"Mechanical properties of a bio-sourced

epoxy resin for composite applications"

Jost, Noah

ABSTRACT

The use of composite materials such as fiber reinforced polymers has become more and more popular
during recent years. In particular, epoxy resins are often selected as polymer matrix due to their excellent
mechanical properties. Nevertheless, their use is related to significant environmental concerns, which is
the reason why bio-sourced epoxy resins with lower carbon footprint are subject of research. Together
with natural fibers, the development of bio-sourced composite materials becomes possible. This work
investigates mechanical properties of the partially bio-sourced InfuGreen810 epoxy resin. Samples
provided by different curing cycles are tested using uni-directional compression and tensile tests. In this
way, the stress-strain response of the material is obtained. The cycles are categorized according to their
curing degree, measured with a DSC-analysis. Different strain rates are performed. Stiffness and yield
stress are higher for lower degrees. A better packing of chains leading to stronger secondary interactions
could be identified as physical origin. In tension, samples with low curing degrees are more brittle than
fully-cured ones, showing an unusual high ductility. The fracture surfaces are investigated with a SEM.
Mechanical properties are further related to the glass transition temperature Tg by studying incomplete
cures. The yield drop and the stiffness both decrease linearly with Tg. Finally, the creep behavior under
uni-axial compression is studied for different loads, all below yield stress. No tertiary creep, i.e. no failure
due to static fatigue could...

CITE THIS VERSION

Jost, Noah. Mechanical properties of a bio-sourced epoxy resin for composite applications. Ecole
polytechnique de Louvain, Université catholique de Louvain, 2020. Prom. : Pardoen, Thomas. http://
hdl.handle.net/2078.1/thesis:25203

Le répertoire DIAL.mem est destiné à l'archivage DIAL.mem is the institutional repository for the
et à la diffusion des mémoires rédigés par les Master theses of the UCLouvain. Usage of this
étudiants de l'UCLouvain. Toute utilisation de ce document for profit or commercial purposes
document à des fins lucratives ou commerciales is stricly prohibited. User agrees to respect
est strictement interdite. L'utilisateur s'engage à copyright, in particular text integrity and credit
respecter les droits d'auteur liés à ce document, to the author. Full content of copyright policy is
notamment le droit à l'intégrité de l'oeuvre et le available at Copyright policy
droit à la paternité. La politique complète de droit
d'auteur est disponible sur la page Copyright
policy

Available at: http://hdl.handle.net/2078.1/thesis:25203 [Downloaded 2025/02/11 at 10:34:38 ]

 École polytechnique de Louvain

Mechanical properties of a
bio-sourced epoxy resin for
composite applications

Author: Noah J OST

Supervisor: Thomas PARDOEN
Readers: Vincent D ESTOOP, Christian BAILLY
Academic year 2019–2020
Master [120] in Chemical and Materials Engineering
 i

Abstract

The use of composite materials such as fiber reinforced polymers has become more and
more popular during recent years. In particular, epoxy resins are often selected as poly-
mer matrix due to their excellent mechanical properties. Nevertheless, their use is related
to significant environmental concerns, which is the reason why bio-sourced epoxy resins
with lower carbon footprint are subject of research. Together with natural fibers, the de-
velopment of bio-sourced composite materials becomes possible.

This work investigates mechanical properties of a partially bio-sourced InfuGreen 810

epoxy resin. Samples provided by different curing cycles are tested using uni-directional
compression and tensile tests. In this way, the stress-strain response of the material is
obtained. The cycles are categorized according their curing degree, measured with a
DSC-analysis. Different strain rates are performed. Stiffness and yield stress are higher
for lower degrees. A better packing of chains leading to stronger secondary interactions
could be identified as physical origin. In tension, samples with low curing degrees are
more brittle than fully-cured ones, showing an unusual high ductility. The fracture sur-
faces are investigated with a SEM. Mechanical properties are further related to Tg by
studying incomplete cures. The yield drop and the stiffness both decrease linearly with
Tg . Finally, the creep behavior under uni-axial compression is studied for different loads,
all below yield stress. No tertiary creep, i.e. no failure due to static fatigue could be ob-
served.

The findings in this study could be a step forward in the development of bio-sourced
fiber reinforced composites.
 iii

Acknowledgments

J’aimerais bien mentionner dans cette section tous les personnes qui m’ont aidé afin de
réaliser ce travail.

Tout d’abord, j’aimerais remercier Thomas Pardoen, Professeur à l’Ecole Polytechnique

de Louvain (EPL) de l’Université Catholique de Louvain (UCLouvain) qui est à l’origine
de ce mémoire et qui n’a pas manqué de curiosité pour les résines bio-sourcées.

J’adresse également mes remerciments à Monsieur Vincent Destoop pour m’avoir épaulé
tout au long de l’année. Ses conseils étaient très précieux. Sans son aide, ce mémoire
n’aurait certainement pas pu être fait. Sa rigueur scientifique et son approche péda-
gogique vont certainement m’impacter au-delà de ce mémoire.

Merci également à Waël Ballout et Pascal van Velthem pour la préparation des échan-
tillons, leur expertise physico-chimique et le thé durant les pauses.

Je remercie également Laurence Ryelandt pour les photos au SEM dans ces conditions
particulières ainsi que Alban Maton pour les usinages.

J’adresse aussi ma gratitude à Charline, qui avait le même sujet de mémoire, pour son
soutien et les échanges constructifs.

Finalement, ce travail marque aussi la fin de mes cinq ans au sein de la communauté
universitaire de l’UCLouvain. Je remercie ici ma famille qui m’a accompagnée toute au
long de cette période ainsi que tous les amis et amies que j’ai pu rencontrer durant mes
études dont certaines sont devenus en quelque sorte une deuxième famille (BK, AK3, ...).
Finalement, je dédie mes efforts durant ces cinq dernières années à mes proches décédés
depuis lors et qui ne voyeront jamais leur fils, petit-fils ou encore leur ami, diplômé.
 iv

Contents

1 Introduction 1

2 State of the art 3

2.1 Bio-sourced composite materials . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.1 The polymer matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.2 The fiber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Architecture of highly crosslinked polymers . . . . . . . . . . . . . . . . . . . 5
2.2.1 Molecular network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Epoxy molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.3 Hardeners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Formation of epoxy-based polymers with hardeners . . . . . . . . . . . . . . 6
2.3.1 Reaction mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3.2 Curing models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Influence of curing on physico-chemical properties . . . . . . . . . . . . . . 10
2.4.1 Glass transition temperature . . . . . . . . . . . . . . . . . . . . . . . 10
2.4.2 Free volume and its influence on aging and rejuvenation . . . . . . . 13
2.4.3 Shrinkage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Mechanical behavior of epoxy resins . . . . . . . . . . . . . . . . . . . . . . . 16
2.6 Micromechanical approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.6.1 Spring-dashpot analogy . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.6.2 Molecular flow theories . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Eyring model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Argon model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.6.3 Localization phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Crazing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Shear yielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Competition between crazing and shear yielding . . . . . . . . . . . 24
2.7 Fracture of epoxy resins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7.1 Brittle fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7.2 Fracture in compression . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.8 Creep behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.8.1 Creep compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.8.2 Creep strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.8.3 Superposition principle . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Time-Temperature superposition . . . . . . . . . . . . . . . . . . . . . 30
Time-Aging superposition . . . . . . . . . . . . . . . . . . . . . . . . . 31
 v

3 Scope of testing 32

4 Materials and methods 34

4.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.1.1 Epoxy-hardener system . . . . . . . . . . . . . . . . . . . . . . . . . . 34
The resin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
The hardener . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.1.2 Physico-chemical properties of the resin . . . . . . . . . . . . . . . . . 34
4.1.3 Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Presentation of Curing cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Differential scanning calorimetry . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4 Mechanical testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4.1 Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4.2 Note on constant crosshead displacement vs. constant true strain rate 41
4.4.3 Testing devices and operational modes . . . . . . . . . . . . . . . . . 42
Uni-axial tensile tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.4.4 Uni-axial compression tests . . . . . . . . . . . . . . . . . . . . . . . . 42
4.5 Creep tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6 Fractography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.7 Incomplete cures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

5 Results 47
5.1 Difficulties in sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . 47
5.2 Physico-chemical testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.3 Mechanical testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.3.1 Elastic behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.3.2 Yield and post-yield behavior . . . . . . . . . . . . . . . . . . . . . . . 50
Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3.3 Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.4 Fractographic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.4.1 Tensile specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.4.2 Compression specimens . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.5 Influence of post-cure on mechanical properties . . . . . . . . . . . . . . . . 58
5.6 Static creep in compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

6 Discussion 60
6.1 Curing process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.2 Note on stress-strain curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.3 Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.4 Yielding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.5 Hardening and fracture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.5.1 Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.5.2 Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.6 Influence of the post-cure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
 vi

6.7 Creep behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

7 Conclusion 70

A Physico-chemical results of InfuGreen810-SD4771 system 72

B Mechanical properties of the RTM6 resin 74

Bibliography 75
 vii

List of Figures

2.1 The interior door panel of a car, made from a hemp-polyethylene com-
pound. From:[6] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 General representation of an epoxy molecule. From: [19] . . . . . . . . . . . 6
2.3 Left: an amine group of the hardener (circled in red) attacks the carbon
atom of the epoxy. A bond between both is formed. Afterwards, the oxygen
atom captures a hydrogen atom to become a hydroxyl. From: [23] . . . . . . 7
2.4 Left: a single hardener molecule can link four epoxy molecules. Right:
Representation of a highly crosslinked network. From: [23] . . . . . . . . . . 7
2.5 The curing degree α in function of time in different isothermal conditions.
Adapted from [24]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.6 TTT-diagram for a thermoset polymer. From: [32] . . . . . . . . . . . . . . . 12
2.7 Molecular view of free volume build up. From: [38] . . . . . . . . . . . . . . 13
2.8 Specific volume in function of temperature during a cooling process. Blue
lines show the cooling paths, the gray surface represents the amount of free
volume. The graph states that for two cooling rates, the faster one (q1 that
is quicker than q2 ) yields more free volume (point A). Moreover, as already
discussed, Tg is different for both cases. When the sample is aged, free
volume is further decreased until point D is reached. It is also shown that
equilibrium (point D) will never be achieved. From: [38] . . . . . . . . . . . 14
2.9 Stress-strain curve for a glassy polymer in a uni-directional compression
test (green curve), with consecutive unloading (red curve) and strain re-
covery (orange line). Source: [46] . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.10 a) Influence of thermal history, adapted from [40] b) Influence of tempera-
ture and strain rate. From: [41] . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.11 Spring-dashpot model of a glassy polymer network. Source: [56] . . . . . . 19
2.12 a) Compression . b) Formation of cavities From: [55] . . . . . . . . . . . . . . 22
2.13 True stress in function of the dwell ratio for a uni-axial tensile test. The
panels show the formation of a neck that is diffused on the whole length of
the reduced section of the sample. From: [40] . . . . . . . . . . . . . . . . . . 23
2.14 True strain ratio in function of chain density. The two regimes can be dis-
tinguished. Higher crosslink density favors yielding [41] . . . . . . . . . . . 24
2.15 true stress - true strain curve of a 1D tensile test for a) Polycarbonate (PC):
as a post-yield response, shear bands are formed and a lowering in cross-
section can be observed on a relative large area. b) Polystyrene (PS): per-
pendicular to the loading direction, localized crazes appear, resulting in
failure of the sample. From: [40] . . . . . . . . . . . . . . . . . . . . . . . . . 25
 viii

2.16 Four different zones of a fracture plane. From: [61] . . . . . . . . . . . . . . . 27

2.17 Failure of compression specimens due to a) shear b) combination between
shear and brittle behavior. From: [46] . . . . . . . . . . . . . . . . . . . . . . 27
2.18 Creep behavior of a material tested with constant load. Three regions are
distinguished: primary, secondary and tertiary creep. Adapted from: [65] . 29
2.19 log - log plot for creep compliance against time of an epoxy. The master
curve is shown in blue (central curve). If the temperature for the creep test
is set above the reference of the master curve, a horizontal shift to the left
by a T is done (red curve). When the sample has been aged for a given time,
a horizontal shift to the right by a E can predict the creep compliance (green
curve). From: [38] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.1 TTT-diagram for the epoxy-hardener system investigated. From [67] . . . . 35

4.2 Testing geometries for a) compression and creep test b) tensile test. Adapted
from [46] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3 Presentation of the different curing cycles . . . . . . . . . . . . . . . . . . . . 38
4.4 Scheme of a DSC machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.5 Strain-stress curve for uni-axial loading. Adapted from: [46] . . . . . . . . . 41
4.6 True strain rate of CCDR vs. CSR for a static compression test. From: [46] . 42
4.7 A specimen deformed during a compression test. Left: with lubricant.
Right: Without lubricant. Source: [46] . . . . . . . . . . . . . . . . . . . . . . 43
4.8 a) Screw-driven testing machine b) LVDT strain sensor c) Subpress d) Creep
testing device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.9 Temperature paths for samples of the incomplete cure study . . . . . . . . . 46

5.1 a) and b): Defects inside the tubes, respectively cracks and air bubbles. c)
Tensile specimen with defect d) Compression specimen with chimney . . . 47
5.2 Thermograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.3 True strain - true stress curves for the different curing cycles at 1mm/min
in compression (left) and tension (right) . . . . . . . . . . . . . . . . . . . . . 49
5.4 Elastic modulus in tension and compression, tested at room temperature. . 49
5.5 Yield values in function of curing cycles  0, 1, ◦ 1, ♦ 10mm/min. a) σY and
σl in compression. b) σY and σy0,2% in tension . . . . . . . . . . . . . . . . . . 50
5.6  0, 1, ◦ 1, ♦ 10mm/min a) True stress in compression and tension. Upper
data are compression, lower ones tensile values. b) True strain at failure
for compression and tension. Yellow color labels compression, green color
tensile results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
 ix

5.7 RTM cycle a) The overall appearance shows only slight brittleness at the
borders of the surface. Long cracks propagate linearly from the defect to
the borders. Those long cracks are accompanied by micro-cracks, as it can
be observed on panel b). Some of them propagate radially whereas oth-
ers seem to be initiated by perpendicular, scar-like extensions of the main
cracks (highlighted by the white arrow). A zoom on the defect is showed
on panel c). For orientation, an arrow is added pointing to the void. Strong
roughness is formed around the defect and the long cracks seem to be built
up there. Between rough cracks and the mirror zone, a moisture-like zone
that is a few tenth of µm in thickness, appears. . . . . . . . . . . . . . . . . . 53
5.8 RTM cycle with defect A pre-existing defect was already visible before
testing. The defect is in this case a huge bubble that can be distinguished
without any zoom on a). The crack deviation inside the mirror zone is
visible on b). At the same level, the parabolic pattern begins. Panel d)
shows the inside of the bubble, no specific texture is visible. Note that
some spots (marked with the arrow on d) are probably dust and therefore
of no interest. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.9 T amb Panel a) reveals that most of the area is covered by a relief. On b),
the defect with the laminar micro-cracks in the center is highlighted (the
bright spot on this micrograph is a dust particle that should be ignored),
followed by a crack multiplication. This first crack propagation appears
still in the mirror-like zone. As it can be deduced on panel c), it seems
that the growth of these cracks is stopped abruptly, upper and lower levels
appear. Therefore, regions that are not in-plane with the initial propaga-
tion appear. Lastly, on the borders of capture d), the propagation becomes
rougher, a wavy pattern is visible following the horizontal line on d). . . . . 55
5.10 Like 5.7, nearly no roughness can be observed. A few long cracks propagate
over several mm. The micro-flow is parallel with them and crack deflection
is visible for both on panel c). Panels b) and d) point out the defect, proba-
bly due to decohesion between a small impurity and the resin. At the edge
between particle and fracture surface, cracks are initiated. . . . . . . . . . . 56
5.11 Two representative compression specimens after test. a) High curing de-
gree, showing many fragments. b) Low curing degree, showing cracks but
no real fragments as the sample did not undergo failure. The white "wings"
on a) and areas on b) are remaining teflon-papers. . . . . . . . . . . . . . . . 57
5.12 a) Thermogram for samples with 0min and 240min of isothermal curing at
80◦ C b) Tg as a function of post-cure time . . . . . . . . . . . . . . . . . . . . 58
5.13 red: 10mm/min blue: 1mm/min a) True stress-strain curves b) Young’s
modulus c) True strain at failure d) True yield stresses as a function of Tg .
Values of KY are plotted and marked on d) as well . . . . . . . . . . . . . . . 59
5.14 Creep increments and steady-state creep for three different loading condi-
tions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

