49 2800 57
90 0.263 0.076 32 3800 119
186 0.286 0.058 30 4050 135
475 0.369 0.083 40 3100 78
717 0.381 0.392 20 2300 104
1000 0.471 0.692 9.5 2150 231
* Interference with the ω-relaxation
)

4.4.1.3 Effects of thermo-oxidative ageing on the flexural behaviour

As shown in the previous section the thermo-oxidative ageing has a significant effect on the
viscoelastic properties of the materials. In order to investigate the correlation between the
physico-chemical and the mechanical behaviour the flexural properties as a function of
thermo-oxidative degradation were studied. The well-established MY721/DDS-system was
used as a reference system in order to benchmark the different epoxy/anhydride systems.
The initial values of the flexural strength and the elastic modulus for the studied systems are
listed in Table 15. It can be seen that the flexural strength values of all three anhydride-
cured systems are comparable at σrm=86±4 MPa. In contrast the amine-cured system shows
a significantly higher flexural strength.

Table 15: Virgin flexural properties of the investigated epoxy-systems

System Flexural strength [MPa] Young’s modulus [GPa]
MY721/MA 86 3.5
MY721/HHPA 82 3.2
MY721/MA/HHPA 90 3.4
MY721/DDS 129 3.71
 Thermo-oxidative degradation behaviour 147

The dynamic changes of the ultimate flexural strength (σrm) as a function of ageing time and
temperature are presented in Figure 105. A general trend for all investigated samples was
obtained. The ultimate strength of the samples first increases until the ductile-brittle
transition is reached. Then the flexural strength passes through a maximum and decreases
again into the ‘brittle domain’. It is obvious that the maximum of the flexural strength
depends on the nature of the used curing agent and related crosslinking structure. The
highest flexural strength was obtained for the MY721/DDS network (σrmax≈150 MPa),
followed by MY721/MA (σrmax≈140 MPa), MY721/MA/HHPA (σrmax≈115 MPa), and
MY721/HHPA-system (σrmax≈100 MPa). Moreover, the obtained σrm-maxima of the
anhydride-cured system are consistent with the Tg-maxima as already discussed in section
4.4.1.1. This fact provides evidence that the mechanical changes can be strongly related to
the crosslink density variations occurring during thermo-oxidative degradation. If the glass
transition temperature increases due to a post-curing effect, the increased crosslink density
induces an increase of the yield stress [62]. At a certain point the yield stress becomes too
high, consequently the mechanism of brittle fracture becomes competitive, and the sample
fractures already in the elastic area at comparatively low deformation resulting in a decrease
of the flexural strength.

140 b) 140
a) 150°C
130 130 150°C
170°C 170°C
120 200°C 120
200°C
110 110
Flexural strength [MPa]

Flexural strength [MPa]

100 100
90 90
80 80
70 70
60 60
50 50
40 40
30 30
20 20

0.1 1 10 100 1000 10000 0.1 1 10 100 1000 10000

Time [h] Time [h]

c) 140 d)
160 150°C
130 150°C 170°C
120 170°C 140 200°C
200°C
110
120
Flexural strength [MPa]
Flexural strength [MPa]

100
90 100
80
80
70
60 60
50
40
40
30 20
20
0
0.1 1 10 100 1000 10000 0.1 1 10 100 1000 10000
Time [h] Time [h]

Figure 105: Flexural strength behaviour as a function of degradation time and temperature in
air for (a) MY721/MA, (b) MY721/HHPA, (c) MY721/MA/HHPA, (d) MY721/DDS.
148 Chapter 4

To compare the long-term mechanical behaviour of the studied materials the related 5000
hours temperature index (TI5000) was calculated according to DIN-IEC216/T1 [63] wherein as
a general end-of-life criterion is defined as a 50% reduction of the initial flexural strength.
Moreover, the analysis of TI5000 data is based on the assumption that there is a linear
relation between the logarithm of time to the end-point and the reciprocal of the
thermodynamic ageing temperature (Arrhenius dependency). The thermal endurance
graphs for the different epoxy-systems are shown in Figure 106 and the calculated TI5000-
values are given in Table 16.
Considering the calculated TI5000 for a lifetime prediction it can be stated that the thermo-
oxidative stability can be ranked as: MY721/MA>MY721/MA/HHPA≥MY721/DDS
>MY721/HHPA. Therefore, it can be seen that the stability sequence is related to the
hardener nature. However, the fact that different anhydride-cured diglycidylester of
bisphenol-A (DGEBA) systems show TI5000-values significantly above 220°C [21, 64, 65],
gives evidence that the currently studied amino-glycidyl resin structure has more impact on
the thermo-oxidative stability than the chemical nature of the used anhydride hardener.
Thus, it can be concluded that the presence of relatively unstable C-N bonds from the
TGDDM-resin monomer is predominantly involved in the molecular breakdown-process.

10000
Time [h]

1000

100
80 90 100 110 120 130 140 150 160 170 180 190 200 210 220
Temperature [°C]

Figure 106: Thermal endurance graph for calculation of the 5000 hours temperature index
(TI5000). The symbols represent () MY721/MA, (0) MY721/HHPA, () MY721/MA/HHPA,
(&) MY721/DDS. The dotted lines represent the extrapolated TI5000-values.
 Thermo-oxidative degradation behaviour 149

Table 16: 5000 hours temperature index (TI5000) of the investigated epoxy-systems. All
TI5000-values were calculated using a 50% reduction of the initial flexural strength as end-
criterion [63, 66].
System TI5000 [°C]
MY721/MA 153
MY721/HHPA 76
MY721/MA/HHPA 137
MY721/DDS 131

4.4.2 Short-term thermo-oxidative stability

The short-term thermal stability in air (thermal-oxidation), and in nitrogen atmosphere was
studied by thermo-gravimetric analysis (TGA). The thermograms, as recorded from
continuous heating experiments, are shown in Figure 107. It can be seen that the
degradation process is accompanied by the release of volatiles above temperatures of
300°C for the systems heated in air (Figure 107(a)-(b)), leading to a decrease in mass. On
the other hand for the system heated in nitrogen, a significant weight increase (up to about
10%) was observed. This behaviour can be probably attributed to a buoyancy effect, though
10% increase in mass is very large. Despite, chemical reactions between the nitrogen
atmosphere and the cured epoxies cannot generally excluded, even if nitrogen is used in
TGA as an inert gas.

1a) 1b)
2
100
HHPA 0
90
MA/HHPA
-2
80 MA
dm/dt [%/min]

70 -4
m/m0 [%]

60 II III
-6
50
-8
40
-10 I
30
-12 MA
20
MA/HHPA
10 -14 HHPA
0
-16
100 200 300 400 500 600 700 800
100 200 300 400 500 600 700 800
Temperature [°C]
Temperature [°C]
150 Chapter 4

2a) 120 2b) 2

110
0
100
90 -2

dm/dt [%/min]
80
-4
m/m0 [%]

70
60 -6
50
-8
40
30 HHPA -10 HHPA
MA/HHPA HHPA/MA
20
MA -12 MA
10
0 -14
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800
Temperature [°C] Temperature [°C]

Figure 107: TGA-run (ϕ=10°C/min) of MY721-systems (r=0.8) cured with different anhydride
hardeners. (a) relative weight-loss; (b) derivative of weight-loss (dm/dt), under (1) air and (2)
nitrogen atmosphere.

For a quantitative comparison of different investigated epoxy-networks, values of the

following parameters were determined: the degradation onset temperature (Ti), the
temperature at 50% weight-loss (T50) and the temperature of maximum extent of
degradation (Tmax). Since the investigated samples exhibited a pronounced double
degradation mechanism under air atmosphere, the two individual mass losses m1 and m2, as
well as the ultimatively mass loss (m∞) after Tmax, were analyzed as well. Additionally, the
corresponding temperature at the maximum rates of mass loss Tm1 and Tm2 for the individual
degradation processes were determined from the peaks of the appropriate differential
thermo-gravimetric (DTG) curves. The characteristic data obtained from dynamic TGA-
experiments in different atmospheres are shown in Table 17 and 18.

Table 17: Characteristic data obtained from dynamic TGA performed in air
System Ti T50 Tm1 Tm2 dm/dt max1 dm/dt max2 m1 m2
[°C] [°C] [°C] [°C] [%/min] [%/min] [%] [%]

MY721/MA 302 491 308 622 6.8 2.0/2.1 69.3 2.6

MY721/HHPA 341 389 358 564 15.1 2.2 49.0 5.6
MY721/MA/HHPA 306 472 340 588 5.2 2.3 61.4 5.3

Table 18: Characteristic data obtained from dynamic TGA performed in nitrogen
System Ti T50 Tmax1 dm/dt max m1 = m∞
[°C] [°C] [°C] [%/min] [%]
MY721/MA 308 549 341 5.6 46.3
MY721/HHPA 331 394 359 12.5 32.5
MY721/MA/HHPA 301 598 342 6.9 47.4
 Thermo-oxidative degradation behaviour 151

It is obvious that a more complex weight-loss behaviour was observed in air (Figure 107;
1(a)-(b)). Based on the number of peaks in the appropriate DTG-thermograms (Figure 107;
1(b)) the decomposition process can be divided at least into a two-stage mechanism. For
maleic anhydride-containing systems a third underlying, but low amplitude stage, located
between the two main stages, is visible. This observation indicates that the presence of MA
results in a more heterogeneous network structure and different susceptibilities to thermal
oxidation. It is proposed that the slightly increased weight-loss of the two MA-cured systems
at temperatures below Ti, can be mainly attributed to the increased moisture content
generated during the curing reaction (cf. Chapter 2).
On the other hand the thermograms in nitrogen atmosphere show that only a one-stage
degradation mechanism occurred (Figure 107; 2(a)-(b)). The residual weight at the ‘pseudo-
plateau’ as well as the fact that the first degradation stage shows a similar mass-loss profile
in both air and nitrogen indicates that presumably thermolysis, i.e. non-oxidative degradation
reactions, seems to dominate the degradation process in this temperature region. This
observation is in agreement with earlier epoxy ageing studies [67-70]. The fact that in
nitrogen atmosphere the high temperature degradation stage is effectively suppressed
provides evidence that this decomposition stage can be most probably attributed to
reactions between oxygen from the air and epoxy (oxidation). In addition it might be possible
that reactions products formed during the first, low temperature degradation step react with
oxygen in the second step. Combined TGA and heat flow measurements that have been
recently carried out by Lutter-Kayser [71] on the same system, revealed a higher exothermal
heat flow for the second stage of degradation. These experimental data confirmed the
conclusion that the high temperature decomposition stage can be most probably related to
substantial oxidation, since oxidation processes generally increase the sample temperature
and the measured heat flow.
The fact that the decomposition temperatures are generally lower (<350°C) than reported
values for DGEBA-based epoxy resins [72, 73] emphasizes the reduction of the thermo-
oxidative stability due the presence of amino-glycidyl-resin. Considering that the two
nitrogen atoms in the TGDDM-adducts are more basic than the DEGBA-counterparts, it is
proposed that the related nitrogen interactions are destabilizing the network structure
resulting in an adverse effect on the thermo-oxidative stability of TGDDM based networks as
also shown for the mechanical endurance in section 4.4.1.3.
The apparent thermo-oxidative stability was assessed by the value of the initial degradation
temperature (Ti). Using this criterion, amongst the neat resin samples, the most stable was
the MY721/HHPA-system. The thermo-oxidative degradation for this system starts at around
340°C. However, compared to the two other investigated systems, the maximum weight-loss
rate of about 15%/min after the onset of degradation, reflected a very rapid degradation
152 Chapter 4