6.1 Yield stresses of different epoxy resins as a function of Tg − T. Values of

InfuGreen are those of high-degree batches. Adapted from [72] . . . . . . . . 63
 x

6.2 Ultimate stress and strains for different epoxy resins as a function of T/Tg
are shown. Values of InfuGreen are those of high-degree batches. Adapted
from [72] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

A.1 Linear dependency between Tg and α. From: [67] . . . . . . . . . . . . . . . . 72

A.2 Thermograms of the InfuGreen810 resin cured at higher temperatures . . . 73
A.3 Isothermal evolution of curing degree with time . . . . . . . . . . . . . . . . 73

B.1 Stress-strain response of RTM6 resin. From [41] . . . . . . . . . . . . . . . . . 74

B.2 Fracture surface of RTM6 resin. From [46] . . . . . . . . . . . . . . . . . . . . 74
 xi

List of Tables

4.1 Different properties of epoxy and hardener . . . . . . . . . . . . . . . . . . . 35

5.1 Ky is the ratio σY /σl , GR the hardening modulus, values of Tg retaken from
section 5.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
 xii

Nomenclature

Abbreviations
BPA Bisphenol-A
CCDR Constant crosshead displacement
CSR Constant true strain rate
DSC Differential scanning calorimetry
FRP Fiber reinforced polymer
FSE Fracture surface energy
MACOBIO Matériaux composites bio-sourcées
LEFM Linear elastic fracture mechanics
LVDT Linear variational displacement transducers
PC Post-cure
PE polyethylene
PP Polypropylene
PLA Polylactic acid
RTM Resin transfer molding
SEM Scanning electron microscope
TTS Time temperature superposition
TTT Time-temperature transition
 xiii

Symbols
α Curing degree (or curing extend)
k Reaction rate constant
T Temperature
f (α) Conversion function
Ea Activation energy
R Ideal gas constant
kB Boltzmann’s constant
β Heating rate
Tg Glass transition temperature
TCure Curing temperature
T0 Freezing temperature
Tg∞ Glass transition temperature at full cure
Tgel Temperature at gelation point
Tgel Time at gelation point
E Young’s modulus
e True strain
ė Strain rate
ėss Steady state creep
eR True strain ratio
e NH Neo-Hookean strain
GR Hardening modulus
λ Draw ratio
l Current length
l0 Initial length
NDe Deborah number
ρe Entanglement density
VE Activation volume
σ True stress
σY Yield stress
σl Lower yield stress
∆H LE Activation barrier for yielding
ac Critical defect size
D Creep compliance
KY Yield drop
 1

Chapter 1

Introduction

The use of composite materials such as fiber reinforced polymers (FRPs) has become more
and more popular the recent years. They combine excellent mechanical properties such
as strength and stiffness with a low weight and are therefore selected as material espe-
cially in domains where lowering weight is related to economical gains, for example in
the aeronautical sector. Epoxy resins are particular suited for the use as polymer matrix.

However, their use is related to some issues. As a matter of fact, their production has
a high carbon footprint, especially because their production requires high amounts of en-
ergy. Conventional epoxy resins are petrol-based, meaning that they originate from finite
sources. In addition, recycling of FRPs is hardly possible. Humanity gets aware of those
problems and researchers are looking for sustainable alternatives.

In this context, the FEDER project "Matériaux composites bio-sourcées" (MACOBIO) tries
to develop low carbon footprint materials such as FRPs that are bio-sourced (for further
information, the reader is invited to consult the websites [1, 2]). The following work is
in the frame of this project. From a mechanical point of view, their purpose is the same
as for conventional FRP: maximizing the ratio between mechanical properties and weight
while being sustainable from an environmental point of view.

An understanding of those materials has to be developed to employ them in real-life

applications.

This Master’s thesis aims to characterize an epoxy resin named InfuGreen810 that is (partly)
bio-sourced. Preliminary work by De Halleux (2019) has given information about its
physico-chemical properties. In the frame of this thesis, first attempts are performed
to investigate the mechanical properties. By doing so, the volume needed for physico-
chemical tests is increased to a larger scale in order to carry out a mechanical characteri-
zation.

The following parts constitute the outline of this Master’s thesis: chapter 2 gives a state-of-
the-art. A short and general overview on bio-sourced epoxy resins is given. Afterwards,
still in chapter 2, different aspects related to the preparation, i.e. the curing of epoxy
samples are explained. In addition, literature about the mechanical response of glassy
polymers is reviewed. Chapter 3 gives a scope of the testing. The materials and methods
Chapter 1. Introduction 2

relative to these experiments are given in chapter 4.

Chapter 5 contains the experimental results, turning around three different objectives:

• It turned out that the preparation of samples, known as curing, is related to diffi-
culties as many defects appeared within the resin. Therefore, different curing cycles
are developed.

• Quasi-static tests are performed in uni-axial tension, compression and creep. Re-
sults are superposed with those of other petrol-based resins like the RTM6 system.

• The mechanical response in compression of incompletely cured resins are investi-

gated as well.

The discussion of those findings can be read in chapter 6. A special emphasize is spent on
the relation between physico-chemical properties and the macroscopic response.

Finally, chapter 7 concludes this work. This thesis, together with complementary work
of Van Innis (2020) [3] is the begin of the mechanical characterization of a yet unknown
resin. Future steps are therefore given as a kind of guideline in the perspectives of the
conclusion.
 3

Chapter 2

State of the art

2.1 Bio-sourced composite materials

To discuss the development of sustainable fiber reinforced polymers, the fiber as well as
the polymer itself have to be investigated. The objective of this section is to give a general
overview on what is found in literature to replace conventional composites by more eco-
friendly alternatives.

2.1.1 The polymer matrix

One concern of petrol-based polymers is that they rely on finite resources. Their produc-
tion requires huge amounts of energy and the carbon footprint is quite high. Therefore,
scientists became aware of the possibility to use bio-sourced polymers, based on chemi-
cals produced by plants or bacteria with lower carbon footprints [4].

Commonly, polymers are divided into two categories: thermoplastic and thermosetting
polymers [5]. In contrast to thermoplastics, thermosets can not be (re)melted, what makes
recycling very complicated. One approach is therefore the use of thermoplastics as ma-
trix for composite materials. A petrol-based candidate is for example polypropylene (PP)
or polyethylene (PE). This second polymer is already used in combination with natural
fibers, for example in the automotive industry as nonstructural parts (see fig 2.1) [6, 7].
One bio-sourced candidate that replaces these thermoplastics is polylactic acid (PLA), a
polymer obtained by condensation of lactic acid [8]. For a same natural fiber, bio-based
PLA composites have shown even better mechanical properties than petrol-based PP ones
[9].

Even though thermoplastics present the benefit to be potentially recyclable, a compromise

has to be established because thermosets, like epoxy resins, are more efficient in terms of
mechanical properties. Moreover, they have a better adhesion with the fibers. The pro-
cessing also differs: thermoplastic composites are often produced by creating small pellets
of the fiber/polymer compound in a shear-extruder. This limits the length of the fibers to
2mm [10]. In contrast, thermosets can be processed by resin transfer molding (RTM) or
flex molding that allows to produce composites with fibers which length is only limited
by the dimensions of the mold itself [11]. However, companies already sell thermoplastic
resins that can be processed in the same way as thermosets [12].
Chapter 2. State of the art 4

F IGURE 2.1: The interior door panel of a car, made from a hemp-
polyethylene compound. From:[6]

Another concern of petrol-based epoxies is their toxicity. Most of them are based on
bisphenol-A (BPA), that is related to some environmental and medical issues [13, 14].
To create sustainable high-performance composites, a subject of research is to find eco-
friendly epoxy systems which are not based on petroleum with at least the same proper-
ties as other conventional epoxy resins like RTM6, found in aeronautical applications.

It is important to insist on the fact that bio-sourced is not a synonym for bio-degradable
or compostable. PLA for example is biodegradable in the sens that specific bacteria and
conditions are necessary for their decomposition, but the same environmental concerns
as for conventional plastics arise if there is no correct disposal [8, 15]. The real ecological
benefit of bio-sourced plastics is still discussed. Even if their carbon footprint is lower,
the growth of crops requires space and huge amounts of water. The reader is invited to
consult the parallel work of Charline van Innis (2020) [3] for more information about these
concerns.

2.1.2 The fiber

Modern composites contain fibers of glass or carbon. Bio-sourced alternatives can be
provided by animals (silk, wool, ...) or plants and have the advantage that they are com-
pletely biodegradable, in contrast to many bio-sourced polymers. Examples for vegetable
fibers are flax, hemp, jute or kenaf but there are many other possible candidates. They
are cheap, renewable, have low densities and are competitive in terms of stiffness and
strength with traditional fibers. Well-established manufacturing processes like RTM are
compatible [16, 9].

In terms of mechanical properties, a challenge to overcome is that natural products have

high deviations regarding their quality. As most of them are provided by crops, the
Chapter 2. State of the art 5

growth condition (weather, geographical origin, ...) has an influence on properties like
stiffness or strength.

In a similar way, the diameter is hardly controllable for fibers provided by nature. This is
in contrast to well-established glass or carbon fibers produced industrially.

Moreover, they are less resistant to specific impacts like fire or water. In fact, natural
fibers are likely to burn, which is once again a drawback compared to some of the artifi-
cial materials. Another issue is water absorption. The fibers mentioned here-above tend
to be hydrophilic because of their high cellulose content. This means that swelling occurs.
Even if it is reported that swelling may have a positive impact on mechanical properties,
dilatation of the fibers due to moisture can induce microcracks in the contact area of the
matrix with the fiber [17, 16].

Another concern is the fact that the adhesion between fiber and resin is poor compared to
classical composites. This is also a consequence of the hydrophilic character of the natu-
ral material with hydrophobic resins. One possible solution is a surface treatment of the
fiber that enhances compatibility. Furthermore, the transport of water molecules inside
the fibers can be prevented by such a treatment. This is done by either modification of the
surface using sophisticate physical techniques or by application of molecules with inter-
mediate properties between fibers and the matrix [9]. More details about those techniques
would go beyond the scope of this work.

2.2 Architecture of highly crosslinked polymers

2.2.1 Molecular network
A curing process is essential to form a polymer corresponding to a solid network that
can be implemented as a structural material. Basically, this process is a polymerization
during which the low molecular weight monomers connect to each other to form a 3D
network with high molecular weight. The interconnecting bonds are known as crosslinks.
Needless to say, a sufficient high density of crosslinks is necessary for solid mechanical
properties.

Two different molecules with different functional groups have to be mentioned to un-
derstand how the molecular network is formed: the epoxy molecules and the hardener
[18].

2.2.2 Epoxy molecules

The epoxy molecule is the monomer of the thermoset, a general representation is given in
figure 2.2. The functional group of an epoxy is a cyclic ether, also known as oxirane group.
A high reactivity is provided by the fact that only three atoms form this ring, the valence
Chapter 2. State of the art 6

angles have therefore an amplitude of 60◦ (by assuming an equilateral triangle) instead
of 109,5◦ which is energetically favored. This induces a certain tension that may break up
the bonds of the ring quite easily. Moreover, because of a difference in electronegativity
between oxygen and carbon atoms, the latter ones can react with nucleophiles.

F IGURE 2.2: General representation of an epoxy molecule. From: [19]

2.2.3 Hardeners
Their role is to initiate a step-polymerization that forms the final 3D network. Various
types of hardeners exist and they are not necessarily specific to a given type of epoxy
molecule. Most commonly, the hardener is a nucleophilic amine or polyamine. Polyamides,
phenolic resins, anhydrides isocyanates and polymercaptans can also be used [18].

The choice of the hardener has, like the epoxy, a direct impact on mechanical properties as
well as on physico-chemical characteristics like the curing kinetics or the glass transition
temperature. It should also be mentioned that in some processes, only the epoxy molecule
and no hardener is used. They are called single-component resins, whereas systems cured
with the help of a hardener are two-component resins [18, 20, 21].

2.3 Formation of epoxy-based polymers with hardeners

2.3.1 Reaction mechanism
To understand how a network is formed out of these two components, it is interesting
to look on the mechanism of the reaction. When, for example, a diamine is used as a
hardener, a so-called cationic ring-opening polymerization reaction occurs [22]: Due to
the high reactivity of epoxies with nucleophilic agents, explained here above, an amine
group can attack a carbon atom of the ring. The ring breaks and a covalent bond between
epoxy and hardener is formed. The oxygen atom of the former epoxy captures a hydro-
gen atom of the amine to end up as a secondary alcohol group (see figure 2.3) [23].

In the case of a hardener containing two diamine groups, a total of four epoxies can be
linked, as shown on the left of figure 2.4. For the fabrication of thermosets, epoxies with
Chapter 2. State of the art 7

F IGURE 2.3: Left: an amine group of the hardener (circled in red) attacks
the carbon atom of the epoxy. A bond between both is formed. Afterwards,
the oxygen atom captures a hydrogen atom to become a hydroxyl. From:
[23]

two oxirane groups are used. Every (primary) amine group becomes consecutively a sec-
ondary and tertiary group while the epoxy monomer is consumed. A huge crosslinked
network, represented on the right of figure 2.4 is formed when time is sufficient.

F IGURE 2.4: Left: a single hardener molecule can link four epoxy
molecules. Right: Representation of a highly crosslinked network. From:
[23]

To get a high crosslinking degree, it is crucial to respect a given epoxy-to-hardener ratio.

If there are too many epoxy molecules, some oxirane groups will not react and molecular
chains will not be crosslinked together. On the other hand, an excess of hardener will also
disturb the formation of the network by the presence of useless amine groups that do not
connect molecules to each other. Therefore, the right proportion has to be known when
thermosets are prepared.
Chapter 2. State of the art 8

Translated into equations, the network formation becomes:

R − E + R 0 − A1 → R 0 − A2 − R
R − E + R0 − A2 − R → R0 − A3 −( R)2
With E the epoxy group and A1 , A2 , A3 primary, secondary and tertiary amino groups
respectively.

The mass balances can be written as follows, using brackets to indicate concentration
and the index 0 to specify the beginning of the reaction [18]:

• [ A1 ] = [ A1 ]0 − [ A2 ] − [ A3 ]

• [ E ] = [ E ]0 − [ A2 ] − 2[ A3 ]

Some reactions also involve a catalyst, but they are not within the scope of this work. It is
also important to mention that other reactions, e.g. etherification occur. As the etherifica-
tion is neglected in many sources, they are not discussed further [24, 20].

2.3.2 Curing models

The crosslinking reactions described in section 2.2.1 need a given amount of energy to
take place. Some curing processes use UV light and an appropriated catalyst as source of
energy but most often it is provided by heat. Different temperature profiles, changing the
heat flux over a given range of time, are known as curing cycles. Depending on the epoxy
and hardener as well as on the application, another curing cycle may be selected. This
means that there is no "universal" cycle that is the best choice in every situation. It is thus
necessary to understand the influence of temperature on reaction kinetics. It is important
to say that the reaction is exothermic, the source of heat is not only external but also due
to the reaction itself [24].

Several approaches are found in literature. As the aim of this work is not to cite them
all, only a brief review is given. Basically, two categories can be distinguished: Phe-
nomenological and mechanistic models [25]. Latter ones adapt the kinetics of free radical
polymerization to the curing of thermosets but they are not discussed in this work. At
the beginning of the 21th century, so-called model-free kinetics (also known as isconver-
sional methods) appeared [25, 26, 27]. Curing kinetics of bio-sourced epoxy resins have
also been reported [28], but as expected, no difference is made between petrol and bio-
sourced resins.

Phenomenological models

This first approach generalizes the well-known Sestak-Berggreen equation [24, 25, 27]:

dα
= (k1 + k2 αm )(1 − α)n (2.1)
dt

With α the curing degree that is defined below, n and m the reaction exponent and k1,2 ( T )
Chapter 2. State of the art 9

the rate constants that can be assimilated by an Arrhenius law:

− Ea
k ( T ) = A exp (2.2)
RT

With A the pre-exponential factor, Ea the activation energy and R the gas constant. As
already mentioned, α is the curing degree (or extend) of the reaction. There are several
ways to define this parameter. As differential scanning calorimetry (DSC) is mainly used
to characterize curing kinetics, α can be defined in the following way [25]:

Q
α= (2.3)
Qu

With Q the heat amount generated by the reaction and Qu the ultimate heat as total area
under the scanning cure thermogram. By using the fraction of epoxy groups that have
reacted, α can be defined as well:

[ E0 ] − [ E] dα −1 d [ E ]
α= ⇔ = (2.4)
[ E0 ] dt E0 dt

Equation 2.4 is rather related to the mechanistic approach, but shows clearly that crosslink-
ing is synonym of a decrease of the concentration of epoxy groups with time.

Note that equation 2.1 reminds a rate function of a single-step, n-order reaction if m = 0.
The reaction is however too complex to assume that m = 0. In fact, the equation would
suggest that the rate reaches its maximum at the beginning of the reaction. However, this
is not the case as the formation is autocatalytic because of formation of hydroxyl groups.
The presence of the parameter m can therefore be related to this autocatalytical mecha-
nisms [24, 25, 27]. By adding the exponent m, the maximum rate is situated at α = mm+n
which is typically around 30% to 40% [25].

To be more general, equation 2.1 can be written in the following way [26, 28]:

dα
= k( T ) f (α) (2.5)
dt
The function f (α) is the conversion function. The parameters A, Ea and f (α) are called
the "kinetic triple" [27].

Moreover, to take into account non-isothermal conditions, the following modification can
be done with β expressing the heating rate [28]:

dα dα
=β (2.6)
dt dT
Chapter 2. State of the art 10

Not only the crosslinking reaction itself, also the diffusion of the molecules inside the sam-
ple have to be taken into account to obtain highly crosslinked networks. This is especially
true by approaching the solid state, i.e. at higher curing degrees. Therefore, a correction
of the rate constant is necessary. This is done by adding a rate constant k d that is related
to diffusion. The constant k c is related to the chemical reaction [25]:

k −1 = k − 1 −1
e + kc (2.7)

Model-free kinetic method

Establishing a meaningful kinetic function f (α) introduced in eq. 2.5 is in practice im-
possible. To overcome this limitation, the isoconversional or model-free kinetic methods
have been developed. Their aim is to evaluate the activation energy Ea in function of
temperature by assuming that at a constant degree of conversion, the curing rate only
depends on the temperature [26, 28, 29]. Combining equations 2.2 and 2.5 and by taking
the derivative, one can obtain the following equation at a fixed degree of conversion [28]:

∂ ln (dα/dt) ∂ ln k ( T ) ∂ ln f (α)
−
|α = −
|α + |α (2.8)
∂T 1 ∂T 1 ∂T −1
As the kinetic function f (α) is not dependent of the temperature, the relation can be sim-
plified [28]:
∂ ln (dα/dt) − Ea
− 1
|α = (2.9)
∂T R
This relation can be solved by various numerical methods.

Taking all these concepts together, the curves in figure 2.5 can be established. They show
that a higher temperature leads to a smaller time that has to be reached to achieve a higher
conversion. After a certain time, the rate becomes smaller and the degree of conversion
arrives at a saturation.

2.4 Influence of curing on physico-chemical properties

2.4.1 Glass transition temperature
During a curing process, two phase transformations can be observed: A first one occurs
when the liquid mixture of epoxy monomers and the hardener transform into a visco-
elastic rubber, known as gelation. This transformation is irreversible. A second one, the
vitrifaction, is due to the transformation of the rubber into a glassy polymer (or vice-versa
as the transformation is reversible).

The glass transition temperature Tg is related to the latter transformation and is reported
as the most important thermal property of a thermoset polymer. At this temperature,
there is a change from glassy to rubbery state due to a sudden increase of chain mobil-
ity. At this rubbery state, the polymer becomes a kind of gel. The transition is marked
by changes of many properties like thermal expansion, Young’s modulus, heat capacity,...
Chapter 2. State of the art 11

F IGURE 2.5: The curing degree α in function of time in different isothermal

conditions. Adapted from [24].

Intuitively, one can state that with increasing degree of curing, the glass transition temper-
ature increases as well because a high density of crosslinks prevents chains from motion,
reducing their mobility and favoring therefore the glassy state. The Tg of a polymer can
thus be seen as a structural parameter [30].

It is also worth to mention that the cooling rate from rubbery to solid state can have
an impact on Tg [31]. This can be seen in figure 2.8 where two different cooling rates are
shown. The slower the cooling rate, the lower Tg .

The nature of the molecules used as epoxy monomers and curing agents plays also a
role in determination of the glass transition temperature. Stiffer chains or electrostatic
interactions between chains increase Tg [30].

A powerful tool to evaluate the influence of time and temperature on Tg (as well as the
curing degree by following the explanation here above) is the time-temperature-transition
diagram (TTT-diagram) which is specific to each epoxy/hardener system. An example of
such a TTT-diagram is given on figure 2.6 for isothermal conditions. The y-axis represents
Tcure , being the temperature at which a given resin is cured. Three different temperatures
are marked [30]:

• T0 : the "freezing" temperature of the unreacted resin. Below this temperature, the
reaction would occur in a frozen state and is thus really slow, even impossible.

• Tg∞ : the glass transition temperature of a resin that is completely cured, meaning
that α is nearly 100%
Chapter 2. State of the art 12

• Tgel : the temperature where no transition regime between gelation and vitrification
can be observed. The gelation point is defined as the intersection between gelation
and vitrification curves and has an associated gel time t gel .

F IGURE 2.6: TTT-diagram for a thermoset polymer. From: [32]

If the curing temperature Tcure is set between Tgel and Tg∞ , the system undergoes gelation
and vitrification afterwards. The TTT-diagram shows also the "full cure" curve (dashed
line), where Tg = Tg∞ . It is quite intuitive and in good agreement with figure 2.5 that a
higher Tcure requires less time to reach the full conversion. If the curing temperature is
sufficiently below Tg∞ , the plateau in fig. 2.5 at which reaction rate becomes zero is low
because the diffusion limits the overall reaction as the viscosity increases due to network
formation [33]. It may also occur that the polymer is quenched before achieving complete
cure so that the reaction is stopped. Both cases lead to an intermediate curing degree.

If Tcure is set equal or just below Tg∞ , vitrification occurs and full cure quickly afterwards.
There is no further reaction possible at this stage. When the temperature is maintained
(or even raised), the sample returns into a gel state.

A phenomena called devitrification can happen. This phenomena is related to the ther-
mal degradation of the network. It results that Tg decreases and mechanical properties
are lowered. Tcure must not exceed a temperature at which Tg∞ is reached and should not
stay for a too long time at a temperature at high temperatures [30, 34].

There is another curve above the devitrification that marks a full degradation of the poly-
mer. In fact, the sample would rather become a piece of coal than a cured polymer when
this line is reached. During the preparation of samples, it has to be avoided to stay on an
increased temperature level for a certain time [30].

The correlation between degree of cure and glass transition temperature can be expressed
Chapter 2. State of the art 13

by the Di Benedetto equation [35, 36]:

λα
Tg = Tg0 + ( Tg∞ − Tg0 ) (2.10)
1 − (1 − λ ) α

With λ a structural parameter that can be determined experimentally by adjusting or be

calculated using the heat capacities at zero and full conversion which goes beyond the
scope of this work [36]. The assumption is made that the curing path has no influence.
The equation shows clearly an increase in Tg with the crosslinking degree [35].

Many curing cycles are made of two phases: a first temperature rise followed by a sec-
ond one at higher temperature, known as post-cure (PC). The idea of the PC is to enhance
chain mobility by increasing temperature so that chains may react even if the reaction was
controlled by diffusion at a lower temperature. Nevertheless, according to the explana-
tion in this section, the temperature should not exceed a given maximum Tg [30, 34].

It is reported that Tg∞ decreases with increasing amount of bio-based molecules [37].

2.4.2 Free volume and its influence on aging and rejuvenation

To understand the phenomenons of aging and rejuvenation, the notion of free volume has
to be introduced firstly.

F IGURE 2.7: Molecular view of free volume build up. From: [38]

It is known that polymers are "living" materials. Polymers in glassy state (below Tg ) are in
fact not in a thermodynamic equilibrium1 . The chains have still a certain mobility because
of the presence of little cavities in those polymers. These cavities build the free volume
and its presence is unavoidable. The degree of packing is complementary to free vol-
ume [31]. Figure 2.7 shows the situation in a graphical way. The close-packed, crystalline
configuration minimizes free energy and free volume, whereas the amorphous structure
1 unless the cooling after curing is really slow, which is not achieved in practice
Chapter 2. State of the art 14

represents cavities. To minimize free energy, the chains tend to reach the equilibrium with
evolution of time. By doing so, their packing is increased and free volume reduced.

The reduction of free volume is known as aging. This process is a kind of annealing
where the temperature of the polymer is held below Tg . The annealing can be enhanced
by increasing the temperature. Figure 2.8 shows the macroscopic evolution of the volume
of a polymer that is cooled from above Tg [38, 39]. In fact, some authors believe that free
volume is annihilated as it is diffused to the surface of a sample [39].

It is important to note that free volume is not related to Brownian motion, as at 0K the
total volume of a sample is composed of the solid volume occupied by matter and free
volume that is non-occupied. Thus, the free-volume exists at 0K as well [38].

F IGURE 2.8: Specific volume in function of temperature during a cooling

process. Blue lines show the cooling paths, the gray surface represents the
amount of free volume. The graph states that for two cooling rates, the
faster one (q1 that is quicker than q2 ) yields more free volume (point A).
Moreover, as already discussed, Tg is different for both cases. When the
sample is aged, free volume is further decreased until point D is reached. It
is also shown that equilibrium (point D) will never be achieved. From: [38]

The opposite of aging is rejuvenation. In this process, the amount of free volume is in-
creased. This can be done by quenching the polymer from a temperature above Tg so
that the chains have nearly no time to arrange during cooling. There is thus more free
volume trapped inside the sample. In fact, passing above Tg erases the thermal history of
the sample, this is why rejuvenation is sometimes called erasure [38].

Thermal history is not the only way free-volume is affected. Mechanically, polymers can
Chapter 2. State of the art 15

be aged/rejuvenated as well. The principle remains the same. By inducing plastic de-
formation in compression, by rolling for example, the chains begin to move. Compact
packaging is thus destroyed and more free volume can be found [40].

In this section, it is assumed that an increase in chain mobility occurs due to a higher
amount of free volume. However, several articles reported major doubts on the validity
of the free volume theory. It can therefore rather be seen as a pedagogic model that allows
a good visualization. For more information about those doubts, the reader is invited to
consult Chevalier’s thesis (2018) [41] that reviews some of those papers.

2.4.3 Shrinkage
It is known that shrinkage is less expressed for thermosets compared to thermoplastics.
Nevertheless, as epoxy resins are very adhesive to the mold or to a fiber in the case of
composites, defects can appear as the polymer sticks on the surface of the mold/fiber
while other parts of the resin still undergo displacement, resulting in residual stresses. It
is reported in literature that those stresses are responsible for warpage, void formation or
cracks. It is thus interesting to have a look on this phenomena to prevent failings in the
curing process [42].

Shrinkage can have two different origins in the case of thermosets cured from a liquid
resin:

• Chemical shrinkage: As already introduced in the previous section, the liquid resin
undergoes gelation and vitrification afterwards. These variations are accompanied
by variations in specific volume. The resin looses therefore volume when the com-
pact 3D-network is formed. This can be imagined by the fact that network forma-
tion is synonym of covalent bond formation between epoxies and hardeners. The
distance between atoms in the case of covalent bonds is much shorter than the dis-
tance between the atoms in the case of weaker electrostatic interactions. The onset
of shrinkage is defined by the gelation curve and shrinkage stops once the reaction
stagnates, i.e. the plateau at which the rate of reaction is zero2 [42, 43].

• Thermal shrinkage: In contrast to the chemical shrinkage, this phenomena is only

related to the mechanical dilatation/contraction of the polymer due to variations
in temperature, like any other material. In particular, the cooling step from a high
curing temperature to ambient temperature in manufacturing of FRPs can be critical
due to a mismatch of thermal expansion coefficients between fibers and resins [43,
44, 45].

2 This definition is in a certain sense misleading as the reduction of specific volume begins with the start

of the reaction, but at low curing degrees, the shrinkage does not lead to residual stresses.
Chapter 2. State of the art 16

2.5 Mechanical behavior of epoxy resins

This section aims to give a general overview about the mechanical response of a glassy
polymer below its glass transition temperature Tg . To understand the macromechanical
response of a glassy polymer below its Tg , several consecutive steps have to be discussed.
Morelle (2015) [46] described the stress-strain curve using 6 phases in figure 2.9. The idea
of following section 2.6.1 is to build up a response based on the behavior of the molecular
chains to investigate its physical origins.

F IGURE 2.9: Stress-strain curve for a glassy polymer in a uni-directional

compression test (green curve), with consecutive unloading (red curve) and
strain recovery (orange line). Source: [46]

1. Elastic-viscoelastic region
The response of the material at small strains is characterized by a linear dependency be-
tween stress and strain. The slope, called Young’s modulus (E), is usually between 1 and
4 GPa. At this state, the deformation remains elastic, the sample comes back into its initial
state [46].

Richeton [47, 48] predicts a rate dependency of the modulus. For a given strain rate ė,
one can write:  
re f ė
E=E 1 + s log (2.11)
ė0

With Ere f the modulus at the reference strain rate ė0 . This means that Young’s modulus
increases by a constant s par decade of strain rate [47].
Chapter 2. State of the art 17

2. Yield transient
Several possibilities can be used to describe the location of the yield point. In this work,
the yield is considered as the local maximum of true stress. The yield transient is influ-
enced by strain rate, temperature and hydrostatic pressure [49].