reaction. An indication that the presence of MA significantly affects the thermal degradation
of the cured epoxy systems is given by the fact that both, the first and the second stage of
the thermal degradation mechanism (see Table 17 and Table 18) are affected. While MA
decreased the onset temperature of the first degradation stage, it retarded the second
thermal degradation step as manifested by the corresponding Tm1 and Tm2-values (cf. Table
17).
Numerous publications have described methods for the determination of kinetic parameters
by mathematically processing the appropriate TGA-data [74-79]. An extensive study on the
degradation kinetics using thermo-gravimetric analysis on the same epoxy systems has
been recently performed by Lutter-Kayser [71]. However, due to very complex and
simultaneously occurring degradation processes, the applied empirical and semi-empirical
methods have not resulted in a satisfactory mathematical description of the overall
decomposition process. In this work the estimation of the overall activation energy for the
weight-loss was used as quantitative measure. For the studied MY721/MA/HHPA-system an
integral method as originally proposed by Flynn-Wall [78] and Ozawa [76, 77] was applied.
This method has been widely used for different epoxy networks [13, 79] as it allows for a
mathematical description without a prior knowledge of the involved mechanisms of
degradation and how those mechanisms change with conversion. The Flynn-Wall relation is
given by eqn (4.67):
 AE a 
log(ϕ) = 
0.457E a
 − 2.315 − (4.66)
 αR  RT
where the apparent activation energy (Ea) and the corresponding pre-exponential factor (A)
can be graphically calculated from a plot of logarithm of heating rate (ϕ) as a function of
reciprocal temperature (1/T), for a constant extent of thermal degradation (α) as defined by
eqn (4.68):
m0 − m
α= (4.67)
m0 − m ∞
where m, m0, m∞, are the actual, initial and ultimate weight of the sample, respectively.
This finally leads to the activation energy as a function of the extent of degradation as shown
for the MY721/MA/HHPA-system in Figure 108. However, it should be noted that
conclusions only drawn from measured values of Ea could lead to erroneous results with
regards to thermal stability. The obtained degradation behaviour shows an almost linear
increase of the activation energy with increasing extent of degradation. The calculated
average activation energies of 174 kJ/mol (0.06<α<0.3) is in a good agreement with
previously reported values on TGDDM based epoxies, which vary between 124 kJ/mol for a
TGDDM/DDS [79] and 191 kJ/mol for a TGDDM/ DDS/DDM-system [67], respectively.
 Thermo-oxidative degradation behaviour 153

220

210

200

Ea [kJ/mol]
190

180

170

160

150
0.05 0.10 0.15 0.20 0.25 0.30
Extent of degradation [-]

Figure 108: Activation energy (Ea) as a function of the extent of thermo-oxidative

degradation as determined by the Flynn-Wall-Ozawa method [78] for the MY721/MA/HHPA-
system (r=0.8).

4.4.2.1 Effects of the molar anhydride to epoxy ratio

The characteristic property profile of epoxy thermosets is strongly coupled to the resin
stoichiometry and related network variations [80]. Therefore, in this section the effects of
variations in the stoichiometry on the thermal–oxidative stability will be investigated more in
detail. Thermograms obtained from dynamic TGA-measurements of the MY721/MA/HHPA-
systems with different anhydride to epoxy ratios are shown in Figure 109. It can be seen that
the characteristic weight-loss and rate curves revealed no significant difference in terms of
thermo-oxidative stability where identified for the r=0.64 and r=1.0 systems, since the onset
of degradation (Ti) started at around 300°C. Only a slightly higher Ti-value (314°C) was
obtained for the r=0.64 system. Therefore, it can be concluded that thermo-oxidative stability
is primarily independent of the molar anhydride to epoxy ratio in the investigated variation
range between 0.64 and 1.0.

a) 110 b)
1 0.64
100
0.8
90 0.64 0 1.0
0.8
80 -1
dm/dt [%/min]

1.0
70
m/m0 [%]

-2
60
50 -3 II III
40
-4
30
20 -5

10 -6 I
0
100 200 300 400 500 600 700 800 100 200 300 400 500 600 700 800

Temperature [°C] Temperature [°C]

Figure 109: Non-isothermal thermo-gravimetric analysis of the MY721/MA/HHPA-system in

air atmosphere using different molar anhydride/epoxy ratios. (a) relative weight-loss; (b)
derivative of weight-loss (dm/dt).
154 Chapter 4

4.4.3 Long-term thermo-oxidative stability

During long term thermal ageing in air, significant color changes appeared in all samples.
The initial red color changed successively to green-blue, then brown and finally black. This
observation, previously reported by others [20, 81], is indicative of chemical network
changes, e.g. increase of conjugation and final carbonization that occurred due to thermo-
oxidation processes.
The specific long-term weight-loss behaviour was examined by isothermal ageing
experiments that were carried out at 150, 170, and 200°C, respectively, and the weight-loss,
m(t), was calculated using eqn (4.69):
m − mo
m( t ) = (4.68)
mo
where m0 and m are the initial weight and the weight at time, t, respectively. The weight-loss
behaviour of the different anhydride-cured systems are shown in Figure 110. In order to
benchmark the different studied epoxy/anhydride-systems the well-investigated
MY721/DDS-system [45-48] was selected as high performance epoxy system.

a) b)
MA 5 MA
MA/HHPA MA/HHPA
2 HHPA
HHPA
DDS 4 DDS
Weight loss [%]

Weight loss [%]

1
2

0
0
0 200 400 600 800 1000 0 200 400 600 800 1000
Time [h] Time [h]

c) 26 d) 26
24 MA/HHPA 24 200°C
22 MA 22 170°C
HHPA 150°C
20 20
DDS calculated
18 18
Weight loss [%]

Weight loss [%]

16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Time [h] Time [h]

Figure 110: Isothermal weight-loss curves (4 mm thick plates) as a function of ageing time
for the unfilled MY721-systems using different hardeners at (a) 150°C, (b) 170°C, and (c)
200°C. The dotted line in graph (d) represents the calculated weight-loss line for the
MY721/MA/HHPA system using a molar mixing rule of the two individual anhydrides
(MA/HHPA=0.7/0.3) and the experimental data.
 Thermo-oxidative degradation behaviour 155

As seen from Figure 111 the thermo-oxidative stability depends strongly on the nature of the
used hardeners. It should be noted that the time-temperature equivalence principles are also
valid, i.e. a short ageing time at higher temperature is comparable with a long ageing time at
lower temperature. Thus, for ageing temperatures of 150°C and 170°C, the same stability
sequence was obtained; see Figure 110(a) and Figure 110(b). The most stable system was
the MY721/HHPA-system (∆m about 0.6% at 150°C and 1.2% at 170°C after 1000 hours)
followed by the MY721/DDS-system (∆m about 0.9% at 150°C and 1.7% at 170°C after
1000 hours). This observation is consistent with the dynamic TGA results, where the highest
measured onset -temperature (Ti) reflects the initial low weight-loss of the MY721/HHPA-
system under isothermal long-term ageing conditions. The MY721/MA-system was the least
thermo-oxidative stable and a weight-loss (∆m) of 2% at 150°C and 5% at 170°C after 1000
hours was obtained.
The MY721/MA/HHPA-system exhibited a fairly good thermo-oxidative stability resulting in
∆m of about 1.8% and 3% after 1000 hours at 150°C and 170°C, respectively. Hence, the
co-anhydride cured system can be considered as a “composite” system consisting of two
underlying sub-network structures. This finally results in an adaptive weight-loss behaviour
of the two individual MY721/MA and MY721/HHPA sub-networks. As a result, a prediction of
the overall weight-loss behaviour by applying a simple molar mixing rule
(MA/HHPA=0.7/0.3) leads to a fairly good agreement with the experimental data at 150 and
170°C (cf. Figure 110). The overall weight-loss in the early ageing period is mainly attributed
to higher thermal susceptibility of the MY721/MA sub-network. In the later ageing period
(>500 hours) the experimentally obtained weight-loss is primarily dominated by the rapid
degradation behaviour of the MY721/HHPA sub-network.
However, the previously obtained sequence of stability changed when the samples were
exposed to an ageing temperature of 200°C. In this case, the thermo-oxidative stability of
the MY721/HHPA-system is only superior compared to the amine-cured system
(MY721/DDS) up to an ageing time of 500 hours. After 500 hours of thermal exposure, the
shape of the corresponding weight-loss curve changed and the weight-loss increased
rapidly from this point on (Figure 110(c)). Moreover, it should be noted that all specimens
were unanimously destroyed between 500 to 1000 hours at 200°C; see Figure 111. Thus, it
can be concluded that a transition of the degradation mechanism for the MY721/HHPA-
system is reached after about 500 hours of thermal ageing at 200°C resulting in a rapid
decomposition reaction. Consequently, a marked reduction of the network density by
intermolecular chain scissions is consistent with a decrease in the corresponding Tg-data (cf.
Figure 102).
156 Chapter 4

Figure 111: Photograph of a MY721/HHPA-sample aged for 1000 hours at 200°C under air
atmosphere.

The MY721/MA-system exhibited different isothermal weight-loss behaviour during

continuous ageing at 200°C. The lower Ti-value from the dynamic TGA-experiments has
already provided a first indication for a decreased long-term thermo-oxidative stability.
Nevertheless, it is remarkable that the sustainable weight-loss has no direct affect on the
network structure as indicated by the analogous glass transition behaviour (cf. Figure 102)
under comparable ageing conditions. Therefore, it is proposed that the observed weight-loss
is predominantly based on the thermal degradation of dangling network chains (presumably
monoacid end groups as identified in section 2.3.4) that are not significantly affecting the
final network density. Nevertheless, such chain scissions contribute to the overall weight-
loss by the formation of volatile fragments.
The MY721/MA/HHPA-system exhibited the maximal weight-loss during the thermal ageing
at 200°C and an almost 25% mass reduction was recorded over the entire ageing period of
1000 hours. Moreover, the validity of weight-loss predictions by using a mixing rule is more
questionable, since the calculated values differ significantly and lower absolute weight-loss
values were calculated.

4.4.3.1 Effects of inorganic fillers

In this section the influence of different disperse inorganic fillers on the long-term weight-loss
behaviour was investigated. The weight-loss behaviour of the neat and filled-systems as a
function of ageing temperature and time are shown in Figure 112.
 Thermo-oxidative degradation behaviour 157

a) 2.0 b) 5
1.8
neat resin
1.6 4 silica
alumina
1.4
Weight loss [%]

neat resin

Weight loss [%]

1.2 silica 3
alumina
1.0

0.8 2

0.6

0.4 1

0.2

0.0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Time [h] Time [h]

c) 25
neat resin
silica
20
alumina
Weight loss [%]

15

10

0
0 200 400 600 800 1000
Time [h]

Figure 112: Isothermal weight-loss of the MY721/MA/HHPA-system (r=0.8) as a function of

filler type at (a) 150°C, (b) 170°C, (c) 200°C. The filler fraction of 40 vol.% was constant for
all samples. The sample thickness was 4 mm.

It can be clearly seen that substantial differences between filled and neat MY721/MA/HHPA-
systems were obtained. As expected, the filled systems exhibited a decreased weight-loss
behaviour in comparison to the neat resin. For instance, filling the MY721/MA/HHPA-system
with 40 vol.% of alumina (Al2O3) reduced the weight-loss by about 1.5% during 1000 hours
of thermal ageing at 150°C. However, even if these results seem trivial since the filled
systems contain less epoxy per volume unit than the neat systems, it is interesting to study
the effects of different fillers on the weight-loss of the resin part of the composites. In this
case, it is possible to quantify the efficiency of the filler on the thermo-oxidative stability.
Since the specimens used in this study have different filler-weight fractions, the comparative
weight-loss (CWL) as defined by equation (4.70) was used for direct comparison of the
different samples:

CWL(Ta , t a ) = m' ( t ) =
m( t )
(4.69)
1 − Vf
where Vf is the volume fraction of inorganic filler in the material, and m(t) is the measured
weight variation of the composite sample as given by eqn (4.69).
158 Chapter 4

The experimental data for silica (SiO2) and alumina (Al2O3) filled systems at different ageing
temperatures are shown in the Figure 113(a)-(c).

a) 2.0 b) 5

1.8 neat resin

silica
1.6 4 alumina

Comparative weight loss [%]

Comparative weight loss [%]

1.4 neat resin

silica
1.2 alumina 3

1.0

0.8 2

0.6

0.4 1

0.2

0.0 0
0 200 400 600 800 1000 0 200 400 600 800 1000

Time [h] Time [h]

c) 25
neat resin
silica
20 alumina
Comparative weight loss [%]

15

10

0
0 200 400 600 800 1000
Time [h]

Figure 113: Comparative weight-loss (CWL) of the MY721/MA/HHPA-system (r=0.8) as a

function of filler type at (a) 150°C, (b) 170°C and (c) 200°C. A filler fraction of 40 vol.% and a
sample thickness of 4 mm were constant for all samples.