There are also other, time and temperature dependent, phenomena having an influence
on the distinctness of the peak, which are already introduced in section 2.4.2, namely ag-
ing and rejuvenation, without explaining the influence on the macroscopic response [40].
As already stated, increased age means that chains are hindered from motion because
free volume is reduced for polymers below Tg . It results that the molecular flow requires
higher stresses to occur, meaning that the peak of the yield transient is higher. The oppo-
site is true for rejuvenation: the peak is lowered as lower stresses are necessary [38, 40].

An increase of temperature lowers the yield point due to thermal activation processes
(see section 2.6.2). When strain rate is increased, it can be shown that the peak is higher,
due to a delayed response of the network.

Combining the influence of aging, temperature and strain-rate, the following figures 2.10
can be obtained:

F IGURE 2.10: a) Influence of thermal history, adapted from [40] b) Influence

of temperature and strain rate. From: [41]

3. Strain softening
During strain softening, a decrease of stress with increasing strain can be observed. The
result is a drop. Strain softening is especially pronounced in compression and shear tests
and a bit less in tension [50]. In compression tests, the local minimum in true stress after
yield transient, also known as lower yield point, is influenced by aging as well but the
influence is considerably less expressed as for the upper yield point [38]. Strain rate rises
the stress at lower yield point by the same amount as the upper one, the drop remains
therefore unchanged by the strain rate [51].
Chapter 2. State of the art 18

The real physical origin of strain softening is not fully understood and still a subject of
research [52].

4. Strain hardening
As explained in section 2.6.1, the strain hardening is related to the network resistance3 .
The strain hardening is not influenced by the strain rate. Temperature has an influence
that can not be neglected. The hardening modulus GR quantifies the slope in this regime
and is directly proportional to crosslinking density. This modulus remains however dif-
ficult to compute by calculating the slope of true stress with true strain due to the non-
linearity of the curves. Therefore, it is more convenient to calculate the hardening modu-
lus as the increase of true stress with the Neo-Hookean strain, defined as follows [53]:

e NH = −(λ2 − 1/λ) (2.12)

With λ = l/l0 the draw ratio, measured as fraction of current sample length over the
initial one. For high strains, a linearity is obtained between true stress and Neo-Hookean
strain, allowing an interpretation of GR [53]. An advanced physical understanding of GR
is given in section 2.6.1.

5. Back-stress during unloading

In contrast to the linear unloading observed in other types of materials, glassy polymers
have important non-linear unloading behavior. This is known as the Bauschinger effect.
Without entering too much into details, one can state that there is still a given amount
of strain that persists but the unloading benefits from mechanisms that facilitate reversed
deformation, leading to non-linearity [41].

6. Strain recovery under unloading

Some amounts of the persisting strains inside the material is recovered with time. This
mechanism is favored when temperature is increased [46]. When a sample is exposed to
a higher temperature so that gelation occurs, plastic deformation is erased, meaning that
they are completely reversible [52].

3 Sometimes, this process is also called strain re-hardening to underline that it is a second increase in stress
Chapter 2. State of the art 19

2.6 Micromechanical approach

2.6.1 Spring-dashpot analogy
One approach to visualize and understand the mechanics of highly crosslinked networks
is to do an analogy using a springdashpot system. This section relies mostly on the work
of Haward and Thackray (1968) [54]. Boyce (1988) [49] and Argon (1973) [55] completed this
work by describing rate and temperature influences. Their articles describe the behavior
of a 1D tensile test for glassy polymers below their glass transition temperature. Many
sources rely on the characterization of thermoplastic polymers. Below Tg , the difference
between both is that connections between chains are entanglements for thermoplastics
and covalent crosslinks for thermosets. Nevertheless, it can be shown that most results
are valid for both types. Nevertheless, the fracture mechanics between both are different,
as it is explained in section 2.6.3 [41].

F IGURE 2.11: Spring-dashpot model of a glassy polymer network. Source:

[56]

The model involves three components and reminds Maxwell’s model, shown in figure
2.11:

1. A Hookean (and thus linear) spring on a first branch

2. A viscous dashpot in connected in series with the spring on branch A

3. A non-linear spring on a parallel branch B

Branch A models the intermolecular barrier. At small strains, its spring stretches. Phys-
ically speaking, the intermolecular distance increases. The force to overcome is due to
Chapter 2. State of the art 20

Van Der Waal’s interaction between the chains (so-called secondary interactions). These
interactions are completely recoverable. If the strain and the energy stored in the spring
on branch A is sufficient, the dashpot is activated. This results in a plastic flow of the
molecular chains that becomes constant at higher strains. A peak is pronounced because
a certain energy barrier must be overpassed. The deformation begins to be irreversible.

The dashpot is also realted to the rate-depandence of strain-stress curves: lower rates lead
to a earlier activation of the dashpot, the stress at yield transient is thus lower. A concept
from rheology, the Deborah Number, can be used to visualize the influence of strain rates:

λr
NDe = (2.13)
ts
With ts the time scale of the manipulation and λr the characteristic time of chain relax-
ation that is high for polymers with low mobility and low for fluids. When NDe is high,
the sample behaves like a solid. This is the case for shorter time scales i.e. higher rates [31].

The spring on the branch B is added to simulate the resistance of the chain alignment due
to the network. When alignment occurs, the chains are stretched and the confirmation
changes. The origin of the resistance is therefore entropic. Mainly primary interactions,
due to covalent bonds between crosslink points, have to be taken into account. The fact
that this spring is non-linear leads to a small resistance at small strain levels. At low
strains, the response is related to motions between chains rather than to the stretch of
segments. When the strain increases, its contribution becomes more and more important.
The stress inside this spring can be written as follows:

σh ∼ ρe k B T (λ2 − λ−1 ) = GR e NH (2.14)

Where ρe is the entanglement density. The hardening modulus introduced in the previous
section is directly proportional to this entanglement density as well as to temperature. An
increase of ρe leads to more covalent bonds inside the system, which means that one gets
a higher stiffness [41, 52].

One may notice that an analogy can be made between the plastic flow of the chains in-
duced by stress in this model and the flow of the chains when temperatures is higher than
Tg . Both phenomena are based on the increase of mobility of the chains.

The spring-dashpot analogy is useful to develop further constitutive models that can be
used to predict the behavior of the material.

2.6.2 Molecular flow theories

Several models can be found in literature that describe how the intermolecular interac-
tions are built up and which role they play in the explanation of yielding. This work does
not aim to review all models, even if it is assumed that they play all a role in the big
picture of micromechanics [46].
Chapter 2. State of the art 21

Eyring model
Earlier in this work, the flow (or high mobility) of chains has been mentioned without
entering in details. This section serves to highlight this phenomena.

Based on his theory in section 2.6.1, Haward (1986) [54] included the model of viscosity
developed by Eyring (1936) [57] to investigate the molecular flow. In fact, once the chains
begin to flow, the condition ∂σ
∂e = 0 is fulfilled. Indeed, this is the case of branch A of
figure 2.11 once the dashpot is activated which is the case when strain is high enough. It
is assumed that this flow is thermally activated, thermal vibration is the key of the phe-
nomena.

For further investigation, the notion of yield stress σY is helpful. As a reminder, this
stress is necessary to overcome the barrier before flow is achieved.

The Eyring equation for the uni-axially applied strain rate close to the yield point σY = σ
can be written as [48]:
−∆H σY V E
ė = e˙0 E exp sinh (2.15)
kT 2k B T
With k B Boltzmann’s constant, V E related to the apparent activation volume and e˙0 E a con-
stant pre-exponential strain rate. The parameter ∆H E can be seen as an energy barrier
that the chains have to overcome, leading to flow.

Eyring’s equation can be simplified by a linear form, using the assumption that
σY V E > 2k B T. Equation 2.15 becomes:

−(∆H − σY V LE )
ė = e˙0 LE exp (2.16)
kB T

With e˙0 LE = e˙0 E /2 and V LE = V E /2. The product between free volume and yield stress
σY V LE lowers the activation barrier to overcome.

It can be stated that the energy barrier is independent of temperature and strain rate:

e˙0 LE
∆H LE = σY V LE + kT ln (2.17)
ė

Argon model
Argon’s model aims to describe the intermolecular resistance to flow by disclination. In
fact, a given chain can be kinked under compression. It results in a tendency of rotation of
the chain (or, to be more precise, a chain segment), which is prevented by the neighboring
chains. This restriction is the origin of resistance in the Argon theory. Surrounding areas
of the chain remain elastic, build up of plasticity is therefore local. A visualization is given
in figure 2.12a where a segment tries to rotate against the resistance by parallel chains [49,
55].
Chapter 2. State of the art 22

F IGURE 2.12: a) Compression . b) Formation of cavities From: [55]

A connection can be done with free volume theory as the rotation of the segment induces
cavities. This is in agreement with the fact that plastic deformation induces an increase of
free volume which is illustrated in 2.12b [49, 55].

2.6.3 Localization phenomena

It is reported in literature that two different mechanisms lead to a localization of plasticity
in glassy polymers, causing failure of polymers. They both have a common origin: a
localized stress concentration that induces a plastic strain heterogeneity. There is a build-
up of hydrostatic stress that can have two different consequences, as discussed below
[51].

Crazing
If the hydrostatic stress is locally higher than a given threshold, micro-voids are formed.
This results in localized fracture by crazes, appearing perpendicular to the loading di-
rection, as the voids continue to grow with increasing stress. The crazes are bridged by
fibrils, typically 5 − 30nm in diameter, which break up once the stress becomes critical
and fracture occurs. The voids already appear in the elastic region. This kind of intrinsic
failure is not observed in uni-axial compression tests as positive triaxiality is required [50,
40, 52, 58].

Shear yielding
If voiding does not occur, shear bands are developed. A condition for the formation of
those shear bands is that strain softening has been reached. The post-yield behavior is
therefore important to understand intrinsic failure of polymer networks. On a macro-
scopic scale, the cross section is lowered when shear bands are formed and necking oc-
curs. In contrast to crazes, the shear bands are oriented at angles around 45◦ to the loading
direction. Generally, the volume is conserved during shear yielding and their form can
vary, depending on testing conditions, e.g. temperature, strain rate, thermo-mechanical
past of the sample as well as cross-linking density as all of these factors influence the be-
havior during yielding [41].
Chapter 2. State of the art 23

Figure 2.13 resumes the link between microscopical strain heterogeneity and the macro-
scopic deformation in the case of a ductile material. The four graphs are the material
response to an uni-axial tensile test. The y-axis is the macroscopic true stress whereas
the x-axis represents the draw ratio λ. Previously, λ has been defined as ratio between
current and initial length. In an equivalent way, using the assumption that the volume is
conserved during plastic deformation, this parameter can be visualized as ratio between
initial and current cross-section. At step (a), the point close to the center of the sample
undergoes strain-softening. The first microscopic shear bands are formed in this region,
resulting in a macroscopic neck. Further, the point in the middle undergoes hardening
(b). Due to this hardening, the reduced section in the neck does not break, even if the
stress becomes higher and higher. Mainly primary interactions play a role in the center
of the sample, whereas secondary interactions are still predominant next to the clamps. If
the material is ductile enough, the neck is propagated on the whole length of the sample,
as can be seen on panel (c). At the end (step d), the neck is propagated on the whole spec-
imen and primary interactions (as explained in section 2.6.1) induce a resistance against
the load until breakage. The entire sample is strain hardened [40, 51].

F IGURE 2.13: True stress in function of the dwell ratio for a uni-axial tensile
test. The panels show the formation of a neck that is diffused on the whole
length of the reduced section of the sample. From: [40]
Chapter 2. State of the art 24

Competition between crazing and shear yielding

The polymer network itself determines the threshold that distinguishes crazing and duc-
tile behavior. Higher crosslink density results in a higher threshold value for the hy-
drostatic stress in the deformation zone to overcome. Some studies on highly crosslinked
epoxy resins reveal that shear bands are predominant in those systems [46, 41, 51]. Crosslinks
prevent the formation of fibrils and crazing is hindered, an intermolecular flow is more
likely to happen than covalent bond breakage. Figure 2.14 shows the true strain ratio eR
in function of chain density νc (that is in direct correlation with the density of crosslinks
ρe defined in section 2.6.1). The value of eR is defined as follows:

lnλ
eR = (2.18)
lnλmax
with λmax representing the maximum extension of a chain. Following this discussion, the
true strain ratio is next to one due to the formation of fibrils that extend until breakage.

Clearly, two regimes can be distinguished: a first one with high values of eR tending
to fail by crazing with low values of νc and a second one with showing shear deformation
with higher chain density [51, 41, 58].

F IGURE 2.14: True strain ratio in function of chain density. The two regimes
can be distinguished. Higher crosslink density favors yielding [41]

As already mentioned, a crucial role that determines failure is given by the response af-
ter yielding, to be more precise the strain softening with consecutive hardening. A more
expressed drop in true stress between peak and minimum in the strain-softening zone
leads to higher pressure build-up. Polystyrene has a lower capacity to be hardened which
results in a brittle failure mode. The difference between both situations can be observed
in figure 2.15 [51, 40].
Chapter 2. State of the art 25

F IGURE 2.15: true stress - true strain curve of a 1D tensile test for a) Polycar-
bonate (PC): as a post-yield response, shear bands are formed and a lower-
ing in cross-section can be observed on a relative large area. b) Polystyrene
(PS): perpendicular to the loading direction, localized crazes appear, result-
ing in failure of the sample. From: [40]

2.7 Fracture of epoxy resins

This section focuses on how a sample breaks in the absence of any pre-cracks. The expla-
nations are brief as more advanced formulations can be found in the parallel work of Van
Innis (2020) [3].

First of all, it can be stated that the fracture mechanism differs significantly with the type
of stress test. In tensile tests, the samples are quite brittle. Especially at low temperatures,
the stress-strain curve is linear until fracture occurs at high stress. If the temperature in-
creases, without overpassing Tg , the behavior becomes more and more ductile. However,
in compression tests, the same material can be largely deformed in the plastic region. The
fracture is not brittle anymore. Because of this differences, both cases are treated sepa-
rately.

2.7.1 Brittle fracture

For a tensile stress, one may imagine that the fracture of a glassy polymer is related to the
cleavage of the covalent bonds inside the fracture plane. In the case of brittle materials,
it means that about 1/10 of Young’s modulus, i.e. about 300MPa of true stress is required
to induce fracture. However, experimental observations show that this value is never
reached. Models that reduce the predicted failure stress have been developed because of
this. The work of Griffith (1921) [59] suggests that defects (artificial or natural ones) are at
the origin of fracture at stresses well below the theoretical predictions based on primary
interactions.
Chapter 2. State of the art 26

Based on the previous assumption and in the frame of linear elastic fracture mechan-
ics (LEFM), Berry (1963) [60] defined the fracture surface energy (FSE) as the amount of
energy necessary to form two new surfaces at the inside of the material by crack propa-
gation. This surface energy is denoted as γ.

For a given crack or defect of size ac , the failure stress σF can be predicted by the fol-
lowing equation [41, 60]:
2Eγ
ac = (2.19)
π (1 − ν2 )σF2
With ν the Poisson ratio. As long as the size of the defect remains smaller than ac , the
growth of the crack remains slow. This regime is known as sub-critical crack growth.
Once the crack size is large enough and reaches ac , the crack propagates fast and fracture
occurs.

It can be assumed that fracture is related to visco-plastic flow next to the tip. Therefore,
a high stress at failure is linked high amounts of energy that can be dissipated by such a
flow. For PMMA, it has been shown that by lowering the temperature or by integration
of crosslinks, the FSE is lowered as well because the flow is hindered [41, 60].