It can be seen that the obtained differences can not only be related to the weight-loss of the
neat resin but also to the nature of fillers used in the composite systems. For example a
weight-loss of 1.78% for the neat resin was obtained after 1000 hours of exposure at 150°C.
Taking into consideration that the volume unit involves 40% of inert inorganic filler a
theoretical weight-loss of about 1.1% for the filled systems would be expected. However, the
experimental data under comparable ageing conditions show only a weight-loss of 0.8% for
the SiO2 and 0.52% for the Al2O3 filled composites. It is proposed that this observation can
be mainly attributed to the extension of the diffusion length as a consequence of adding
impenetrable particles. Hence the diffusion of oxygen into the bulk and the migration of
volatile reaction products out of the sample are significantly obstructed, as schematically
shown in Figure 114. Further, effects due to filler surface interactions or variations in the
particle size distribution cannot be excluded, but stated as minor alterations. Finally, based
on the experimental results, it can be concluded that Al2O3 is more efficient in protecting the
MY721/MA/HHPA-system against thermo-oxidative degradation than the SiO2.
 Thermo-oxidative degradation behaviour 159

a) b)
Figure 114: Schematically diffusion path for (a) neat resin and (b) particle filled systems

4.4.3.2 Modeling the isothermal weight-loss behaviour

In order to describe the degradation kinetics and the related lifetime under service
conditions, it is necessary to accurately model the weight-loss and thus to estimate the long-
term performance of the examined systems. However, to express the degradation process
in polymeric materials quantitatively, phenomenological approaches are generally adopted,
owing to the complexity of the degradation mechanism. Thus, it is assumed that the rate of
weight-loss (dm/dt) is generally given by:
dm
= − kf (m) (4.70)
dt
Where k is an overall kinetic constant that follows an Arrhenius law with respect to
temperature, and f(m) is the model equation used to describe the dependence of the
degradation rate on weight-loss. A considerable number of different models have been
postulated over the last years [13, 73, 82, 83]. Among these, two theoretical models were
applied in order to describe the different investigated anhydride/epoxy-systems. The first
model is the double-stage mechanism that has been originally proposed by Le Huy [33]. The
second one is a Fickian analogy considering diffusional processes.

4.4.3.3 The Double-Stage Mechanism (DSM)

The performed long-term weight-loss experiments allow verifying whether the experimental
data can be described by the double-stage mechanism (DSM). The DSM-approach
generally assumes two distinct occurring kinetic stages to describe the characteristic shape
of the global weight-loss curve (see Figure 115):
(I) a first stage (m1) where the rate of weight-loss (dm/dt) is comparatively high but
decreases with ageing time, describing a self-retarded step.
(II) a second degradation stage (m2) whereas the whole of the sample degrades with
a quasi-constant weight-loss rate (dm/dt=constant).
160 Chapter 4

Cumulative weight-loss
m=total measured weight-loss

m2=final slow weight-loss (k2)

m1=initial fast weight-loss (k1)

Time

Figure 115: Representation of the Double-Stage Mechanism (DSM). The curves represent
the two independent weight-loss processes (m1 and m2) utilizing different reaction rates and
the cumulative weight-loss curve (m).

If one denotes m as the weight-loss (%) of a given epoxy system, the two individual stages
(m1 and m2) can be described by using the following cumulative relation (4.71):
m = m1 + m 2 (4.71)

Assuming that in the first degradation stage the studied epoxy-systems exhibit a limited
oxygen permeability, the thermo-oxidative degradation is predominately oxygen-
concentration controlled. Under normal conditions, the mass of the degradable species on
the polymer are in excess with respect to oxygen, and hence approaching a limiting, residual
weight-loss value. Considering a first order homogeneous reaction order model, this effect is
quantified by the expression:
m1( t ) = B e (−k1t ) + m f (4.72)

where k1 is a pseudo-first order rate constant. mf denotes the residual mass of the reactive

species at time t→∞. From the initial mass that m=m0 at t=0, it follows that the pre-

exponential factor B is equal to m0-mf , i.e. the mass of the reactive species that has been
degraded by the end of the process. In terms of relative conversion (α), since α=(1-m/m0),
αf=(1-mf/m0), and m=m0(1-α), equation (4.72) can be rearranged to:
(
α = α f 1 − e −k1t ) (4.73)

Finally, recalling that α=(m0-m)/m0, the weight-loss conversion equation can be written as
eqn (4.74):
(
m1 (t ) = (m0 − mf ) 1 − e −k1t ) (4.74)

where (mo-mf) is a conversion parameter.

 Thermo-oxidative degradation behaviour 161

For the second weight-loss stage, m2(t), the corresponding weight-loss rate shifts towards
an almost constant value that can be accurately described by zero-order kinetics such that
eqn (4.75) applies:
m 2 (t ) = − k 2 t (4.75)

with k2 being the pseudo-zero order rate constant.

Consequently, the global weight-loss behaviour can be described by a combination of a first-
and a zero-order process (eqns (4.74) and (4.75)) as:
m(t ) = (m 0 − m f )[1 − exp(− k 1t )] − k 2 t (4.76)

A non-linear least square curve regression allows fitting the DMS-model to the
experimentally obtained isothermal weight-loss data, and thus the determination of the
characteristic kinetic parameters as given in Table 19.
In addition, the specific reaction rates (r1 and r2) of each individual degradation stage can be
finally determined by the derivation of the previous part reactions leading to eqn (4.77):
dm1( t ) (m0 − m f ) (m 0 − m f ) (4.77)
r1 = = + k2 = + r2
dt t→0 k1 k1
dm 2 ( t )
r2 = = k2
dt t→∞
The finally calculated reaction rates (r1 and r2) are listed in Table 19 as well.
 162 Chapter 4

Table 19: Characteristic kinetic parameters of thermo-oxidative degradation of the

MY721/MA/HHPA-system (r=0.8) obtained by best fitting to the double-stage mechanism.
Ta=150°C (m0–mf) (%) k1 (h-1) k2 (h-1) = r1 (%h-1)
r2 (%h-1)
MY721/HHPA/MA 1.0544 13.8150 0.0008 0.0771
MY721/HHPA 0.3563 0.9940 0.0003 0.3587
MY721/MA 1.1287 16.5723 0.0010 0.0691
MY721/DDS 0.6120 1.1319 0.0004 0.5410
MY721/HHPA/MA/Al2O3 (40 vol.%) 0.0369 1.0052 0.0003 0.0370
MY721/HHPA/MA/SiO2 (40 vol.%) 0.0978 1.0169 0.0004 0.0966
Ta=170°C
MY721/HHPA/MA 1.3359 15.3717 0.0020 0.0889
MY721/HHPA 0.4050 1.0490 0.0009 0.3870
MY721/MA 1.1674 15.1061 0.0004 0.0777
MY721/DDS 0.8050 1.2945 0.0010 0.6229
MY721/HHPA/MA/Al2O3 (40vol.%) 0.0764 1.0132 0.0009 0.0763
MY721/HHPA/MA/SiO2 (40vol.%) 0.1781 1.0485 0.0010 0.1708
Ta=200°C
MY721/HHPA/MA 6.0353 91.8784 0.0181 0.0838
MY721/HHPA 0.3492 1.4227 0.0113 0.2568
MY721/MA 8.2791 140.1040 0.0101 0.0691
MY721/DDS 1.3896 19.1890 0.0022 0.0746
MY721/HHPA/MA/Al2O3 (40vol.%) 0.4044 1.721 0.0049 0.2399
MY721/HHPA/MA/SiO2 (40vol.%) 0.4497 1.6631 0.0056 0.2760

A comparison between the experimental data and the theoretical Double-Stage Mechanism
for the MY721/MA/HHPA-system at different ageing temperatures demonstrates the fairly
good agreement as shown in Figure 116.
4
a) b)
2.0 MY721/HHPA/MA Ta=150°C
MY721/HHPA/MA Ta=170°C

3
1.5
Weight loss (%)

Weight loss (%)

2
1.0 Double-stage mechanism model
Experimental data Double-stage mechanism model
Experimental data
0.5 1

0.0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Ageing time (h) Ageing time (h)
 Thermo-oxidative degradation behaviour 163

c)
25
MY721/HHPA/MA Ta=200°C

20
Weight loss (%)

15

10 Double-stage mechanism model

Experimental data

0
0 200 400 600 800 1000
Ageing time (h)

Figure 116: Comparison between the experimental weight-loss data and the calculated DSM
for the MY721/MA/HHPA-system (r=0.8) at an ageing temperature of (a) 150°C, (b) 170°C,
and (c) 200°C, respectively.

Furthermore, it is assumed that the temperature dependence of the specific reaction rates
(r1 and r2) obey an Arrhenius relation as given by eqn (4.78):
 E 
ri = A i exp − ai  with i = (1,2) (4.78)
 RT 
Where Ai is the pre-exponential factor, Eai is the apparent activation energy of the thermo-
oxidative degradation process, R is the universal gas constant and T is the absolute
temperature. The subscript “i” denotes the two individual degradation stages.
According to the previous mathematical expression, the apparent activation energy (Eai) is
given by the slope of the curve ln(ri)=f(1/T) and the constant Ai is obtained from the
intersection at the origin. An example of such an Arrhenius plot for the second degradation
stage of the MY721/MA/HHPA- system is shown in Figure 117. The finally calculated Eai and
the corresponding Ai for the investigated systems are listed in Table 20.

-2.0
-2.5
-3.0
-3.5
-4.0
ln(r2) [%h ]
-1

-4.5
-5.0
-5.5
-6.0
-6.5
-7.0
-7.5
-8.0
-3 -3 -3 -3 -3 -3 -3
2.1x10 2.2x10 2.2x10 2.3x10 2.3x10 2.4x10 2.4x10
-1
Temperature [K ]

Figure 117: Arrhenius plot for the second degradation stage of the neat MY721/MA/HHPA-
system (r=0.8). Calculation of the activation energy (Ea2) and the reaction rate (A2) were
obtained from the linear regression.
164 Chapter 4

Table 20: Characteristic kinetic data for thermo-oxidative degradation of different MY721
based epoxy system as experimentally obtained form determined weight-loss data (4 mm
thick samples) by applying the Double-Stage Mechanism [33].

This table is not available on line. Please consult the hardcopy thesis
available from the QUT Library.

It should be noted that some of the calculated Arrhenius parameters for the first degradation
stage (Ea1) in Table 20 must be treated with caution, because the experimental values did
not result in an Arrhenius plot with a high correlation coefficient (R<0.9). Nevertheless, the
comparison of Ea-values for the anhydride-cured systems shows that the MY721/MA-
network exhibits the lowest apparent activation energy. The MY721/DDS-system revealed
the lowest activation energy for the second degradation stage (Ea2) and consequently the
lowest overall activation energy (Ea).
Besides, the data provide evidence that the microstructure of the MY721/MA/HHPA-system
is a combination of an independent MY721/MA and MY721/HHPA sub-network. Hence, a
simple mixing rule that considers the molar anhydride ratio (MA/HHPA=0.7/0.3) was applied
to predict the apparent activation energy (Ea) of the co-anhydride cured system in a
reasonable manner (cf. eqn (4.79)).

E MY
a
721/ HHPA / MA
= 57.7 kJ / mol (4.79)

0.3 * E MY
a
721/ HHPA
+ 0.7 * E MY
a
721/ MA
=

kJ kJ kJ
0.3 * 64 + 0.7 * 46 = 56.4 ≅ E MY
a
721/ HHPA / MA
mol mol mol

In addition, the obtained DSM-results allow a correlation to the involved degradation

mechanisms. It can be summarized that the first degradation stage can occur in only a minor
proportion of the virgin material. Consequently, the fast degradation stage has both a lower
activation energy and a lower A-factor than the slow (second) degradation process. In
practical terms, it is proposed that the fast degradation process is associated with a
sequential loss of unstable moieties such as structural irregularities (presumably acidic
dangling chain ends, cf. section 2.3.4) that are initially present. Since these volatile
fragments gradually reduced during thermo-oxidative degradation, the experimentally
 Thermo-oxidative degradation behaviour 165

observed weight-loss rate decreased as a function of ageing time. Moreover, the higher (m0-
mf)-values, see Table 19, for MA-cured systems, give evidence of a higher initial amount of
‘irregular’ structures at the beginning of the degradation process. Consequently, an
increased weight-loss rate was observed for MA-cured samples during the initial
degradation stage. In addition, the predominant decomposition of dangling chains within the
first degradation stage gives a reasonable explanation, while the network density as
expressed by the related Tg remains almost unaffected (cf. Figure 102).
In contrast, the whole of the sample is susceptible to degradation by the second, slow
degradation process that is attributed to random chain scissions of more stable backbone
structures. Therefore, it has a higher activation energy and a higher pre-exponential factor
than the fast process.
A schematic presentation of the proposed network changes is given in Figure 118. Even if
the hypothesis has not been experimentally confirmed for the studied epoxy/anhydride
networks, it is proposed that a general indication of the two-stage degradation mechanism is
valid.
a)
Q Q Q Q Q Q

Q • Q
Q • Q

Q Q Q Q
Q Q
Q Q

Q Q
• •
• • •
Q Q

• •
Q Q
Q Q

Q Q
•
Q Q

b) Q Q Q Q Q Q

Q • Q
Q • Q

Q
Q Q
Q • • Q Q

••• •
Q Q

Q Q Q Q

• •
Q Q

Q Q

Q Q
Q Q

Figure 118: Schematization of a chain scission processes yielding to volatile fragments

occurring (a) during the first degradation stage in dangling chains, and (b) during the second
degradation stage in more stable backbone structures.

4.4.3.4 The Fickian diffusion model

The thermo-oxidative degradation generates volatile degradation products that evaporate
according to a diffusion-controlled process that constitute the rate-determining stage.
Therefore, the specific isothermal weight-loss behaviour of the investigated glassy epoxy
systems has been frequently analyzed by analogy to diffusional processes as originally
166 Chapter 4

introduced by Fick’s second law of diffusion [84]. This theoretical description based on a
differential mass balance, which assumes that the time rate of concentration change is
related to the second derivative of the concentration gradient through the diffusion
coefficient:
dC x ∂  ∂C x  ∂ 2C x (4.80)
=−  − D  = −D ⋅
dt ∂x  ∂x  ∂x 2

where D is the diffusion coefficient, x is the space coordinate measured normal the sample
thickness, and t denotes the time. If the material has a uniform initial diffusant concentration
(Co) and the surface is kept at a constant concentration (Cmax) the solution of eqn (4.80) is:

C − C0 ∞
( −1)n  D(2n + 1)2 π 2 t 
= 1−
4
∑ exp −  cos (2n + 1)πx (4.81)
Cmax − C 0  
π n=0 (2n + 1)
2
 h  h
The total amount of fragments diffusing out of the polymeric network (m) as a function of
time is given by the integral of eqn (4.81) across the thickness (h):
∞  π 2 (2n + 1)2 
∑ (2n + 1) exp −
m 8 1 Dt (4.82)
= 1− 2
mmax π h 2 
n =0  
where mmax is the maximum quantity of the diffusing substance at infinite time. A more
simplified form of (4.82) for values of m/mmax lower than 0.6 can be given by:
m 4
= Dt (4.83)
mmax h t
Consequently, if the weight-loss obeys Fickian behaviour, a linear plot of m against t0.5 in its
initial part can be considered as a validation criterion of Fick’s law. From the slope of the
linear part, the coefficient of diffusion can be calculated. The rearranged weight-loss curves
for the MY721/MA/HHPA-system and the linear regression curves are shown in Figure 119.
a) b)
4
2.0
MY721/HHPA/MA Ta=170°C
MY721/HHPA/MA Ta=150°C
3
1.5
Weight loss (%)
Weight loss (%)

2
1.0

1
0.5 1/2
p(t)=0.10884.tR=0.98
1/2
p(t)=0.06694.t R=0.92

0
0.0 0 5 10 15 20 25 30 35
0 5 10 15 20 25 30 35
1/2
1/2 t
t
 Thermo-oxidative degradation behaviour 167

c)
25
MY721/HHPA/MATa=200°C

20
Weight loss (%)

15

10

1/2
5 p(t)=0.71173.tR=0.997

0
0 5 10 15 20 25 30 35
1/2
t
Figure 119: Comparison between the experimental weight-loss data (solid line) and the
theoretical Fickian model (dotted line) for the MY721/MA/HHPA-system (r=0.8) at an ageing
temperature of (a) 150°C, (b) 170°C, and (c) 200°C, respectively.

From the gravimetric analysis it can be seen that with increasing ageing temperature the
weight-loss curves can be more precisely described by the Fickian dependency. Thus, the
more accurate agreement with the Fickian diffusion behaviour, even in the initial stage, was
obtained for higher ageing temperatures. This observation is consistent with the fact that
diffusion processes are generally favoured by elevated temperature.
Considering further that the diffusion coefficient (D) obeys an Arrhenius dependency as
given by eqn (4.84):
 E 
D = D0 . exp − a  (4.84)
 RT 
and rearranging eqn (4.84) by taking the logarithm leads to eqn (4.85):
E  1
ln(D) = − a  ⋅ + ln D0 ( ) (4.85)
R T
Thus the global activation energy (Ea) and the appropriate rate coefficient (D0) can be
obtained by plotting the ln(D0) versus 1/T for the three studied ageing temperatures. The
obtained kinetic degradation data are summarized in Table 21.

Table 21: Characteristic kinetic data for thermo-oxidative degradation as calculated by the
Fickian method [84] for different MY721-systems (r=0.8)
1/2
Systems Ea (kJ/mol) D0 (%/h ) R2
MY721/HHPA/MA 80 4.6*108 0.94
8
MY721/HHPA 84 4.5*10 0.95
7
MY721/MA 69 2.1*10 0.99
3
MY721/DDS 39 2.5*10 0.99
9
MY721/ HHPA/MA/Al2O3 (70%wt) 92 2.1*10 0.99
8
MY721/ HHPA/MA/SiO2 (60%wt) 80 1.0*10 0.99

Even if the Fickian model used is based on different theoretical assumptions than the
Double-Stage Mechanism concept, the calculated kinetic parameters allow the prediction of
168 Chapter 4

the same stability range in terms weight-loss. Thus, a consistent sequence was obtained for
the Fickian- and the DSM-model. Nonetheless, the absolute values of the activation energy
(Ea) slightly differ between the two applied models and generally higher values were
obtained for the Fickian model. Moreover, a common behaviour of both models is that the
quality of the results increased with increasing temperature, as it can be seen in Figure 116
and Figure 119, respectively. Finally, it can be stated that diffusional processes must be
unambiguously included when describing the experimental degradation data.

4.4.4 Thermo-oxidative degradation studied by FT-IR spectroscopy

Numerous analytical techniques, such as gas chromatography [85], mass spectroscopy [86],
radiochemical analysis [85, 86], or combinations of these methods have been used to study
the degradation mechanisms of polymeric materials. However, a limitation of these
techniques arises from the fact that a relatively high degradation temperature is required to
produce sufficient amounts of volatile products. Moreover, the degradation products itself
may not be stable and react in a second step to more complex structures. Thus, the product
analysis may not always be unambiguous to interpret mechanistically and consequently may
result in incorrect conclusions. These difficulties can be overcome by FT-IR spectroscopy
since this technique examines the behaviour of the polymeric substrate rather than the
volatile products formed. The high sensitivity of FT-IR spectroscopy additionally allows an
approximate localization and a quantitative determination of chemical species in particular
when the chemical degradation reactions are relatively slow [87, 88].
In the present section the thermo-oxidative ageing behaviour of the different anhydride-
cured MY721-systems was studied by FT-IR micro ATR-spectroscopy. The interpretation of
the recorded spectra and the band assignments are adapted from the information related to
the pure components. The used band assignments rely predominately on the
comprehensive work of Antoon and Koenig that was based on the characteristic group
frequency approach [89, 90]. Further interpretation of the spectra is based on the work of
Bellamy [91], Colthup et al. [92], and Pretsch et al. [93], as well as the work presented in
Chapter 2 within this thesis.
During the isothermal ageing at 150°C the characteristic spectra significantly changed with
increasing ageing time, and hence the spectral features allow an extraction of the relevant
molecular information. At higher ageing temperatures the spectra rapidly became
featureless and consequently a meaningful interpretation was not possible. The
representative time-resolved IR-spectra of the different anhydride/epoxy-systems are shown
in Figure 120. The infrared analysis of the crosslinked systems revealed that a variety of
molecular changes occurred during thermo-oxidative ageing. Almost all peaks exhibited
 Thermo-oxidative degradation behaviour 169

intensity changes or peak-shifts when compared to the initial spectra. However, with
exception of the fingerprint region, the general changes in relative intensities for all studied
systems were similar and independent on the anhydride curing agents that were used.
For all investigated samples, an intensity increase of the broad band between 3100 and
3600 cm-1 was clearly visible with progressing ageing. Generally, this broad band is related
to [νO-H] or [νN-H] if both are present, however, an assignment to hydroxyl seems in this
case more reasonable. Nonetheless, it cannot be further differentiated between hydroxyl
due to water that is dissociated and bound in the network or to water as a reaction product.
The observed shift of the [νO-H] peak maxima at ∼3500 cm-1 to a low frequency component
centered at ∼3360 cm-1 indicates the formation of new species absorbing in the lower side of
the O-H range. These oxygenated species might be tentatively assigned to hydroperoxide
groups and carboxylic acids [91, 94]. However, for quantitative considerations the spectra in
this area were disregarded, since all degraded samples provided a significant baseline shift
in this region, due to an intense discoloration of the severely degraded material.
An analysis of the data in the area between 800 and 2000 cm-1 revealed that the asymmetric
carbonyl stretch of cyclic anhydrides [νasC=O] at 1798 cm-1 in the HHPA-cured, and at 1780
cm-1 in the MA/HHPA-cured system, respectively, decreased with increasing exposure time.
This observation is related to a further consumption or vaporization of partially unreacted
HHPA. On the other hand for purely MA-cured system no vibrational band was observed
that confirms the presence of unreacted maleic anhydride.
A peak at ∼1775 cm-1 (indicated by arrows in Figure 120) gradually increased with ageing.
The peak position is in agreement with earlier studies by Le Huy et al. who investigated the
thermal degradation behaviour of differently flexibilized anhydride-cured epoxies [33-35]. Le
Huy et al. assigned the peak to [νas C=O] of reformed anhydride species. However, since the
characteristic symmetric carbonyl stretch of cyclic anhydrides at ∼1850 cm-1, [νsym C=O] was
not detectable, the 1775 cm-1 band assignment to reformed anhydride species is highly
questionable. Even if the peak assignment at 1775 cm-1 is not unambiguous, it is here
proposed that this intensity is tentatively assigned to the carbonyl stretch of highly
oxygenated species, such as lactones [95].
The distinct carbonyl stretch of the aromatic ester at ∼1730 cm-1 exhibited a gradual
decrease and shifts towards lower frequencies at 1710 cm-1 (indicated by the dotted line in
Figure 120). This behaviour is probably related to an extensive oxidation of the ester groups
and partial transformation into new aldehyde and ketone species. Moreover, the broadened
peak appearance is likely related to the multiplicity of different carbonyl species formed
and/or due to the occurrence of molecular interactions, e.g. hydrogen bonding, which
generally decreases the frequency of the carbonyl stretching absorption. In addition,
170 Chapter 4