Even if aging has been reported in this work to enhance some mechanical properties like
yield stress, it induces a more brittle fracture of the sample. The rearrangement of the
chains is slower, the stress can therefore not be reduced easily [31].

Fracture surfaces can be observed using a scanning electron microscope (SEM). As a gen-
eral rule for epoxy resins, Cantwell (1988) [61] distinguished four different zones in the
case of tensile specimens. A schematic representation is given in figure 2.16. First, a de-
fect can be described. In the case of the scheme it is situated on the edge, which is however
not always the case because defects can be located in the center of the sample as well. The
surroundings are marked by a smooth, mirror-like appearance. In this zone, crack prop-
agation is sub-critical and slow. Afterwards, secondary cracks can initiate at the tip of the
crack. This induces a parabolic pattern in the third zone. The main crack branches more
and more and a superposition of these secondary cracks is the reason why a rough zone
appears [61].
Chapter 2. State of the art 27

F IGURE 2.16: Four different zones of a fracture plane. From: [61]

2.7.2 Fracture in compression

Two different scenarios can be distinguished to describe damage in the case of compres-
sion tests [46]:

Firstly, the failure is purely related to yielding (2.17a). A small amount of cracks propa-
gate through the sample. They are oriented at 45◦ from the loading direction. The number
of fragments produced by this kind of fracture is quite small.

A second option is a combination between shear localization and brittle failure (2.17b).
A higher number of cracks propagate through the sample. Their nucleation starts at the
surfaces in contact with the plates and already begins at relatively small amounts of plas-
tic strain. The failure is sudden and a large number of fragments is produced [46].

F IGURE 2.17: Failure of compression specimens due to a) shear b) combi-

nation between shear and brittle behavior. From: [46]
Chapter 2. State of the art 28

2.8 Creep behavior

The resistance to creep is an important criterion in many material selection processes
for structural materials. Epoxies have, thanks to their visco-elastic properties, a time-
dependent behavior [62]. In this work only loads within the elastic regime are considered,
the initial response of the material is thus only elastic. They are referred as pre-yield creep
tests. Tests at different temperatures are popular to investigate creep. However, it must
to be taken into account that the yield stress decreases with increasing temperature.

2.8.1 Creep compliance

During a creep test, a material is submitted to a constant load that induces a given strain.
The aim of a creep test is to investigate this strain as a function of time. As it will be ex-
plained shortly, the testing temperature plays an important role in this behavior.

The material compliance in function of time, as defined as follows, is often given as a

property [63]:
e(t, T )
D ( t − t0 ) = (2.20)
σ0
Where t0 is the start of the creep test. The initial compliance is the inverse of the initial
Young’s modulus [63].

Often, a formalism based on relaxation with τv the characteristic relaxation time and β v
the stretching exponent is used to describe the evolution of the creep compliance with
time:  βv
t
D (t) = D0 exp (2.21)
τv

This equation is also known as KWW equation [38].

It is possible to predict the strain at a given time in terms of creep compliance when the
applied load is not constant. This approach is known as Boltzmann superposition princi-
ple and can indeed be helpful when complex loading cycles have to be studied [38]:
Z t
dσ 0
e(t) = D (t − t0 ) dt (2.22)
0 dt0

2.8.2 Creep strain

The creep strain is the contribution of three different strains. The elastic strain eel that
is governed by the initial and instantaneous Young’s modulus, the plastic strain e p that is
developed with time and eR that can be seen as an amount of strain that is recoverable.
One can thus write [64]:
e = eel + e p + eR (2.23)
Chapter 2. State of the art 29

F IGURE 2.18: Creep behavior of a material tested with constant load. Three
regions are distinguished: primary, secondary and tertiary creep. Adapted
from: [65]

As mentioned earlier, loads are applied below the yield stress of the material. The plastic
strain introduced here above is therefore not induced as an immediate response to stress.

Regardless the nature of the material investigated, the same type of curve is obtained
when the strain is plotted as a function of time as it can be seen on figure 2.18.

First of all, the submission to the load induces an immediate increase in strain, as it is
the case in uni-axial compression/tensile tests. Afterwards, three different regimes can
be observed [62]:

1. Primary creep
At the beginning of a creep test, the strain increment i.e. the slope of the curve, is
quite high.

2. Secondary (steady-state) creep

In this region, the strain increases proportionally to time. The creep exponent n can
be defined as follows:
ėss = Bσn (2.24)
Once again, this phenomena relies on the capacity of the chains to flow under the
stress. A better diffusion increases the slope on figure 2.18. By following the devel-
opments introduced in section 2.6.2, the following relation can be written [62]:

− Ec
ėss = C exp (2.25)
RT
The activation energy for creep is denoted as Ec . Combining equations 2.24 and
2.25, the following relation that is widely used to characterize the creep behavior of
Chapter 2. State of the art 30

a material can be deduced [62].

n
− Ec

σ
ėss = ė0 exp (2.26)
σ0 RT

The constants ė0 and σ0 are, like n and Ec , characteristic for the material. It is worth
to say that this relation is not specific for polymers, also metals and ceramics follow
this rule, even if their creep behavior is not governed by chain diffusion [62].

3. Tertiary creep
The creep rate increases at a given moment. This is due to an upcoming failure of
the sample because more and more damage is build up inside the material. Using
a simple equation, known as Monkman-Grant law, it is possible to estimate the time
that has to pass until rupture, denoted as tr [62]:

ėss tr = K (2.27)

The constant K is usually between 0.05 and 0.3. For any material where it is pos-
sible to estimate the steady-state strain increment with formula 2.26, it is possible
to estimate the time to failure even if the material has not been tested under those
conditions. This is of particular interest since creep tests require a lot of time [62].

The time to failure is the build up by two successive steps related to two differ-
ent times, namely ti that is the time required for crack initiation and t p the time for
its propagation, one can write [64]:

tr = ti + t p (2.28)

2.8.3 Superposition principle

Time-Temperature superposition
As already discussed, temperature has an important influence on the creep test. A su-
perposition that aims to develop a mastercurve can be developed to predict the creep
response of a material at a given temperature without that any test has been actually per-
formed under this condition by a time temperature superposition (TTS).

In practice, a test can be performed at a reference temperature TR . A log-log plot with

the creep compliance as a function of time can be established. Experimentally, one would
observe that the compliance at T1 > TR is, in this log-log plot, shifted horizontally from
the curve at TR . Beside this horizontal offset, the curves are identical. The shift can be
computed theoretically by the shift factor a T [38]:
  
ETTS 1 1
a T = exp − (2.29)
R T TR

With ETTS as an activation temperature that has to be determined experimentally. Nev-

ertheless, it has to be distinguished between the characteristics of the material below and
Chapter 2. State of the art 31

above Tg as mechanical properties undergo a sudden change. The discussion given in this
work is only valid for temperature below Tg , it is considered that the material is not used
in structures that can exceed the glassy state [38].

Time-Aging superposition
In a similar way, aging can also be predicted. The principle remains the same, but the
shift factor is of course defined in a different way:
τv
aE = (2.30)
τvR

As a reminder, τv has already been defined in equation 2.21 and is the characteristic relax-
ation time, the index R marks the reference test [38].

F IGURE 2.19: log - log plot for creep compliance against time of an epoxy.
The master curve is shown in blue (central curve). If the temperature for
the creep test is set above the reference of the master curve, a horizontal
shift to the left by a T is done (red curve). When the sample has been aged
for a given time, a horizontal shift to the right by a E can predict the creep
compliance (green curve). From: [38]
 32

Chapter 3

Scope of testing

The aim of this short section is to explain which tests have been performed in this work
and to clarify some choices. Afterwards, chapter 4 presents the materials and methods
used to carry out those tests which results are presented in chapter 5.

The results are complementary to the Master’s thesis of Van Innis (2020) [3]. The test-
ing of the bio-sourced resin is therefore split into two distinct Master’s thesis.

In the work of Van Innis, the resin is analyzed using dynamic tests. In particular, an
investigation of the fatigue behavior and the toughness can be found. The propagation of
cracks in SENB specimens is analyzed using a SEM. A constitutive model is developed as
well, based on the results in the following part of this present thesis. This model can be
used for finite element simulations. A cross-check between the experimental results using
a direct image correlation and the finite element model is performed.

In this work, mainly static tests are performed.

First, physico-chemical properties like Tg and α are obtained using a DSC analysis. These
results are presented firstly as they are used for better representations of the following
ones.

Uni-axial compression and tensile tests are performed to obtain the strain-stress response
of the material. In a first approach, a total of five different curing cycles are tested as the
selection of a cycle is within the objectives of this work. They are categorized into "low
degree" cycles and "high degree" cycles, depending on the results of the DSC analysis.

At the beginning of the mechanical results, the general true strain - true stress curves for
compression and tensile tests of the cycles are given for a better orientation. Afterwards,
more details about elasticity, yield transient, post-yield response as well as hardening fol-
lowed by failure are subject of the study. The results and discussions are done following
the chronology of typical stress-strain curves. As it is known for polymers that the me-
chanical response of a material is rate-dependent, different strain rates are imposed.

A fractography of tensile specimens is done for some representative specimens. A link

could be done between those observations and the results provided by the tensile tests.
Chapter 3. Scope of testing 33

In a similar way, compression specimens after testing are analyzed.

In a second approach, samples are cured with different post-cure times. Resulting partly
cured specimens are tested in compression. By doing so, specimens with four different Tg
are obtained and mechanical properties are given in function of its Tg . An evolution with
curing extend can be established. In fact, the first approach distinguishes only between
two categories, namely "low degree" and "high degree" cycles as unavoidable batch-to-
batch variations would give invalid results. As explained in the next chapter, those ex-
perimental errors could widely be avoided using the second approach.

Lastly, specimens provided by the same batch and cured in a similar way were inves-
tigated under uni-axial static compression over a period of several days to analysis their
creep behavior.
 34

Chapter 4

Materials and methods

4.1 Materials
4.1.1 Epoxy-hardener system
The resin
The epoxy resin studied in this work is the InfuGreen810 provided by the company SicominTM
(Chateauneuf les Martigues, France). The carbon of the epoxy molecules are to 38% bio-
sourced. This is the reason why the hole system can be denoted as bio-sourced. The resin
is designed in particular for transfer processes, especially injection and infusion.

According to the producer, the carbon footprint of this system is about 40% lower com-
pared to conventional petrol-based ones. No bio-degradability is reported [4].

The hardener
InfuGreen810 is a resin that has to be cured by a hardener. Several types of hardener can
be used. As the hardener can have a huge impact on the kinetics, it has to be chosen care-
fully. In the preparation of the samples in this work, it was decided to use the SD4771,
produced by SicominTM as well. It is categorized as a "ultra-slow" hardener. It is worth
to say that the molecules are not bio-sourced, meaning that the percentage of bio-sourced
carbon atoms in the cured end-product is lower than 38%.

The proportion between epoxy and hardener is 100 : 22, 5 in terms of mass (meaning
that for 100g of epoxy mixture, 29g of hardener have to be added).

Several properties are given in table 4.1. For more information, the reader is invited to
consult the datasheets [4, 66].

4.1.2 Physico-chemical properties of the resin

Some preliminary work on InfuGreen810 cured with SD4771 has been done by De Halleux
[67]. Thanks to his experiments, a majority of the physico-chemical properties and infor-
mation about the curing kinetics could be used to extend the datasheet.
Chapter 4. Materials and methods 35

InfuGreen 810 SD4771 Ultra-Slow

Bisphenol A diglycidyl ether (DGEBA) Etherdiamine
Composition Formaldehyde 4-methylcyclohexane-1,3-diamine
1,4 Butanediol diglycidyl ether 2-methylcyclohexane-1,3-diamine
Density (20◦ C) 1,16 0,944
Viscosity 1350 11
(20◦ C) [mPa·s]
Bio-Carbon 38% 0%

TABLE 4.1: Different properties of epoxy and hardener

Based on his physico-chemical experiments, De Halleux developed a TTT-diagram of the

resin, represented in figure 4.1.

The diagram contains neither the devitrification curve nor the full-cure curve but pro-
vides most importantly Tg0 = −59, 29◦ C, Tg∞ = 80, 54◦ C and Tgel = 49◦ C.

This diagram helped to develop the curing cycles, given in section 4.2.

F IGURE 4.1: TTT-diagram for the epoxy-hardener system investigated.

From [67]
Chapter 4. Materials and methods 36

4.1.3 Sample preparation

Several consecutive steps have to be performed to obtain specimens that can be tested
afterwards:

1. Mixture preparation
Epoxy and hardener are stored separately, otherwise a reaction would occur at room
temperature. Once the two components are poured together, typically in a beaker,
the reaction begins but as temperature is too low and time too short, no immediate
network formation can be observed. Therefore, the resin remains liquid. Constant
stirring has to be performed to obtain a homogeneous mixture.

2. Degasing
Afterwards, it is absolutely necessary to apply a vacuum to degase the resin. This
step is common for many resins. In fact, there is plenty of air dissolved in the liquid
mixture that has to be removed. Otherwise, air bubbles will be present in the final
product which is of course not desirable. The low pressure draws the air out of the
resin. This is done between 15 and 20 minutes, it can be decided by observing the
mixture with the naked eye if degasing is sufficient. The pump in use was a Pascal
2010 SD.

3. Molding
Then, the mixture is poured in the mold that is used for the curing process itself.
Glass test tubes (13mm inner diameter, 160mm height) have been chosen. Before,
the inner wall has been coated twice with a release agent (LoctiteTM Frekote 700-NC),
otherwise it would be difficult to remove the cured resin from the glass due to its
severe adhesion.

4. Curing1
Once the resin, still in a liquid state, is inside the tubes, the curing cycle can begin.
During the first 20 minutes, a second degasing step is done to make sure that no air
left trapped in the mixture. The different curing cycles are presented in section 4.2.
Curing was performed in a Heraeus oven (see figure 4.2) that allows the operator to
impose temperature and heating rates.

5. Machining
Once the resin is cured, they have to be removed out of the glass tubes (a hammer
may be useful to break the glass). Afterwards, machining is necessary to obtain the
testing specimens. Two different geometries are used:

• Compression specimens that are used for creep tests as well (see fig.4.2a). The
form is a cylinder with height and diameter equal to 12mm.
• Tensile specimens (see fig.4.2b). The diameter and length of the reduced section
are 6mm and 34mm respectively. The total length of such a specimen is 100mm.
1 Note: the post curing is included in this step
Chapter 4. Materials and methods 37

F IGURE 4.2: Testing geometries for a) compression and creep test b) tensile
test. Adapted from [46]

4.2 Presentation of Curing cycles

In the following part, the five curing cycles tested mechanically are presented.

A) 4h at 45◦ C + 4h at 80◦ C: During a first step, the liquid resin is heated at a rate of
0, 5◦ C/min. Then, the temperature is kept constant during 4h at 45◦ C on a first
plateau. A slope of again 0, 5◦ C/min leads to a plateau at 80◦ C, held during 4h.
This second plateau is the post-cure (PC). For ease of representation, this cycle is
denoted 45◦ C + PC.

B) 36h at room temperature + 4h at 80 ◦ C: Here, the liquid resin is kept at ambient tem-
perature (∼ 18◦ C) during 36h. The resin is afterwards post-cured similar to the
previous cycle, i.e. heated at 0, 5◦ C/min until 80◦ C is reached and held at this tem-
perature during 4h. In the following part, the name T amb + PC is attributed to this
profile.

C) Ambient temperature: This cycle does not involve any heating, the samples are just
kept at ambient temperature. In this work, the cycle is denoted by T amb.