potential C-O and C-N chain scissions yield to an increase of the polar O-H and N-H groups
that generally favor hydrogen bonding.
A new absorption band near 1670 cm-1, characteristic for enamine groups, increased in
intensity during ageing due to the attack of non-stable C-N bonds by oxygen [94, 96-98].
Although polyaromatic carbonaceous compounds that are formed in the TGDDM-resin at
high temperature, could also have an IR absorption near 1670 cm-1, this potential
contribution can be neglected, because of the relatively low ageing temperature and short
ageing time used in this study [79]. Additionally, the C=C indication in MA-cured systems at
1640 cm-1, which can be assigned to monoacids of partially reacted MA, decreased with
progressive ageing.
After longer exposure the bonds between nitrogen and phenyl groups may rupture. The
observed decrease in peak intensities at ∼1165 cm-1 for HHPA-, and at ∼1158 cm-1 in the
case of MA-containing networks, respectively, can be attributed to the cleavage of C-N
bonds [99]. Further evidence of C-N cleavage is given by the change of the [ν Φ-N] peak at
about 1380 cm-1 [28, 83, 100].
Indication of a cyclization process to heterocyclic compounds is provided by the changing
substitution pattern of the aromatic TGDDM-structure. The intensity shift of the aromatic
TGDDM-structure at 1615 cm-1 [quadrant, Φ], and 1515 cm-1 [semicircle stretching, Φ] (see
dotted line in Figure 120) towards lower wavenumbers can be attributed to a cyclization to
quinoline-structures [101, 102]. This observation is in agreement with previously performed
degradation studies of TGDDM based networks, where a backbiting reaction to 1,2,3,4-
tetrahedra-3-hydroxyquiniline structures was postulated [35, 102]. Additional support for this
hypothesis is given by the increased intensities above 3000 cm-1 that correspond to the
formation of the hydroquinoline-compounds. However, this contribution could not be further
differentiated from the hydroxyl and amine absorbancies in the same region. With higher
exposure times a significant loss of the aromatic character is visible by a decrease of the
corresponding intensities.
Evidence for oxidation of the methylene groups linking the two-phenyl groups in the
TGDDM-resin (cf. structure in Table 2) is given by the decrease of the characteristic in-plane
deformation of the methylene band [δCH2] at ∼1450 cm-1 (see Figure 120) [103]. In addition,
a progressive decrease of the stretching vibration of the CH2-deformation at 2970 and 2850
cm-1 confirmed the degradation of aliphatic CH2-groups.
Additionally, a gradual growth of a new absorption at ∼830 cm-1 is noted. The hydro-peroxide
group is known to exhibit a characteristic peak at ∼830 cm-1, corresponding to the –O-O-
stretching mode [91, 93]. Hence, the observed peak is associated with a steady increase of
the R-O-O-H concentration, which is consistent with the increase of the OH stretching range.
 Thermo-oxidative degradation behaviour 171

This observation is in agreement with previously performed thermo-oxidative studies where

evidence for a free radical formation was postulated [106].
At wavenumbers below 800 cm-1 a meaningful peak analysis was not possible and therefore
no discussion about this part of the spectra was performed.

I) I)
1730 1615
3500

d) 1775

d)

c)
Absorbance [a.u.]

Absorbance [a.u.]
c)

b)
b)

1515

1640 1380
1450
a)
a)

2000 1800 1600 1400 1200 1000

3600 3400 3200 3000 2800 2600
-1
-1 Wavenumber [cm ]
Wavenumber [cm ]

II) II)
3500 1730
1615

1775
d)
d)
Absorbance [a.u.]
Absorbance [a.u.]

c)

c)

b)
b)

1515 1380

1790 1450
a)

a)

3600 3400 3200 3000 2800 2600 2000 1800 1600 1400 1200 1000
-1 -1
Wavenumber [cm ] Wavenumber [cm ]
 172 Chapter 4

III) III) 1730

1615
3460

d)

d)

c)

Absorbance [a.u.]
Absorbance [a.u.]

c)

b)
b)

1640 1515
a)
1380
1780 1450
a)

3600 3400 3200 3000 2800 2600 2000 1800 1600 1400 1200 1000
-1 -1
Wavenumber [cm ] Wavenumber [cm ]

Figure 120: Selected micro-ATR spectra from one surface area of different anhydride-cured
MY721-systems: (I) MY721/MA, (II) MY721/HHPA and (III) MY721/MA/HHPA; as a function
of ageing time (a) initial, (b) 90 hrs, (c) 1000 hrs, (d) 7420 hrs at 150°C.

A summary of the characteristic spectral changes and the related interpretations of the
investigated anhydride-cured epoxy systems is given in Table 22.

Table 22: Overview of changing FT-IR intensities during thermo-oxidative ageing of

anhydride-cured epoxy systems
Band Assignment Response on thermo- Interpretation
-1
[cm ] oxidative degradation
3100- [νO-H] gradual increase and formation of new hydroxyl or acid
3600 shift lower wavenumbers groups
1798 [νasC=O] HHPA decrease during early crosslinking reaction of HHPA
exposure in HHPA-cured
systems
1775 [νasC=O] of lactones gradual increase formation of new oxidized lactone
structures
1730 [ν(C=O) aromatic gradual decrease and reduction of aromatic ester groups
ester] shift to ∼1710 cm -1

1670 [νasC=O] of amine gradual increase formation of new amine groups

1650 [νasC=O] of ketone gradual increase formation of new aldehyde or ketonic
or aldehyde groups groups
1640 [νC=C-COOH] gradual decrease in MA- loss of conjugated acids (e.g.
cured systems monoacids)
 Thermo-oxidative degradation behaviour 173

1615 [quadrant, Φ] gradual decrease and cyclization to quinoline-structures

shift to ∼1600 cm -1

1515 [semicircle gradual decrease and cyclization to quinoline-structures

stretching, Φ] shift to ∼1500 cm -1

∼1450 [δCH2] gradual decrease degradation/oxidation of methylene

groups
1380 [νΦ-N] gradual decrease cleveage of C-N bonds
∼830 [νO-O] gradual increase formation of hydro-peroxides
species

4.4.4.1 Thermo-oxidation profiles

As already discussed in the previous sections, thermo-oxidative ageing of polymeric
materials often starts by oxidation that is kinetically controlled by oxygen diffusion. When
such a kinetic approach exists the whole oxidation process becomes a decreasing function
of the initiation rate as a function of temperature. Hence, the depth-distribution of the
oxidation products becomes heterogeneous and might be limited to superficial layers [104,
105]. In this case an oxygen-concentration maximum can be observed in the surface zone
that decreases continuously over the sample depth. Numerous kinetic models for diffusion-
controlled oxidation have been proposed that principally allow predicting the thickness of the
oxidized layer (TOL) [34, 106, 107]. However, the main disadvantage of these models is that
they mainly consider the oxidation process to be independent on oxygen diffusivity.
Moreover, oxidation is a radical chain process and there is yet no consensus on the kinetic
scheme [106]. Consequently, many authors have reduced the reaction mechanism to a
standard mechanistic scheme for radical chain processes that are initiated by thermolysis or
due to hydro-peroxide decomposition [106-108].
In this section the spectroscopic micro-ATR technique was used to determine the spatial
oxidation profile of aged samples by line mapping the cross section in selected intervals.
Figure 121 shows the measured spectra for the differently cured MY721-systems aged at
170°C. In the obtained spectra the appearance of lactone species, related to the IR-intensity
at ∼1775 cm-1 [ν C=O] (indicated by the dotted line in Figure 121) was monitored as a
function of ageing time. The lactone formation is accompanied by the shift of the aromatic
peak at 1615 cm-1 [quadrant, Φ] as indicated by an arrow in Figure 121. It is proposed that
the formation of these highly oxidized superficial layers occurred by grafting oxygen to the
macromolecular network and some chain breaking reactions with the diffusion of volatile
compounds out of the sample. The elimination of the volatile by-products promotes the
experimentally observed weight-loss (cf. section 4.4.2.).
174 Chapter 4

e)
e)
Absorbance [a.u.]

Absorbance [a.u.]
d)
d)

c)
c)

b)
b)

a)
a)

3800 3600 3400 3200 3000 2800 2600 2000 1800 1600 1400 1200 1000
-1 -1
Wavenumber [cm ] Wavenumber [cm ]

Figure 121: Selected micro-ATR spectra from a line map across a section of the
MY721/MA/HHPA-system (r=0.8) aged at 170°C for 7420 hrs. Distance from one surface:
(a) 200, (b) 400, (c) 600, (d) 800, and (e) 1000 micron.

Oxidation profiles, for visualizing the depth distribution in which the material degrade, were
constructed. The determination of the depth distribution of oxidation products over the
sample cross section was experimentally established by taking the deconvoluted peak
height at ∼1775 cm-1 in ratio to the internal standard at ∼1615 cm-1. The latter band was
assigned to the quadrant-stretching mode of the aromatic groups and it was not significantly
affected by thermo-oxidative degradation in the investigated time frame. Consequently, the
reduced local oxygen concentration, (C/Cs), represented by the lactone intensity was
calculated according to eqn (4.86):

C  I1775 I1775 
 (4.86)
= 1615t 0
/ 1615
C s  I I 

where Io, It, are the relative intensities of reactive species at time zero, and after a certain
time interval, t, of the oxidation process. C represents the appropriate lactone concentration
as a function of time, and Cs denotes the maximum lactone concentration on the sample
surface at infinite time.
The experimentally determined oxidation profiles as a function of ageing time and
temperature under atmospheric pressure are given in Figure 122(a)-(c). It can be seen that
the oxidation is heterogeneously distributed within the sample thickness and a pseudo-
hyperbolic lactone/oxygen concentration was obtained.
 Thermo-oxidative degradation behaviour 175

After about 400 hours of exposure, which approximately reflects the end of the first
degradation stage obtained from the isothermal weight-loss curve (cf. Figure 111), an
oxidized layer of about 200-400 µm thickness was formed. However, during the second
degradation stage the TOL increased considerably and reached values of ∼1400 µm after
7420 hours of thermo-oxidative exposure. The obtained shape of the TOL-profile changed
during exposure indicating that the apparent oxidation rate is a decreasing function of the
sample thickness. Additionally, it was apparent that the TOL is a decreasing function of
temperature (see Figure 122 (d)). The TOL, and consequently the oxidation susceptibility
can be finally ranked in the following order: MY721/HHPA>MY721/MA/HHPA≥MY721/MA.
The ranking of the oxidation layer thickness, which probably acts as an intrinsic diffusion
barrier layer and limits the loss of volatile products, is consistent with the experimentally
obtained weight-loss data in the same temperature range. Moreover, the results provide
evidence that oxidation is the most important degradation mechanism during thermo-
oxidative ageing.

a) b)

1.0 90hrs 1.0

90hrs
1000hrs
1000hrs
7420hrs
0.8 0.8 7420hrs

0.6 0.6
C/Cs
C/Cs

0.4 0.4

0.2 0.2

0.0 0.0
0 500 1000 1500 2000 0 500 1000 1500 2000

Sample depth [microns] Sample depth [microns]

c) d)
1.0 12
90hrs
1000hrs 8
7420hrs
0.8
4
TOL [microns *10 ]
2

T=170°C
0.6 0
C/Cs

0.4 12

8
0.2
4
T=150°C
0.0 0
0 500 1000 1500 2000
Sample depth [microns] 0 1000 2000 3000 4000 5000 6000 7000 8000
Time [h]

Figure 122: Oxidation profile as determined by the normalized lactone concentration (C/Cs)
as a function of curing agent and ageing interval at 170°C; (a) MY721/MA, (b)
MY721/HHPA, (c) MY721/MA/HHPA; (d) thickness of the oxidized layer (TOL) of 4 mm thick
samples as a function of the ageing time and ageing temperature for () MY721/MA, (0)
MY721/HHPA, and () MY721/MA/HHPA
176 Chapter 4

4.4.5 Thermo-oxidative degradation mechanism

All investigated anhydride/epoxy samples revealed no significant differences in the thermo-

oxidative degradation mechanisms. Even if the thermo-oxidative degradation of the
anhydride-cured TGDDM-networks studied here is very complex and involves several
competitive mechanisms, the obtained changes are mainly related to the formation of highly
oxidized and carbonized network structures involving a large content of aromatic structures.
During degradation the networks are below their glass transition temperature and
consequently in the highly crosslinked structure the molecular movement of any part of the
network relative to another is severely hindered. Despite this, the thermo-mechanical results
of the ageing studies revealed that the networks predominantly degrade by the scission of
chemical network bonds. Hence, the proposed reactions outlined below illustrate
predominantly routes that are basically limited to the consideration of chemical bond
scissions and to how such reactions lead to the formation of degradation products that have
been identified by FT-IR analysis within this study (cf. section 4.4.4). The exceptional
reaction is the dehydration of the network.
In terms of bond dissociation energies, the weakest points in the crosslinked network are the
C-N and C-O bonds [26]. However, the relatively high concentration of hydroxyl groups in
the cured resin favors localized hydrogen bonding, which consequently promotes initial
dehydration. Hence, it is proposed that the thermo-oxidative degradation initiates by the
dehydration in agreement with previously reported degradation studies [20, 29, 79, 109,
110]. Moreover, available information emphasize that dehydration precedes, to a large
extent, any further degradation of the network [30, 111]. The formation of water from ether
oxygen of the network is not reasonable because of the relatively high stability of the aryl-
oxygen bond [27]. Consequently, it is more likely that the dehydration of the secondary
alcohols, as schematically shown in Figure 123, is primarily involved in this initial
degradation reaction [27, 29-31]. The driving force behind these reactions is the large
increase in stability of the systems, which follows double bond formation. This is particularly
true when the formed double bonds are pseudo-conjugated by a nitrogen atom in
conjunction with an aromatic ring [26, 27]. It is postulated that degradation results in the
formation of the two unsaturated structures (I) and (II) as depicted in Figure 123. Structure
(II) is probably the more favored moiety, because oxygen has a greater negative inductive
effect than nitrogen so a hydrogen atom is more easily withdrawn from the CH2 group
adjacent to the oxygen atom. However, increasing IR-absorption in the region at
1650 to 1690 cm-1 suggests that the enamine structure (structure I) is also formed to a
certain extent. In addition, the formation of a highly conjugated structure would be consistent
with the observed discoloration of the degraded epoxy-samples. Besides, the formed
 Thermo-oxidative degradation behaviour 177

conjugation induces stiffness into the polymer chains, which can be finally attributed to
observed embrittlement and the increase in the glass transition temperature during the early
ageing period (cf. section 4.4.1).

C N C N
H2 H
C C C O H2
H2 H2 C C C O
H H H2
OH
-H2O (I)

C N
H2
C C C O
H2 H H

(II)
Figure 123: Initial dehydration mechanism for thermo-oxidative ageing of the investigated
MY721/MA/HHPA-system.

The allylic hydrogen position in structure (I) and (II) weakens the network structure [31] and
hence the C-N and C-O bonds are likely to undergo further chain scissions as it
accumulates in the first, low temperature decomposition stage.
The enamine structure (I) would favor scissions of the allyl-oxygen bonds followed by
cyclization to quinoline type structures (III) as shown in Figure 124. The shift of the
intensities at ∼1615 and ∼1515 cm-1 indicates the change of the aromatic substitution pattern
that can be assigned to cyclization of quinoline-derivatives. The simultaneously generated
oxygen radical (IV) would be stabilized by hydrogen abstraction yielding new hydroxylated
species (V). This reaction accounts for the experimentally observed increase of the related
hydroxyl intensities.
In the enamine (structure II) the allyl-nitrogen chain cleavage occurs preferentially at the
position indicated in Figure 125. Evidence for the C-N cleavage was obtained by the
decreasing intensity of the corresponding [νΦ-N] peak at 1380 cm-1. The obtained radicals
are stabilized by hydrogen abstraction yielding primary amines (VI) that could consequently
be further converted to secondary amine structures (VII). The experimentally identified C-O
and C-N bond scissions mainly caused the instantaneous cleavage of the molecular network
structure resulting in a significant decrease of the glass transition temperature, as
particularly pronounced for HHPA-cured samples (cf. Figure 102).
In addition, the weak allylic hydrogen in structure (I) and (II) can be easily abstracted by
radical species, which are present in trace amounts in the epoxy material (e.g. formed in the
high temperature curing or even directly generated by O2 attack). For instance the proposed
auto-oxidation sequence leads to hydrogen abstraction of the methylene groups between
the aromatic TGDDM-units as shown by structure (VIII) in Figure 124. Finally, stabilized
178 Chapter 4

radicals may be formed as shown by structure (IX). This reaction would finally account for
the gradually absorption decrease of the methylene groups [δCH2] at
1450 cm-1.

NH2

(VII)

+H
.
+ CH3 CH CH O

N
H

(VI)

+H
.

C N (II)
H2
CH2 CH CH O

C N
H2 (I)
CH CH CH2 O
-H
.

-H
.

N
C
H
N
CH + .O (IV)
CH2 CH CH O
(III) C C
(VIII) H2 H
.
+H

HO (V)

C C N
H C C C O
(IX) H H H2

Figure 124: Proposed degradation sequences for thermo-oxidative ageing of the

investigated anhydride-cured epoxy-system.

Besides, if bond cleavage occurs before dehydration Patterson-Jones [30] postulated that a
radical formed by scissions of the C-N undergoes rearrangement followed by elimination to
form aldehyde species as shown in Figure 125. In any case the reaction yields to the
formation of new aldehyde species (structure X), which was experimentally confirmed by the
shift of the related carbonyl intensities to 1650 cm-1.
 Thermo-oxidative degradation behaviour 179

C N
H2 H
C C C O
H2 H2
OH

O
C N + + . CH2 O
H2 . CH3
H
(X)
aldehyde
. +H
.
+H

C N H3C O
H2
H
Figure 125: Proposed degradation sequences for thermo-oxidative degradation of the
investigated anhydride-cured epoxy-systems.

4.5 Concluding remarks

In this chapter the thermo-oxidative ageing behaviour of different anhydride-cured amino-

glycidyl resins was investigated. The effects of long-term ageing on thermo-mechanical
properties, thermo-gravimetric behaviour, and architectural network changes were assessed
and discussed.
The behaviour of the viscoelastic properties during thermo-oxidative ageing was studied by
means of low frequency dynamic mechanical spectroscopy. All TGDDM-based formulations
led to highly crosslinked networks (Tg>200°C). However, depending on the type of hardener,
a different dynamic mechanical behaviour was observed. It was demonstrated that for
maleic anhydride-cured samples the appearance of a secondary ω-relaxation in the
intermediate temperature range between 80 to 150°C is associated with network-
irregularities (e.g. dangling chain ends). The relaxational behaviour was described as a
function of ageing time by means of Cole-Cole plots. The gradual decrease of the ω-
relaxation with increasing ageing time can be seen in the formation of a mono-relaxational
Cole-Cole plot appearance. It is proposed that this effect can be directly related to the
volatilization and loss of primarily dangling chain fragments. Moreover, it explains the fact
that the glass transition temperature, and thus the associated network-density, of the purely
MA-cured network remains almost unaffected during 1000 hours of ageing. On the other
hand hexahydrophthalic anhydride-cured systems revealed a combination of two
independent ageing mechanisms. During the early ageing period the viscoelastic behaviour
clearly indicates a predominant post-curing process. However, with progressing ageing time,
an irreversible degradation mechanism was superimposed leading to a deleterious alteration
180 Chapter 4

of the thermo-mechanical property profile and the initial network structure. Based on the
experimental results it is emphasized that the application of the physical Perez model is
meaningful to interpret the viscoelastic property changes in terms of structural network
variations.
Thermo-gravimetric analysis of neat and particulate-filled materials revealed a double stage
degradation mechanism under air atmosphere. It was demonstrated that the low
temperature degradation stage is related to thermolysis and hence independent of the
gaseous environment. The second, high temperature degradation stage is primarily
attributed to oxidation processes, since it was not detected under nitrogen atmosphere. To
determine kinetic parameters experimental long-term weight-loss data at 150, 170, and
200°C were modeled by a double stage mechanism and a diffusion related analogy. The
extracted kinetic degradation parameters permit to outline a more generalized two-stage
degradation mechanism. The first degradation stage predominantly reflects the degradation
of structural irregularities, such as untangled chains or end-groups of dangling chains. Thus,
this process has only a secondary influence on the final network density. Even through the
accompanied weight-loss (e.g. for maleic anhydride cured systems) was relatively high an
almost constant glass transition temperature was observed. The second degradation stage
then includes the chemical degradation of the more stable backbone structure, and hence
significant network changes. However, differences in the thermal stability ranking were
observed at higher ageing temperatures. In particular the MY721/HHPA-network was
drastically perturbed when ageing was carried out at 200°C. The obtained weight-loss
considerably increased after 450 hours reflecting significant chemical modifications of the
network structure, which is consistent with the marked decrease in the related glass
transition temperature.
Mechanical endurance tests were performed in order to predict the lifetime under service
conditions. It was highlighted by the calculated temperature index that even though MA-
cured networks are more susceptible to the first degradation step, the entire thermo-
oxidative stability is superior to HHPA and DDS-cured networks. Therefore, the overall
thermo-oxidative stability sequence can be ranked as: MY721/MA>MY721/MA/HHPA
≥MY721/DDS>MY721/HHPA. Evidence that the introduction of inorganic disperse silica
(SiO2) and alumina (Al2O3) particles improved the thermo-oxidative stability and weight-loss
rate, was deduced from the experimental measurements of hybrid systems. It was
demonstrated that Al2O3-filler outperformed SiO2-particles.
The obtained initial properties as well as the thermal stability could be estimated, with a fairly
good accuracy, by a simple mixing rule and ranked between the two individual anhydride-
cured systems. Hence, the experimental results demonstrated that the studied co-
 Thermo-oxidative degradation behaviour 181

anhydride-cured system involves at least an MY721/MA and MY721/HHPA sub-network as

schematic shown in Figure 126.

Figure 126: Proposed network structure for the MY721/MA/HHPA-system. The solid lines
represent the MA-segments and the dotted lines the HHPA-segments.

The degradation mechanism and the corresponding oxidation profile were studied by FT-IR
micro-ATR spectroscopy. The spectral information provided strong evidence that an initial
dehydration-process is mainly followed by aliphatic C-N and C-O chain scissions that
triggered more complex degradation processes. Chemical modification of the studied
materials induced by thermo-oxidative ageing can be clearly seen by the formation of highly
oxidized carbonyl structures (e.g. lactones, ketones and aldehydes) in combination with a
cyclization involving the aromatic backbone structure. We have not seen any evidence of an
alternative mechanistic degradation pathway to be caused by the presence of different
anhydride types or variations in the stoichiometry of the studied networks.