D) 2h at 70◦ C + 2h at 120◦ C: This cycle was proposed by De Halleux [67] for a RTM
process, specifically for the InfuGreen 810 resin. During 1h, the resin is held at 30◦ C.
A fast heating (3◦ C/min) leads to a plateau at 70◦ C for 2h. A post-cure at 120◦ C is
Chapter 4. Materials and methods 38

performed, heating is again 3◦ C/min. The cycle is denoted as RTM as it is inspired

by other cycles suited for RTM-processes.

For all the cycles, the cooling rate at the end of the process is estimated to be approxi-
mately 2◦ C/min.

( A ) 45◦ C + PC ( B ) T amb + PC

( C ) T amb ( D ) RTM

( E ) 45◦ C w/o PC

F IGURE 4.3: Presentation of the different curing cycles

Chapter 4. Materials and methods 39

4.3 Differential scanning calorimetry

Output

Cured samples are subject of this study. As discussed in chapter 2.4.1, the glass tran-
sition temperature and the curing degree are important material properties and a DSC
analysis allows to access them.

Passing from the glassy to the rubbery state requires an input of energy. Therefore, this
process is associated with an endothermic (downwards) peak on a DSC thermogram. The
midpoint (between onset and minimum of this peak) is taken as Tg .

If the polymer is not fully cured, the crosslinking reaction continues once the rubbery
state is reached and as the chemical reaction is exothermic, an upwards peak is observ-
able.

As explained, the thermograms allow to compute the curing degree α. One can assume,
based on the TTT-diagram, that the degree of conversion is zero if Tg = −59◦ C (no re-
action below this temperature is possible for the InfuGreen810 system) and that full cure
is achieved if Tg = 80, 54◦ C. The Di Benedetto equation (eq.2.10) allows to establish the
relation between α and Tg . De Halleux predicted a linear relation. Therefore, one can write
the following eqution:

( Tg∞ − Tg0 )λα

Tg = Tg0 + = Tg0 + ( Tg∞ − Tg0 )α
1 − (1 − λ ) α

A second method to compute the curing degree, that is actually more common, consists in
measuring the residual enthalpies. De Halleux demonstrated a good agreement between
both method (for more details, see appendix A). It was opted for the first method that
is actually more precise in this case because only a few points could be used to perform
numerical integration, which induces a computational error.

Experimental device

A DSC contains two different samples that are placed simultaneously in a cell: the sample
itself, which properties are investigated and another one that serves as a reference. The
setup is shown in figure 4.4.

The cell is heated by a given temperature profile that can be chosen by the operator. Just
below the support, a thermocouple is placed to monitor the temperature of the material
itself. As mass and heat capacity of the reference are known, the heat flux, expressed
in [W/mgsample ] can be computed. When, for example, heat production occurs due to
an exothermic reaction of the sample, the temperature measured by the thermocouple
increases as well. By comparing sample and reference, a signal of the heat flux for the
sample can be obtained as a function of temperature [68, 69].
Chapter 4. Materials and methods 40

The machine used in our tests is a DSC810 furnished by Mettler Toledo.

F IGURE 4.4: Scheme of a DSC machine

Operational mode

Very small samples, typically between 5mg and 10mg, are used. The materials are trapped
inside a small aluminum support. A flux of inert gas, in our case N2 , sweeps above the
two materials to purge gases that are eventually formed during heating of the polymers
(Note that chemical reactions may occur as the resin is not fully cured). The temperature
range varied linearly between 0◦ C and 250◦ C and the heating rate was 10◦ C/min.

4.4 Mechanical testing

4.4.1 Nomenclature
A typical true strain - true stress curve is given in figure 4.5. Roman letters in fig.4.5 refer
to the steps described with more details in section 2.5.

Mean values are plotted in the result with the standard deviation for better interpretation.
They are defined as follows:

• Young’s modulus E calculated as the secant slope between 500 and 2500 µStrain as
suggested by ASTM-d368 standard [70]. Displacement rate is 1mm/min.

• Yield stress σy0.2% determined as intersection between the curve and a line with
slope E starting at 0, 2% of strain

• Peak yield stress σY , the local maximum after the elastic regime
Chapter 4. Materials and methods 41

F IGURE 4.5: Strain-stress curve for uni-axial loading. Adapted from: [46]

• Lower yield stress σl , the local minimum after peak yield

• Failure stress and strain at failure σ f and e f , taken as ultimate values of the test

Based on those definitions, other properties can be defined:

• The yield drop KY is defined as the ratio between mean values of upper and lower
true yield stresses at 1mm/min: KY = σY /σl . The higher this parameter, the higher
is the drop after yielding during strain softening.

• The hardening modulus GR . As a reminder of section 2.5, the increase of true strain
with the Neo-Hookean strain defines this modulus. Values presented in the fol-
lowing section are linear interpolations of e NH − σ curves between 90% and 95% of
failure stress.

4.4.2 Note on constant crosshead displacement vs. constant true strain rate
When static tests are performed, two different operational modes can be considered for
the deformation of the sample:

• Constant crosshead displacement (CCDR): when the test parameters are set using
the software of the machine, the deformation is expressed in [mm/min]. As the
name suggests, the crossheads move with constant velocity

• Constant true strain rate (CSR): this mode is more appropriate to measure rate-
dependency but involves a feedback based on the current gauge length during the
test. The rate is expressed in [s−1 ].

The difference between both in terms of evolution compared to constant strain rate is il-
lustrated in figure 4.6. In the case of the experiments which results are shown in the next
chapter, only CCDR measurements are taken.
Chapter 4. Materials and methods 42

F IGURE 4.6: True strain rate of CCDR vs. CSR for a static compression test.
From: [46]

4.4.3 Testing devices and operational modes

Uni-axial tensile tests
The tensile tests have been performed on a screw-driven Zwick and Roell machine with a
250kN external load cell (see figure 4.8a). The specimen geometry is represented in fig.
4.2. Scissor-extensometers are used to measure the strain.

Tests are done under CCDR with deformation rates of 1mm/min and 10mm/min.

A pre-charge of 3[ N ] is applied. The role of a pre-charge is to eliminate the "toe" of the

stress-strain curve that appears due to undesired artifacts like alignment or seating of the
specimen or to non-uniformity at take-up [46].

The tests are inspired by the ASTM-d368 test standards (see [70]).

Data originate from at least three valid tests per testing condition.

4.4.4 Uni-axial compression tests

The cyclindrical specimens in figure 4.2 are used to investigate the behavior in com-
pression, once again on a screw-driven Zwick and Roell machine. CCDR is applied with
0, 1mm/min, 1mm/min and 10mm/min of displacement rate.

The testing device has to be completed by a subpress by Wyoming instruments (figure 4.8c)
Chapter 4. Materials and methods 43

F IGURE 4.7: A specimen deformed during a compression test. Left: with

lubricant. Right: Without lubricant. Source: [46]

in which the specimen is jammed to ensure good alignment. The piston has a diameter of
1 inch = 2, 54cm.

A first way to collect the displacement is to use the displacement of the machine itself.
In this case, it is assumed that the stiffness of the machine parts is not infinite. In fact,
when load is applied by the machine, its metal parts also deform due to a given compli-
ance. For this reason, the so-called machine displacement is subtracted from the overall
displacement to get only the response by the material itself.

uspecimen = umeasured − umachine compliance (4.1)

This is done by compressing the device without any specimen inside the subpress. The
resulting curve is interpolated afterwards so that, for any force, the machine-compliance
displacement can be estimated.

Another option would be the use of Linear Variational Displacement Transducers (LVDT),
represented in fig. 4.8b. The system transforms the deformation of springs into an electri-
cal signal that is connected via a data system to the computer. In the device used in our
experiences, three of them are mounted around the specimen on a home-made device, the
mean of the signal can be computed and interpreted afterwards.

However, the amplitude of deformation is limited so that only small strains could be
investigated. As it is opted to take measurements up to high strains, their use is not rec-
ommendable.

Barreling is another phenomenon typical for compression tests that has to be avoided.
Especially for higher strains, friction occurs between the subpress and the sample. This is
the cause for barreling of the compression cylinder. To avoid this effect, a solid lubricant
in form of thin teflon papers (70m thin) is placed at the contact areas. For the same amount
of strain, the stress will be lowered when teflon papers are used [46]. The differences be-
tween both situations are shown in figure 4.7. Barreling provokes triaxiality that is not
desired in uni-axial compression tests. A graph showing differences in the stress-strain
response can be consulted in the appendix B. The pre-charge is 30kN. Still, a toe could be
observed. Values obtained by a linear regression of the elastic response (between 0 and
25 MPa) replace the part of the curve where a toe could be identified.
Chapter 4. Materials and methods 44

The results provide from at least three valid tests.

F IGURE 4.8: a) Screw-driven testing machine b) LVDT strain sensor c) Sub-

press d) Creep testing device

4.5 Creep tests

Output

Measurements during at least 7 days under constant load were taken. The three speci-
mens originated from the same batch. As no parallel testing was possible, about 2 months
separate each test in the order listed here-above. As the tests require a lot of time, only
one single specimen per loading condition was tested.

A thermometer provides eventual thermal fluctuations.

The output of the test is given as strain increment in function of time, meaning that the
immediate elastic response is removed. The strain at the origin of the graph is therefore
zero even if the material is already deformed elastically. The steady-state creep is given
as a linear interpolation of the strain increment of the last 10% of time.
Chapter 4. Materials and methods 45

Experimental device

The creep tests were not held on a machine like tensile or compression tests. Instead,
a mechanical system was employed that exerts a pressure thanks to weights that are at-
tached via a lever to the subpress. The only testing parameter that can be imposed by the
operator is therefore the mass of the weights. The hole system can be seen on figure 4.8d.

Operational mode

For those manipulations, it was opted to work under constant loads of 20MPa, 41MPa
and 49MPa and at room temperature.

4.6 Fractography
Output

Captures of tested specimens is the output in this case. A total of four tensile specimens
are tested. The chosen specimens are presented close to the results for better clearness.
For compression, only two specimens are observed as each of them is representative for
low and high degree cycles. All samples were tested previously at 1mm/min.

Experimental device

The fracture surfaces of tensile specimens were investigated with a scanning electron
microscope (SEM). This form of microscopy uses field emission gun (FEG) creating an
incident electron beam focused by magnetic lenses on the surface of the sample. An inter-
action between electrons of the beam and atoms creates secondary electrons with different
kinetic energies. They are detected and the signal emerging as a consequence is used to
reproduce the surface of the sample. As electrons of the incident beam can not penetrate
far into the material, only the topmost layers of the surface are scanned. The incident
beam is created by applying a voltage on a tungsten filament.

In the case of compression specimens, a simple camera captures is sufficient to take pic-
tures of the bullets after testing.

Operational mode

In the case of the SEM, 15kV are chosen as voltage on the tungsten filament. The sam-
ples have to be covered by a fine gold coating as the surface would not be conductive.
Chapter 4. Materials and methods 46

4.7 Incomplete cures

To obtain samples as a function of post-curing duration, tubes following the curing path
of 45◦ C + PC are extracted from the oven at given times. Those extractions are all made
on the plateau at 80◦ C. The first ones are taken from the oven immediately after reaching
the post-cure temperature of 80◦ C so that the curing is stopped. They are denoted ”0min”.
Other samples are extracted after 60min, 100min and 240min at 80◦ C (the final ones corre-
spond therefore to the entire cycle 45◦ C + PC), as visualized in figure 4.9. In this way, the
resin with four different post-curing times and different curing degrees are obtained.

F IGURE 4.9: Temperature paths for samples of the incomplete cure study

Afterwards, specimens were machined to cylinders and tested in compression in the same
way as described here-above. As this testing was done within two days, no effects due to
differences in age or even irregularities in the preparation of the mixture are relevant. At
least four samples per post-curing time are tested at 1mm/min and 10mm/min.

The determination of Tg was performed using DSC analysis in the case of samples ex-
tracted after 0min and 240min. No accurate DSC measurements could be obtained in the
frame of this work for the samples after 60min and 100min.

They could nevertheless be predicted thanks to the work by De Halleux [67] (2019) with
a model-free kinetic method (see section 2.3.2), giving access to Tg as a function of time.
The isothermal curve at 80◦ C is retaken in appendix A.
 47

Chapter 5

Results

5.1 Difficulties in sample preparation

The preparation is, despite many attempts and different approaches, more difficult com-
pared to other epoxy resins like RTM6. In fact, the filled tubes contained defects, some of
them are presented in figure 5.1. With similar problems, silicone molds have been tested
as well. It can be seen that, even with degasing steps, the cured material contains larger
air inclusions. Sometimes, cracks have propagated inside the material. Another type of
defects are "chimneys" that developed on the tube walls.

Mainly the cycle RTM during which the temperature was the highest, develops large cav-
ities. A problem related with networks cured at ambient temperature is their brittleness.
In this case, demolding can lead to fracture of the specimens.

F IGURE 5.1: a) and b): Defects inside the tubes, respectively cracks and air
bubbles. c) Tensile specimen with defect d) Compression specimen with
chimney
Chapter 5. Results 48

5.2 Physico-chemical testing

The thermograms for the different cycles are given in figure 5.2.

F IGURE 5.2: Thermograms

The following table provides Tg that is directly obtained by the thermogram as well as
α, obtained using the linear relation (see section 4.3):

45◦ C + PC 45◦ C w/o PC RTM T amb T amb + PC

Tg [◦ C ] 73 54 67 51 71
α 95% 81% 90% 78% 93%

In the following part, the results are categorized in two: on the one hand the two cycles T
amb and 45◦ C w/o PC that are "low-degree" cycles, on the other hand the three remaining
ones that are labeled "high-degree" cycles.

5.3 Mechanical testing

The results for different strain rates are shown. The following color-code is used in the
graphs to distinguish them:
• 0, 1mm/min: 
• 1mm/min: ◦
• 10mm/min: ♦
As a reminder of section 4.4.2, only CCDR measurements are taken.

The true stress-strain curves for the five curing cycles are given in figure 5.3. Unfortu-
nately, no measurements in tension of 45◦ C w/o PC could be taken.
Chapter 5. Results 49

F IGURE 5.3: True strain - true stress curves for the different curing cycles at
1mm/min in compression (left) and tension (right)

5.3.1 Elastic behavior

Young’s modulus is given in fig. 5.4 for both uni-axial compression and tensile test.

Only values for a displacement rate of 1mm/min are shown. Higher displacement rates
lead to values that are sometimes a tenth of MPa higher or lower compared other rates.

The x-axis shows the different curing cycles introduced in section 4.2. The curing de-
grees, taken from the previous section, are also plotted in form of blue bars below the
data representing the modulus. The results for each cycle originate from the same batch,

F IGURE 5.4: Elastic modulus in tension and compression, tested at room

temperature.

except for the compression measurements of 45◦ C + PC and RTM. Batch variability for
the former is about 7% and negligible for the latter one.
Chapter 5. Results 50

5.3.2 Yield and post-yield behavior

This subsection investigates the yield behavior for the different cycles. Unlike the results
for the elastic behavior, compression and tension results are shown separately.

Compression
The yield stresses at peak (upper) and the lower stresses defined as minimum after drop
(lower values) are shown in fig.5.5a).

One can observe the different deformation rates on the graph (for the T amb and 45C
+ PC cycles).