4.6 References

[1] Bryk M.T., Babutkina K.A., Kemp T.J., Degradation of Filled Polymers, Ellis Horward,
West Sussex, (1991).
[2] Lee H., Neville K., Handbook of Epoxy Resins, McGraw-Hill, New York, (1967).
[3] Wright W.W., Lee W.A., Progress in Polymers, CRC Press, Cleveland, (1968).
[4] Struik L.C.E., Physical Ageing in Amorphous Polymers and other Materials, Elsevier,
Amsterdam, (1978).
[5] Kong E.S., Adv. Polym. Sci., 80, 125 (1986).
[6] Choy I.C., Plazek D.J., J. Polym. Sci. Part B, 24, 1303 (1986).
[7] Shmorhun M., Jamieson A.M., Simha R., Polymer, 31, 812 (1990).
[8] Lin Y.G., Sautereau H., Pascault J.P., J. Appl. Polym. Sci, 32, 4595 (1986).
[9] Cook W.D., Mehrabi M., Edward G.H., Polymer, 40, 1209 (1999).
[10] Maddox S.L., Gillham J.K., J. Appl. Polym. Sci., 64, 55 (1997).
[11] Ophir Z., Emerson J.A., Wilkens G.L., J. Appl. Phys., 49, 5032 (1978).
[12] Lee A., McKenna G.B., Polymer, 31, 423 (1990).
182 Chapter 4

[13] Barral L., Cano J., Lopez J., Lopez-Bueno I., Nogueira P., Ramirez C., Torres A.,
Abad M.J., J. Therm. Anal., 56, 1025 (1999).
[14] Xiao G.Z., Shanahan M.E.R., J. Appl. Polym. Sci, 69, 363 (1997).
[15] Chang T.D., Brittain J.O., Polym. Eng. Sci, 22, 1228 (1982).
[16] Mijovic J., Polym. Eng. Sci., 34, 381 (1994).
[17] Hutchinson J.M., Prog. Polym. Sci, 20, 703 (1995).
[18] Plazek D.J., Z.N. F., J. Polym. Sci. Polym. Phys., 28, 431 (1990).
[19] Le Huy H.M., Bellenger V., Verdu J., Paris M., Polym. Degrad. Stab., 41, 149 (1993).
[20] Tsotsis T.K., J. Comp. Mat., 29, 410 (1995).
[21] Förster P., Hausschildt K.R., Huber J., Piecha G., Pohl D., Wilhelm D., Polym.
Degrad. Stab., 24, 269 (1989).
[22] Barral L., Cano J., Lopez J., Lopez A.J., Nogueira P., Ramirez C., Thermochim. Acta
, 269/270, 253 (1995).
[23] Takezawa Y., Itoh Y., J. Appl. Polym. Sci, 78, 14 (2000).
[24] Lee L.H., J. Polym. Sci., 3, 859 (1965).
[25] Lee L.H., J. Appl. Polym. Sci, 9, 1981 (1965).
[26] Keenan M.A., Smith D.A., J. Appl. Polym. Sci, 11, 1009 (1967).
[27] Leisegang E.C., Stephen A.M., Patterson-Jones J.C., J. Appl. Polym. Sci, 14, 1961
(1970).
[28] Levchik S.V., Camino G., Luda M.P., Costa L., Muller G., Costes B., Polym. Degrad.
Stab., 60, 169 (1998).
[29] Levchik S.V., Camino G., Luda M.P., Costa L., Costes B., Henry Y., Muller G., Morel
E., Polym. Degrad. Stab., 48, 359 (1995).
[30] Paterson-Jones J.C., J. Appl. Polym. Sci, 19, 1539 (1975).
[31] Grassie N., Guy M.I., Tennent N.H., Polym. Degrad. Stab., 14, 125 (1986).
[32] Costes B., Reyx D., Platzer N., Macromol. Chem., 190, 349 (1989).
[33] Le Huy H.M., PhD-Thesis, Ecole Nationale Supérieure des Mines de Nancy, Nancy
(F), (1990)
[34] Le Huy H.M., Bellenger V., Paris M., Verdu J., Polym. Degrad. Stab., 35, 171 (1992).
[35] Le Huy H.M., Bellenger V., Verdu J., Polym. Degrad. Stab., 35, 77 (1992).
[36] Cole K.S., Cole R.H., J. Chem. Phys., 9, 341 (1941).
[37] Macdonald J.R., Inpedance Spectroscopy: Emphasizing Solid Materials and
Systems, John Wiley & Son, New York, (1987).
[38] Bower D.I., An Introduction to Polymer Physics, Cambridge University Press,
Cambridge, (2002).
[39] Tcharkhtchi A., Lucas A.S., Trotignon J.P., Verdu J., Polymer, 39, 1233 (1998).
[40] Havrilak S., S. N., Polymer, 8, 161 (1967).
[41] Perez J., Physique et Mécanique des Polymères Amorphes, TEC and DOC-
Lavoisier, Paris, (1992).
[42] Cavaillé J.Y., Perez J., Johari G.P., J. Non-Cryst. Soilds, 172-174, 1028 (1989).
[43] Sindt O., Perez J., Gerard J.F., Polymer, 37, 2989 (1996).
[44] Perez J., Cavaillé J.Y., Jourdan C., Rev. Phys. Appl., 23, 125 (1988).
[45] Pearce P.J., Ennis B.C., Morris C.E.M., Polym. Commun., 29, 93 (1988).
[46] Morgan R.T. In: Advances in Polymer Science (ed. Dusek K.), vol. 72, Springer-
Verlag, Berlin, pp. 1, (1985).
[47] Morgan R.J., Mones E.T., J. Appl. Polym. Sci., 33, 999 (1987).
[48] Morgan R.T., Mones E.T., Steele W.J., Polymer, 23, 295 (1982).
[49] DIN, Bestimmung der Biegeeigenschaften von Kunststoffen, DIN EN ISO 178, Berlin,
1997
[50] Harrick N.J., Internal Reflection Spectroscopy, Wiley Interscience, New York, (1967).
[51] Do T.T., Celina M., Fredricks P.M., Polym. Degrad. Stab., 77, 417 (2002).
[52] Mikols W.J., Seferis J.C., Apicella A., Nicolais L., Polym. Compos., 3, 118 (1982).
[53] Colombini D., Martinez-Vega J.J., Merle G., Polymer, 43, 4479 (2002).
[54] Maxwell I.D., Pethrick R.A., J. Appl. Polym. Sci, 28, 2363 (1983).
[55] Pognay G.A., Brit. Polym. J., 1, 177 (1969).
 Thermo-oxidative degradation behaviour 183

[56] Arridge R.G.C., Polymer, 13, 450 (1972).

[57] Nogueira P., Ramirez C., Torres A., Abad M.J., Cano J., Lopez J., Lopez-Bueno I.,
Barral L., J. Appl. Polym. Sci, 80, 71 (2001).
[58] Pethrick R.A., Hollins E.A., McEwan I., Pollock E.A., Hayward D., Johncock P.,
Polym. Int., 39, 275 (1996).
[59] Pascault J.P., Sautereau H., Verdu J., Williams R.J.J., Thermosetting Polymers,
Marcel Dekker, New York, (2002).
[60] Morel E., Bellenger V., Bocquet M., Verdu J., J. Mater. Sci., 24, 69 (1989).
[61] Cavaillé J.Y., Perez J., Johari G.P., J. Non-Cryst. Soilds, 131-133, 935 (1991).
[62] Kambour R.P., Macromolecules, 1, 393 (1968).
[63] DIN, Electrical insulating materials – Properties of thermal endurance, DIN-IEC
60216/T1, Berlin, 2001
[64] Product data sheet, CY225/HY925, Vantico, Switzerland, 2001
[65] Product data sheet, CY225/HY227, Vantico, Switzerland, 2001
[66] DIN, Electrical insulating materials - Thermal endurance properties - Part 3:
Instructions for calculating thermal endurance characteristics, DIN-IEC 60216-3,
2002
[67] Buch X., Shanahan M.E.R., Polym. Degrad. Stab., 68, 403 (2000).
[68] Wu C.S., Liu Y.L., Chiu Y.C., Chiu Y.S., Polym. Degrad. Stab., 78, 41 (2002).
[69] Wang W.J., Perng L.H., Hsiue G.H., Chang F.C., Polymer, 41, 6113 (2000).
[70] Dyakonov T., Mann P.J., Chen Y., Stevenson T.K., Polym. Degrad. Stab., 54, 67
(1996).
[71] Lutter-Kayser H., Diploma-Thesis, University of Applied Science, Osnabrück, (2001)
[72] Montserrat S., Malek J., Colomer P., Thermochim. Acta , 336, 65 (1999).
[73] Montserrat S., Malek J., Colomer P., Thermochim. Acta , 313, 83 (1998).
[74] Wunderlich B., Thermal Analysis, Academic Press Inc., San Diego, (1990).
[75] Kissinger H.E., Anal. Chem., 29, 1702 (1957).
[76] Ozawa T., J. Therm. Anal., 2, 301 (1970).
[77] Ozawa T., Bull. Chem. Soc., 38, 1881 (1965).
[78] Flynn J.H., Wall L.A., Polym. Lett., 4, 323 (1965).
[79] Rose N., Bras L.M., Bourbigot S., Delobel R., Costes B., Polym. Degrad. Stab., 54,
355 (1996).
[80] Dyakonov T., Chen Y., Holland K., Drbohlav J., Burns D., Vander Velde D., Seib L.,
Soloski E.J., Kuhn J., Mann P.J., Stevenson T.K., Polym. Degrad. Stab., 53, 217
(1996).
[81] Greer R.H., Proc. 24th Nat. SAMPE Symp. Exhib., Azusa (CA), Book 2, pp. 1039,
(1979).
[82] Patel S.P., Patel R.G., Thermochim. Acta , 202, 97 (1992).
[83] Rose N., M. L.B., Delobel R., Costes B., Henry Y., Polym. Degrad. Stab., 42, 307
(1993).
[84] Crank J., The Mathematics of Diffusion, 2nd edition, Oxford University Press,
London, (1975).
[85] Bishop D.P., Smith D.A., J. Appl. Polym. Sci., 14, 205 (1970).
[86] Lee L.H., J. Appl. Polym. Sci., 9, 1981 (1960).
[87] Koenig J.L., Appl. Spectrosc., 29, 293 (1975).
[88] Mailhot B., Morlat S., Gardette J.L., Polymer, 41, 198 (2000).
[89] Antoon M.K., PhD-Thesis, Case Western Reserve University, Cleveland (OH),
(1980)
[90] Antoon M.K., Koenig J.L., J. Polym. Sci., 19, 549 (1981).
[91] Bellamy L.J., The Infrared Spectra of Complex Molecules, Chapman and Hall, New
York, (1975).
[92] Colthup N.B., Daly L.H., Wiberley S.E., Introduction to Infrared and Raman
Spectroscopy, Academic Press, San Diego, (1990).
[93] Pretsch E., Bühlmann P., Affolter C., Structure Determination of Organic
Compounds, Springer Publisher, Berlin, (2000).
184 Chapter 4