F IGURE 5.5: Yield values in function of curing cycles  0, 1, ◦ 1, ♦

10mm/min. a) σY and σl in compression. b) σY and σy0,2% in tension

Table 5.1 shows values for Ky and GR (see section 4.4.1). Note that instabilities at the end
of compression tests lead to high variations within values of GR . Therefore, outlayers are
not retained in the table.

at [1mm/min] T amb 45◦ C w/o PC RTM T amb + PC 45◦ C + PC

Ky [ / ] 1,83 1,77 1,27 1,35 1,36
GR [ MPa] 44 44 104 78 80
Tg [◦ C ] 51 54 67 71 73

TABLE 5.1: Ky is the ratio σY /σl , GR the hardening modulus, values of Tg

retaken from section 5.2

Tension
The representation compared to compression is slightly different: as no real (re-)hardening
occurs, a value for the lower yield stress could not be defined and is therefore not shown.
Instead, as it is common in literature, the stresses at 0, 2% of plasticity are plotted below
Chapter 5. Results 51

the peak values in fig.5.5b)

Samples from batch T amb show no real peak as rupture happens before reaching it. There-
fore, only the value at 0, 2% of plastic strain is shown.

5.3.3 Failure
The true failure stresses and strains at failure are shown. Both compression and tension
results are presented on the same graphs (see 5.6). Panel a) shows the ultimate failure
stresses and b) the true strains for the corresponding stresses. Note that on b), not the
mean but the median values are presented and the interval corresponds min/max values
of the measurement.

F IGURE 5.6:  0, 1, ◦ 1, ♦ 10mm/min a) True stress in compression and ten-

sion. Upper data are compression, lower ones tensile values. b) True strain
at failure for compression and tension. Yellow color labels compression,
green color tensile results

The outer parts of the testing cylinders with low α were squeezed out from the region
where the piston of the subpress exerts a pressure on the sample, comparable to a wax
seal on paper. With other words, the ductility for the samples is too high as their diameter
at the end of the tests exceeds the diameter of the piston of the subpress, without that
failure occured during this test. Those values have to be interpreted differently.

It is estimated that this phenomenon occurs at a true strain of about 1, 45[/], based on
volume conservation and the dimensions of the samples and the subpress. Therefore,
as no real physical interpretation could be done, no values exceeding this critical defor-
mation are presented. The stresses are more an ultimate value for the test than a real
estimation of the failure. At the end of the test, the specimens show cracks but one could
still apply slight flexion without that the flattened cylinder breaks.
Chapter 5. Results 52

5.4 Fractographic analysis

5.4.1 Tensile specimens
The four samples analyzed using a SEM provide from the following batches:

1. RTM cycle (fig.5.7 on page 53)

2. RTM cycle but a huge bubble (∼ 1mm in diameter) was visible on the edge even
before testing (fig.5.8 on page 54). The sample failed before reaching the peak yield.
Ultimate stress was 3, 01MPa.

3. Tamb cycle (fig.5.9 on page 55)

4. 45◦ C + PC cycle (fig.5.10 on page 56)

Samples in figures 5.8 and 5.9 show a defect followed by a mirror-like zone. This zone is
marked by smooth, laminar micro-cracks propagating radially from their onset. Interest-
ingly, an arc causes a deviation of those micro-cracks in the mirror zone for fig.5.8, where
the parabolic crack propagation begins about 100µm afterwards. For fig.5.9, this crack
deviation marks the end of the mirror zone. Nevertheless, most of the area is covered by
a rough relief.

In contrast, the surfaces of samples 5.7 and 5.10 are mostly smooth. Only the borders
show a three dimensional relief. Long cracks originating from the defects run up to the
borders of the sample. The cracks are, once again, slightly deviated.

The observations of the fracture surfaces are discussed in direct relation with data rel-
ative to fracture in section 6.5.
Chapter 5. Results 53

F IGURE 5.7: RTM cycle a) The overall appearance shows only slight brit-
tleness at the borders of the surface. Long cracks propagate linearly from
the defect to the borders. Those long cracks are accompanied by micro-
cracks, as it can be observed on panel b). Some of them propagate radially
whereas others seem to be initiated by perpendicular, scar-like extensions
of the main cracks (highlighted by the white arrow). A zoom on the defect
is showed on panel c). For orientation, an arrow is added pointing to the
void. Strong roughness is formed around the defect and the long cracks
seem to be built up there. Between rough cracks and the mirror zone, a
moisture-like zone that is a few tenth of µm in thickness, appears.
Chapter 5. Results 54

F IGURE 5.8: RTM cycle with defect A pre-existing defect was already vis-
ible before testing. The defect is in this case a huge bubble that can be dis-
tinguished without any zoom on a). The crack deviation inside the mirror
zone is visible on b). At the same level, the parabolic pattern begins. Panel
d) shows the inside of the bubble, no specific texture is visible. Note that
some spots (marked with the arrow on d) are probably dust and therefore
of no interest.
Chapter 5. Results 55

F IGURE 5.9: T amb Panel a) reveals that most of the area is covered by a
relief. On b), the defect with the laminar micro-cracks in the center is high-
lighted (the bright spot on this micrograph is a dust particle that should
be ignored), followed by a crack multiplication. This first crack propaga-
tion appears still in the mirror-like zone. As it can be deduced on panel
c), it seems that the growth of these cracks is stopped abruptly, upper and
lower levels appear. Therefore, regions that are not in-plane with the initial
propagation appear. Lastly, on the borders of capture d), the propagation
becomes rougher, a wavy pattern is visible following the horizontal line on
d).
Chapter 5. Results 56

F IGURE 5.10: Like 5.7, nearly no roughness can be observed. A few long
cracks propagate over several mm. The micro-flow is parallel with them
and crack deflection is visible for both on panel c). Panels b) and d) point
out the defect, probably due to decohesion between a small impurity and
the resin. At the edge between particle and fracture surface, cracks are
initiated.
Chapter 5. Results 57

5.4.2 Compression specimens

The two pictures in figure 5.11 show compression cylinders (initially 12 × 12mm) after the
tests. Sample in fig.5.11a) originates from a 45◦ C + PC batch, fig.5.11b) from a T amb one.

Many cracks propagate radially through the sample on fig.5.11a). The failure occurs sud-
denly leading to many fragments with rough surfaces.

No real failure occurs for the more ductile sample in 5.11b). As already explained here-
above, cracks propagate through the sample. The black arrow highlights the part that is
squeezed out of the loading area, the mark of the edge of the subpress is visible as well.

F IGURE 5.11: Two representative compression specimens after test. a) High

curing degree, showing many fragments. b) Low curing degree, showing
cracks but no real fragments as the sample did not undergo failure. The
white "wings" on a) and areas on b) are remaining teflon-papers.
Chapter 5. Results 58

5.5 Influence of post-cure on mechanical properties

The thermograms of samples post-cured during 0 and 240min are shown in fig. 5.12a).
Based on those thermograms as well as the model-free kinetic method, the four Tg s can
be computed (see section 4.7). They are given as a function of post-cure time in fig. 5.12b).

F IGURE 5.12: a) Thermogram for samples with 0min and 240min of isother-
mal curing at 80◦ C b) Tg as a function of post-cure time

Results of the mechanical testing are given in fig.5.13. Stress-strain curves are shown on
panel a). It is chosen to look closer on b) the stiffness, c) ultimate strain and d) the yield
values as function of Tg . Young’s modulus is plotted only for measurements at 1mm/min.
Those at 10mm/min are some tenth of MPa higher or lower as it is the case for the previ-
ous testing series.

5.6 Static creep in compression

The creep strain as a function of the time are given in fig. 5.14 for the three different
constant loading conditions. Regarding the results presented previously, the loads are all
below the yield stress of the material. In the range of the graphs, no thermal variations
exceeding ±0, 8◦ C to the mean temperature (19, 5◦ C) could be detected according to the
thermometer.
Chapter 5. Results 59

F IGURE 5.13: red: 10mm/min blue: 1mm/min a) True stress-strain curves

b) Young’s modulus c) True strain at failure d) True yield stresses as a func-
tion of Tg . Values of KY are plotted and marked on d) as well

F IGURE 5.14: Creep increments and steady-state creep for three different
loading conditions.
 60

Chapter 6

Discussion

6.1 Curing process

Before the mechanical properties are discussed, it is worth to analyze the different curing
cycles from a physico-chemical point of view. Some elements in this discussion are essen-
tial to make the link with findings from mechanical tests afterwards.

The appearance of voids and other defects, as pointed out in section 5.1, posed problems
in the preparation of the samples. The motivation to try curing cycles with lower tem-
peratures compared to other epoxy resins like RTM6 comes from the fact that, at higher
temperatures, more defects appeared.

An emission of CO2 above a certain temperature is not excluded for bio-sourced resins.
This hypothesis could however not be confirmed as curing the resin inside a DSC reveals
no peak that could be related to such a reaction according to De Halleux (2019) [67]. Other
organic volatile stored in the mixture could lead to the development of the defects as well.

The two cycles at low curing degree, namely T amb and 45◦ C w/o PC have the particularity
that their curing temperature stays below the gel point during the whole time, meaning
that there is a transition from liquid to glassy state. In the glassy state, the reaction is
stopped due to limitations in chain diffusion. In fact, as the system is frozen, the chains
can no longer migrate. The reaction processes with a slow rate. Chains have thus the
possibility to arrange in a way such that the degree of packing is high and free volume is
reduced. The thermograms for both cycles show a well-expressed peak for the glass tran-
sition. The fact that this peak is endothermic indicates that, during the DSC-experiment, a
certain amount of energy has to be provided to pass from the glassy to the rubbery state.
Their low glass transition temperature limits however their use: applications exceeding
50◦ C should not be considered.

The cycle 20◦ C + PC is inspired by the producer itself [66]. Initially, the samples are in
the same state as for the T amb ones. However, heating to 80◦ C allows to reach the rub-
bery state, where the crosslinking reaction can go on as diffusion is no longer hindered.
Therefore, nearly full conversion is obtained as the DSC reveals.

For the RTM cycle, Tg∞ is overpassed by about 40◦ C during the post-cure at 120◦ C. One
Chapter 6. Discussion 61

may believe that devitrification could occur. However, parallel work by Ballout (see ap-
pendix A) shows for the same resin no devitrification, even at higher temperatures. There
is no lowering in Tg due to too long exposure at high temperature. Nevertheless, the fact
that the sample is post-cured well above Tg could include free volume: the chains are free
to move and once the temperature is lowered, voids between the chains are frozen inside
the network.

The 45◦ C + PC cycle tries to stay at a low temperature (45◦ C) at the beginning to avoid
any formation of voids inside the tubes that is more likely to occur when temperature is
higher. The longer duration of curing time to reach a sufficient conversion is a drawback.
After 4h at 45◦ C, a post-curing step at 80◦ C (just below Tg∞ ) bridges remaining chains
together. At the end of the post-curing step, samples reach nearly maximum conversion.
The Tg is almost the same as for T amb + PC and in both cases, cooling-down is made from
80◦ C. Therefore, similar results are expected.

It turned out that 45◦ C + PC and T amb + PC cycles presented the less defects and are
analyzed using dynamic tests in the parallel work by van Innis (2020) [3].

6.2 Note on stress-strain curves

For the tensile curves: During the softening, one may get the impression that the strain
increment is negative is there is a steep decrease of stress. This shows a drawback of the
use of scissor-extensometers as only a mean value of deformation inside the gauge region
is obtained as output. When the localization (i.e. the neck) forms outside of this region,
their distancing becomes smaller as both parts of the extensometer are approached due to
the strain softening. Their use does not allow to access the intrinsic deformation for this
purpose, direct-image-correlation would be a better choice.

6.3 Elasticity
First of all, the "high-degree" samples have a modulus slightly above 3GP, which is a typ-
ical value for epoxy resins [46].

The findings that an incomplete cure results in higher stiffness seem at the first look not
obvious. As a reminder, strong secondary interactions, i.e. Van Der Waal’s forces, lead
to higher stiffness. Crosslinking builds up better connected networks but the network
stiffness itself plays only a role in the re-hardening regime, where Young’s modulus is not
defined.

This behavior is however not only observable for this bio-sourced resin. For the well-
studied RTM6-system, similar findings are reported [71, 46].

The fact that the elastic modulus measured in compression is significantly lower for T
amb compared to its values in tension is however unusual. Some sources report similar,
Chapter 6. Discussion 62

or even higher stiffness in compression, a slight asymmetry is not uncommon [71]. An

experimental error for these compression tests is not excluded, especially because some
irregularities in the force-displacement curve, which is the direct output of the testing
machine, are observable. This could also be the reason why the stiffness is lower than it is
the case for 45◦ C w/o PC specimens, even if no physical reason behind this phenomenon
could be identified.

The Young’s modulus is insensitive to the strain rate. However, Richeton’s model (see
eq.2.11) predicts a linear increase, which is not in agreement with the tests results.

6.4 Yielding
Tension

The rate sensitivity could hardly be discussed. In fact, the variance inside the series are
too high so that interpretations are difficult. The subject of the following paragraphs is
rather the effect of the selected cycle.

For the samples cured at ambient temperature T amb, a yield point is never reached dur-
ing a tensile test and therefore not further discussed in this section. Nevertheless, values
of true stress at 0, 2% of plasticity are higher than those of other cycles. This observation
is discussed more detailed for compression tests here-below, the physical reasons behind
are the same.

A strain softening is observed for high curing degrees. As a reminder of chapter 2, there
is a neck that propagates through the material at this stage. The stress drops by about
30%. The peak value is almost the same for each cycle.

Compression

Before looking closer to the differences between the cycles, the strain rate is subject of
interest in this discussion.

The strain rate has an influence on both peak σY and lower yield σl . As predicted by the
studies of other epoxies a linear relation is observable between yield stress and logarithm
of the strain rate, even if in the frame of this work only CCDR and no CSR measurements
are taken. A faster displacement leads to less time for chains to relax, the applied stress to
reach visco-plastic flow has to be higher in this case (see section 2.6.2).

Beside displacement rates, the selected curing cycle has an influence as well. At the first
look, the results seem to be in contrast to other studies. For example, Klavzer (2019) [72]
studied several epoxy systems by analyzing the influence of Tg on yield stress. His work
shows for different resins a linear increase of σY and σl with Tg or, to be more precise, a
Chapter 6. Discussion 63

linear increase of both values with Tg − T where T is the testing temperature. The ex-
planation in his work is that applying a load could be imagined as lowering Tg . Once
T = Tg , molecular chains are able to flow, that is the definition of yielding following
Eyring’s model. However, in the present work, another explanation has to be given in or-
der to explain decrease of yield stress with increasing Tg . As a matter of fact, differences
in Tg arise from incomplete cures, which is in contrast to comparative studies of Klavzer.
The question that has to be answered is: Why does a low curing degree, i.e. low Tg , leads
to a higher yield stress?

As a reminder of the discussion about the cycles, a close-packed structure with really
small amounts of free volume is obtained for low curing degrees and the DSC-thermogram
reveals an endothermic/relaxation peak at glass transition, meaning that a higher energy
barrier has to be overcome. Indeed, a high yield stress indicates that molecular flow is
more hindered because a higher stress has to be applied to enable this flow. Therefore,
the DSC analysis is in agreement with the mechanical tests, both reveal higher values in
the case of T amb and 45◦ C w/o PC.

Comparing values of the fully cured samples (RTM, T amb + PC and 45◦ C + PC) with
other epoxy systems having similar Tg s in compression, the obtained peak stresses be-
tween 80MPa and 90MPa are quite usual. The RTM6 system has, at room temperature
and same CCDR, a yield stress of around 120MPa [46]. The comparative study by Klavzer
shows that the fully-cured InfuGreen810 is in good alignment with other collected data
when not Tg but Tg − T is considered.