[94] Lin S.C., Bulkin B.J., Pearce E.M., J. Polym. Sci., 17, 3121 (1979).
[95] Olliver-Dureault V., Gosse B., J. Appl. Polym. Sci, 70, 1221 (1998).
[96] Bellenger V., Verdu J., J. Appl. Polym. Sci., 30, 363 (1985).
[97] Bellenger V., Verdu J., J. Appl. Polym. Sci., 28, 2677 (1983).
[98] Bellenger V., Verdu J., J. Appl. Polym. Sci., 28, 2599 (1983).
[99] Smith B., Infrared Spectral Interpretation - A Systematic Approach, 1 ed., CRC
Press, New York, (1999).
[100] Costes B., Rexy D., Macromol. Chem., 190, 349 (1989).
[101] Wilson E., Phys. Rev., 45, 706 (1934).
[102] De Bakker C.J., George G.A., St John N.A., Spectrochim. Acta, 49A, 739 (1993).
[103] Atherton N.M., Banks L.G., Ellis B., J. Appl. Polym. Sci, 27, 2015 (1982).
[104] Parvatareddy H., Wang J.Z., Lesko J.J., Dillard D.A., Reifsnider K.L., J. Comp. Mat.,
30, 210 (1996).
[105] Madhukar M., Bowles K.J., Papadopoulos D.S., J. Comp. Mat., 31, 596 (1997).
[106] Colin X., Marais C., Verdu J., Polym. Degrad. Stab., 78, 545 (2002).
[107] Decelle J., Huet N., Bellenger V., Polym. Degrad. Stab., 81, 239 (2003).
[108] Audouin L., Langlois V., Verdu J., De Bruijn J.M.C., J. Mater. Sci., 29, 569 (1994).
[109] Buch X., Shanahan M.E.R., J. Appl. Polym. Sci, 76, 987 (2000).
[110] Bourbigot S., Le Bras M., Delobel R., Costes B., J. Fire Sci, 13, 3 (1995).
[111] Paterson-Jones J.C., PhD-thesis, University of Cap Town, Cap Town (South Africa),
(1969)
[112] Zhang G., Pitt W.G., Goates S.R., Owen N.L., J. Appl. Polym. Sci., 54, 419 (1994).
[113] Musto P., Ragosta G., Russo P., Mascia L., Macromol. Chem. Phys., 202, 3445
(2001).
General conclusions and perspectives 185

Chapter 5

5 General conclusions and perspectives

In this thesis novel co-anhydride cured epoxy formulations have been investigated in order
to assess the advantages and limitations of these new thermosetting materials in industrial
applications. The main objective of this work was to develop a fundamental understanding
of:
the principal cure mechanism and kinetics
the general thermo-mechanical deformation and fracture behaviour
the thermo-oxidative ageing performance and mechanism

1.) Cure mechanism and kinetics

The thermal polymerization as a function of time and temperature was systematically
studied by means of differential scanning calorimetry, rheometry and FT-Raman
spectroscopy. It was shown that the investigated anhydride-cured epoxy system reacts
mainly by polyesterification. Possible side reactions including the homopolymerization of
MA were considered. The main identified side reaction was decarboxylation of the
anhydrides that may produce some autocatalysis, but this is a minor contribution to the
cure kinetics in the investigated temperature range between 55 to 120°C. The applied in-
situ methods revealed that isothermal conditions are essential to provide reliable and
meaningful results for analysis of the curing reaction. A simple first order reaction scheme
for both epoxy and anhydride with carboxylic acid and hydroxyl groups as reactive
intermediates described the reaction profile until the polymerization was influenced by
chemo-rheological changes due to gelation and vitrification transitions.
The use of a temperature modulated DSC (TMDSC) technique allows the differentiation of
the MY721/MA and MY721/HHPA reaction rates that consequently provides evidence for
the existence of an independent MY721/MA and MY721/HHPA sub-network in the cured
state. Additionally, chemo-rheological studies provided insight in the gelation and a
vitrification phenomenon, which finally allows the reconstruction of a principal time-
temperature-transformation diagram in order to predict the entire isothermal curing
process.
186 Chapter 5

2.) Thermo-mechanical deformation and fracture behaviour

The characteristic deformation and fracture behaviour of neat and silica-filled
MY721/MA/HHPA-samples was studied as a function of temperature, stress rate, filler
fraction, particle-size and filler-matrix adhesion.
Microscopic studies of fractured samples showed that diffuse matrix shear yielding and
micro-shear banding are the dominating deformation mechanisms. These processes are
amplified by higher temperature and increased stress rates. Moreover, with increasing
silica content, the tensile and compression modulus of composites increased whereas the
corresponding strength values remain quasi-constant.
The crack propagation and fracture behaviour was examined by applying the concept of
linear elastic fracture mechanics. The critical stress intensity factor (KIc) and the critical
strain energy release rate (GIc) significantly decreased at temperatures above 100°C. It
was demonstrated that this unconventional behaviour is related to the fact that the
MY721/HHPA sub-network was exposed to temperatures above its glass transition
temperature.
Attempts to improve the fracture toughness while maintaining the thermo-mechanical
property profile has been addressed by using inorganic hard particles, epoxy
functionalized hyperbranched polymers (HBPs), and hybrid-composites through a modifier
combination. Increasing the inorganic filler fraction resulted in an effective enhancement of
the fracture toughness due to the implementation of the “crack-pinning” mechanism. The
corresponding KIc value increased from 0.6 MPam0.5 for the neat resin to 1.6 MPam0.5 for
the silica-filled (60 wt.%) system.
The shell chemistry of different HBP-modifiers and the related morphology-property
relationship were thoroughly investigated. Tailoring the number of functional shell groups
can be effectively used to control the final morphology of the materials. Increasing the
number of functional epoxy end-groups increases the solubility of the HBP modifier in the
matrix. Thus, it is possible to adapt blends with variable morphologies ranging from phase-
separated to totally homogenous. The phase-separated blends revealed a pronounced
anisotropic structure in the cured state due to density differences between the modifier and
matrix. Consequently, only a moderate improvement in toughness was obtained. On the
other hand the homogenous blends showed the best improvement in toughness and the
corresponding critical stress intensity factor was increased by about 50% from 0.6 to 0.9
MPam0.5. Synergistic effects in hybrid-systems were not observed and the inorganic filler
dominated the fracture behaviour. Attempts to establish a chemically induced phase
separation process to precisely control the morphology were limited, because of
inadequate entropic and enthalpic changes in the investigated anhydride-cured epoxies.
General conclusions and perspectives 187

3.) Thermo-oxidative degradation

An important aspect in this thesis was to explore the effects of thermo-oxidative long-term
degradation on thermo-mechanical properties, thermo-gravimetric behaviour, and the
related architectural network changes.
A new approach for monitoring the ageing process by combining mechanical Cole-Cole
plots with the theoretical Perez model has been successfully applied. It was demonstrated
that this method provides a meaningful correlation between the relaxational behaviour and
the intrinsic network variations. It was shown that the appearance of a secondary ω-
relaxation is primarily associated with network-irregularities (e.g. dangling chains) that are
related to the presence of maleic anhydride.
Thermo-gravimetric analysis of neat and particulate-filled materials revealed a double
stage degradation mechanism under air atmosphere. The low temperature degradation
stage predominantly reflects the degradation of structural irregularities, such as untangled
chains or end-groups of dangling chains. The followed high temperature degradation stage
includes the chemical degradation of the more stable backbone structure.
Thermal endurance of the flexural behaviour showed that entire thermo-oxidative stability
of MA-cured networks is superior to HHPA-cured networks. The thermo-oxidative
performance of the MY721/MA-system proved to be an economical and environmental
alternative to the well-established TGDDM/DDS-system.
The experimental measurements on hybrid-systems showed that the introduction of
inorganic particles (e.g. disperse silica (SiO2) or alumina (Al2O3)) improved the thermo-
oxidative stability and the weight-loss rate. The addition of alumina-filler outperformed
silica-particles.
The degradation mechanism and the corresponding oxidation profile were studied by FT-
IR micro-ATR spectroscopy. The spectral information provided strong evidence that the
attack of the relatively weak C-N and C-O bonds in TGDDM-based formulations generally
limits the use of amino-glycidyl based systems for long-term applications at temperatures
above 150°C.
188 Chapter 6

Chapter 6

6 Acknowledgments
First of all I would like to gratefully thank my principal supervisor Professor Graeme George
for providing me the opportunity, the motivation and the guidance throughout this thesis.
The advice and constructive criticism of my doctoral committee, Professor Peter Fredericks,
Professor John Bell and Dr. Felix Greuter have been very helpful in supporting my research
work and the preparation of this thesis. I wish to express my gratitude to Professor Friedrich
Vohwinkel for many fruitful discussions.

I would like to acknowledge my colleagues in the Polymer group at ABB Corporate

Research in Switzerland. In particular Dr. Andrej Krivda, Dr. Stefan Förster, Patrick Meier
and Leopold Ritzer helped me to start up my research project and provide me with many
insights. Furthermore, I would like to thank Dr. Yvo Dirix who helped me to correct the
manuscript.

I am very grateful to the members of the Polymer-Group at the Queensland University of

Technology (QUT), with whom I had many fruitful discussions. In particular I would like to
express my appreciation for the assistance received from Dr. Llew Rintoul and Dr. Greg
Cash who have been involved in many experiments during my stay at QUT. Dr. Imelda Keen
is gratefully acknowledged for the assistance by providing valuable support on the FT-
Raman experiments at QUT. Lars Ischtschuk, Sören Neugebauer, and Claudia Hofschrörer
from the University of Applied Science, Osnabrück, are thanked for the development of
some of the experimental methodology in this work while undertaking research at QUT.

Last but not least I would like to thank my wife Barbara, our son Lucas Simon, and both of
our families for their patience and understanding during this study. Their support has been
invaluable.
Curriculum Vitae 189

Chapter 7

7 Curriculum Vitae (CV)

The author of this thesis was born in Osnabrück, Federal Republic of Germany, on 1st
February 1972. After graduation in 1991, he spend one year in military service. In 1993 he
started to study Material Science at the University of Applied Science in Osnabrück. Under
supervision of Prof. F. Vohwinkel and Dr. F. Greuter he obtained his degree in 1997 on the
project “A comparison of Silicone with EPDM for electrical outdoor insulation“.

In the same year he joined the Polymer and Process Technologies Group of the Corporate
Research Center of Asea Brown Boveri (ABB) in Baden, Switzerland, where he worked as a
research scientist on new polymeric outdoor insulation as well as ageing phenomena of
polymers.

In 2000, he started working on his doctoral degree in a joint project between the School of
Physical and Chemical Sciences, which is part of the Faculty of Science at the Queensland
University of Technology in Brisbane, Australia and ABB Corporate Research in
Switzerland. The results on the work “Characterization of Novel Co-Anhydride cured Epoxy
Resins” are described in this thesis.

In May 2002, he has been promoted to a position of a project leader at ABB Corporate
Research. Since the beginning of 2004 he has taken over the lead of the group “Insulation
and Polymer Technology”. His recent work is mainly focused on the investigation and
development of novel polymeric insulation principles and optimization of existing polymer
processing technologies in the field of electrical high voltage equipment.

Research Publications

J. Rocks, M. Halter, G. George, F. Vohwinkel, Calorimetric and rheological characterization

of a high-performance epoxy curable at low temperatures, Polym. Int., 52, 1749-1757, 2003

J. Rocks, G. George, F. Vohwinkel, Curing kinetics and thermo-mechanical behaviour of co-

anhydride cured aminoglycidyl epoxy resins, Polym. Int., 52, 1758-1766, 2003

J. Rocks, L.Rintoul, F. Vohwinkel, G. George, The kinetics and mechanism of cure of an

amino-glycidyl epoxy resin by a co-anhydride as studied by FT-Raman spectroscopy,
submitted to Polymer
190 Chapter 7

Conference presentations

Rocks J., George G., Förster S., Vohwinkel F., Novel Anhydride Cured Epoxy Resins, 24th
Australasian Polymer Symposium, Beechworth, Victoria, Australia, 2001

Rocks J., George G., Curing Characterization of an Anhydride cured Epoxy-System by

Chemorheological methods, 25th Australasian Polymer Symposium, Armidale, NSW,
Australia, 2002

Rocks J., Rintoul L., George G.,The kinetics and mechanism of cure of a high performance
epoxy resin as studied by FT-Raman spectroscopy, 19th International Conference on
Raman Spectroscopy, Gold Coast, QLD, Australia, 2004

Patents

In the last five years, more than fifteen international patent applications have been filed in
the fields of polymeric insulation, polymer processing technologies and electrical equipment