F IGURE 6.1: Yield stresses of different epoxy resins as a function of Tg − T.

Values of InfuGreen are those of high-degree batches. Adapted from [72]

The yield stress is however slightly lower for RTM samples. A possible explanation is
that free volume is trapped inside the specimens (see 6.1), leading to lower peak values.
Secondly, the equipment (oven, pump and release agent) for the preparation of RTM sam-
ples was different, what could affect the results as well.
Chapter 6. Discussion 64

Following the stress-strain curve, strain softening takes place once the peak has been over-
passed.

An opposite behavior can be found for the lower yield stress σl . A higher value for the
samples with high curing degree is observed. Only network connectivity has an influence
at this stage as flow is already achieved, σl is independent of thermal history, which is re-
ported for other studies as well [53].

Even if the peak is lower for RTM samples, the lower value is almost identical to the
other high-degree cycles. This reinforces the assumption that free volume is responsible
for the smaller peak values as thermo-mechanical history plays a role rather than the net-
work itself.

The yield drop marks the transition between two regimes governed by packaging (sec-
ondary bonds) and network connectivity (primary bonds) afterwards as both σY and σl
are involved. To investigate the yield drop, a parameter Ky is introduced. The higher
Ky , the higher the softening due to the formation of shear-bands. Values of Ky are more
important for low curing degrees, which is due to both higher σY and lower σl . Similar
results are reported for other studies of incomplete resins by Detwiler (2009) [53].

6.5 Hardening and fracture

This section investigates the evolution of the ultimate stress and strain values in function
of the different testing parameters. The hardening, occurring once strain softening is
completed, is discussed in this section as well. Compression and tensile tests have to be
discussed separately, as their outline is different. The findings are confronted with the
fractographic analysis taken by SEM in the case of tensile specimens and a conventional
camera in the case of compression bullets. After reaching a local minimum in the stress-
strain curve, a (re)-hardening can be observed for all cycles in compression and for the
high degree samples in tension. However, for the latter ones, the increase of stress after
lower yield stress does only take place on a small range. Due to this, this part is limited
to compression where complete stress-strain curves are obtained.

6.5.1 Tension
For samples with high curing degrees (RTM, T amb + PC and 45◦ C + PC), a neck is formed
and the localization propagates, resulting in a higher ductility. This softening is quite un-
usual for an epoxy resin. The RTM6-system, as an example, does not reach the yield point
as it is brittle and breaks before. Nevertheless, in terms of ultimate strains, both resins can
be deformed up to a strain of around 7%, keeping in mind that the elastic response of
RTM6 is more expressed because of a higher yield strength (see appendix B). The brittle
behavior is similar to the T amb curves. In this case, as packaging is high, less energy can
be dissipated by molecular flow, resulting in higher stress concentrations at these defects.

It is worth noticing that a single sample of the 45◦ C + PC batch showed an ultimate stress
Chapter 6. Discussion 65

of nearly 12% which corresponds to twice the value of other samples in this category. As
the fracture originates from defects that are already present in the microstructure of the
samples, reproducible results are difficult to obtain. Especially because many bubbles,
hardly detectable, are present in the specimens due to difficulties in their preparation.
The fact that extensometers are used provides an additional difficulty because the exact
localization of the neck influences the gauge length and therefore the captured data (as
discussed in 6.2). Moreover, testing was not performed on the same day, meaning that the
condition of the extensometers may have changed.

An influence of the displacement rate in tension can be observed. For the T amb batch,
higher rates lead to higher ultimate stresses and to higher amounts of plastic strain. Sim-
ilar to yielding in compression, the chains have less time to flow with higher rates, the
material stays for a longer time in a solid state compared to lower rates where there is
more time for the chains to move so that, mechanically speaking, the response is more
viscous. More valid results have to be collected to discuss its effect on the high-degree
cycles.

Not only mechanical tests, also the microrgaphs taken by SEM suggest a different be-
havior for the highly cured, ductile resins (RTM, 45◦ C + PC and T amb + PC) compared to
the more brittle ones like RTM6 or T amb specimens tested in this work. The results can
be summarized by the observation that the brittle samples are rough whereas the ductile
ones are smooth. In the following part, the two cases are treated separately:

Brittle fracture surfaces

The brittle specimens T amb as well as the RTM with the defect present before tested
in tension have surfaces that are marked mostly by a rough appearance.

In the case of the T amb fracture surface, many similarities can be seen with studies of
other epoxy resins like the RTM6 done by Chevalier (2018) [41]. Both have similar fracture
behavior regarding their stress-strain curve as rupture occurs without softening. Their
surface pattern is the same as well and corresponds to those described by Cantwell (1988)
[61]. Note that a true stress-strain curve and a fracture surface of RTM6 resin can be found
in the appendix B.

The fact that the crack propagation is deviated on higher and lower planes could not
be interpreted but this observation changes little on the fact that the fracture surface is
covered mostly by a three-dimensional relief.

The second specimen with a rough surface is cured by the RTM-cycle. However, the
sample contained a visible defect even before tensile tests have been performed. Similar
to the T amb sample described here-above, fracture happened just after leaving the elastic
regime. In contrast to the other samples of the batch, the mirror-like zone is small, the
rough pattern develops quickly. This means that the crack growth becomes critical in an
early stage. The presence of major defects is therefore responsible for brittle behavior.
Chapter 6. Discussion 66

Ductile fracture surfaces

In contrast, the ductile specimens RTM and 45◦ C + PC studied under SEM have a fracture
surface that is mainly marked by a mirror-like zone and only a few big cracks propagate
through the whole area. There is no real crack multiplication and no real relief is visible.

Understanding the origin of the smoother surfaces means understanding the reason be-
hind the high ductility as both observations are related. In fact, once yielding is com-
pleted, there is an enhanced chain mobility. Visco-plasticity at the chain tip is the main
reason for energy dissipation (see section 2.7.1), therefore crack multiplication does not
take place. The crack propagation in the smooth zone is sub-critical according to Cantwell
(1988) [61]. This means that the cracks propagate slowly through the material. Only at
the very end, a slightly rougher morphology is observed. This indicates that the growth
is accelerated.

In a similar way, Cantwell identified that an increase in temperature leads also to big-
ger mirror-like plateaus which can be related to the higher ultimate strain in tension1 .
A reason behind the fact that the bio-based resin has a similar appearance at room tem-
perature could be the relative low Tg . As a matter of fact, the resin Cantwell describes
as well as the RTM6-system have Tg s that are respectively 50◦ C and 150◦ C higher than
the bio-based one with a Tg ∼ 70◦ C. Both RTM6 and Cantwell’s resins tested at room
temperature show much rougher surfaces as the Tg − T is higher. Unfortunately, litera-
ture is poor in comparisons of fracture surfaces with different Tg for uni-axial tensile tests.

To summarize the fracture for specimens 45◦ C and RTM: the high ductility is the result of
a low span between Tg and T. The crack propagation is mainly sub-critical because chains
can flow at the crack tip. This slow growth is in direct link with the smooth fracture sur-
faces.

As a perspective: To confirm these results, tensile tests at different temperatures have to

be performed. A parameter that is not yet investigated by fractography is the strain rate.
But increasing the rate has a similar effects to lowering the temperature, fast displacement
could lead to a rougher surface [61].

6.5.2 Compression
In compression, the fracture mechanisms differ. Opening of microcavities is inhibited. By
the nature of compression itself and because a solid lubricant preventing barreling, the
stress-strain response can expand further.

1 Even
if this work is limited to tests at room temperature, the ultimate stress for higher temperatures is
commonly observed for thermosets as long as Tg is not overpassed. This is visualized in chapter 2 in fig.2.10
Chapter 6. Discussion 67

Before fracture occurs, hardening happens and is discussed in the following part. Interest-
ingly, the plateau at which nearly no variation of the stress with strain could be observed,
is broader in the case of low curing degrees. This means that the onset of hardening and
thus the regime where covalent bonds determine the stress-strain curve occurs later.

As a reminder, during hardening the increase in stress with strain is characterized by

the hardening modulus GR = ρe RT with ρe the crosslinking density. At this stage, the
chains are aligned and only ρe impacts its value. The mechanical tests indicate that the
highest amount of crosslinks are present in the case of RTM samples as GR is the highest.
However, a high curing degree α is synonym of a high ρe . This forms a contrast with the
DSC-analysis that states that the curing degree of RTM samples should be lower than
those of 45◦ C + PC and T amb + PC. The interpretation of mechanical tests and the DSC
analysis is thus contradictory in the specific case of the RTM-samples.

As expected, the cycles with low curing degrees have the lowest GR ; to be accurate 45◦ C
w/o PC has a slightly higher value than T amb.

One could try to link ρe with α not only qualitatively but also quantitatively. As a re-
minder, the curing degree can be visualized as the fraction of oxirane rings that are con-
verted into covalent bonds with amine groups of the hardener.

This approach fails because the crosslinking density is, according to the hardening ap-
proach, twice as high for high curing degrees compared to lower ones, even if the differ-
ence between the curing degrees is only about 15%. With other words, the mechanically
accessed gap in curing extend between cycles is overestimated compared to physico-
chemical based methods. Moreover, the resin is a mixture containing several different
hardener and epoxy molecules, making it difficult to relate ρe and α. In fact, α is mea-
sured using a DSC, meaning that heat produced by the reaction provides its value. How-
ever, this energy released by the exothermic reaction is not necessarily the same for every
molecule and varies between the first and second reaction that a single nitrogen atom can
complete. The rapid rise of GR is reported in other works [53, 73]. Charlesworth (1988) [73]
developed an advanced model, but this goes beyond the scope of this work.

The findings regarding the ductility are different comparing compression and tension:
in compression, samples with low curing degree have the highest ductility. In fact, as
their Tg is smaller and thus closer to the testing temperature, a similar effect observed by
testing at higher temperatures is present. In fact, for small spans of Tg − T, the chain mo-
tion is enhanced, explaining why the low-degree (=low Tg ) bullets are squeezed out from
the piston of the subpress. The analogy to a wax seal on paper can be retaken. If the wax
is not heated, applying a pressure results in formation of cracks, the material is likely to
break. However, an increased temperature allows the molecules of the wax to flow. When
a pressure is applied in this situation, there is no rupture and the seal squeezes out from
the stamp.

Stress at failure has not been discussed yet. It is difficult to do interpretations because
Chapter 6. Discussion 68

the samples of the low degrees do not really fail at the end of compression tests.

One could ask the question: why have samples with low Tg higher yield stresses because
molecular flow is hindered and meanwhile a better flow leading to high amounts of plas-
tic strains? This is not contradictory as the first phenomenon is due to compact packing
and small free volume. Once yielding is completed, the free volume does not play a role
anymore and only the difference between glass transition temperature and testing tem-
perature governs the ease of molecular flow.

Regarding the ultimate stress and strain values of fully-cured samples, a good agreement
is found with other epoxy systems, as it can be seen in fig. 6.2.

F IGURE 6.2: Ultimate stress and strains for different epoxy resins as a
function of T/Tg are shown. Values of InfuGreen are those of high-degree
batches. Adapted from [72]

6.6 Influence of the post-cure

Physical explanations of the previous section are still accurate and not repeated in the
following part. First of all, it can be confirmed by analyzing the DSC-thermograms that
samples cured incompletely have an endothermic peak that is more expressed, which is
in alignment with the comparison of incompletely versus completely cured samples.

A linear relation between Tg and E can be demonstrated by studying the effect of post-
curing which reinforces the idea that secondary interactions decrease with Tg

No real relation could be established for the yield stress. There is a non-linear decrease,
values for σY seem to reach a lower limit arriving at almost full cure. Staying at 80◦ C for
more than 100min instead of 240min changes nothing in terms of yield stress.
Chapter 6. Discussion 69

The curing extend also influences the yield drop and ultimate strain. There is a linear
correlation between KY and Tg . The linear correlation for the strain at rupture could be
explained by the fact that the plateau after peak expends further for decreasing Tg , the
strain accumulated before reaching this plateau is not significantly different to explain
this observation.

A given incertitude remains for the Tg calculated by the model-free kinetic method as
differences between the theoretical model and experimental errors could be expected.

The findings that Young’s modulus and KY decrease linearly whereas peak yield de-
creases asymptotically with Tg is opposite to those published by Chang (1982) [74] for
another epoxy resin. Detwiler (2009) [53] reports results that are similar to those in this
study. Carbas (2013) [30] who investigated incomplete cures for three different epoxy sys-
tems, showed that two of them have a stiffness and a peak stress that increase also linearly
with Tg , i.e. that becomes higher with higher curing degrees. This reinforces the idea that
no general theory valid for all kind of epoxy systems can be established.

6.7 Creep behavior

A first regime, known as primary creep, is followed by the secondary creep that shows a
linear correlation between creep strain with time.

At the end of the secondary creep, a tertiary creep might occur where the creep strain
increases exponentially with time. The tertiary creep can be seen as the onset of failure.
Such a regime can however not be observed in the frame of these tests, even for a longer
duration of about two months for the sample tested at 49MPa.

Tests at several temperatures and different charges have to be accomplished to charac-

terize the entire creep-behavior of the resin. Moreover, due to the diffusion controlled
nature of creep tests, testing of aged samples where chains have a lower ability to flow
could be interesting.
 70

Chapter 7

Conclusion

This work, parallel to the Master’s thesis of Van Innis, is a first mechanical characteriza-
tion of the partly bio-sourced InfuGreen810 epoxy resin. The most challenging part of this
work consisted to find an adapted procedure to cure the resin without appearance of de-
fects. It was found that bubbles appeared if the curing temperature was high, whereas
cracks due to the brittle behavior caused damage in the case of low curing temperatures.

The stress-strain response of the material tested under different displacement rates was
obtained for five different curing cycles. In addition, a study establishing relations be-
tween Tg and some mechanical properties was performed. The resin studied has good
mechanical characteristics with yield stresses above 80MPa and Young’s modulus above
3GPa, compared to other high-performance systems like the RTM6. A sufficient resis-
tance to creep could be identified as well.

One of the findings is a decrease of stiffness with increasing curing degree. This decrease
was found to be linear. A lower peak yield for higher curing extends was also observed.
In both cases, a link with physico-chemical properties could be obtained by a DSC anal-
ysis that revealed a well-pronounced endothermic peak. This peak constitutes an energy
barrier hindering yielding and is synonym of strong secondary interactions.

In terms of post-yield and failure behavior, distinction has to be done between tensile
and compression tests. Former ones revealed that samples with high curing degree show
strain-softening, which is unusual for glassy polymers, and therefore a high ductility. In
contrast, no real failure could be observed during compression tests with low-degree sam-
ples as the specimens were squeezed out from the testing device.

Most properties are in good agreement with other collected data for other systems, as
it can be seen by extending the comparative study of Klavzer with the results taken in this
work.

It could not be said that one of the tested cycles is the best choice in any case. The se-
lection can be done depending the application and the required properties.

Unfortunately, the glass transition temperature of less than 80◦ C limits the use of the
Chapter 7. Conclusion 71

InfuGreen810 to applications with relative low operational temperatures. They are there-
fore probably not suited for aeronautical composites. This conclusion is quite general for
(partly) bio-sourced resins as they tend to have a lower Tg compared to conventional ones.

The real ecological benefit is difficult to measure. Even if the carbon footprint is lower
compared to petrol-based resins, the end-of-life concerns still remain as the resin is nei-
ther bio-degradable nor recyclable.

This Master’s thesis was the first attempt to characterize this novel resin mechanically.
However, more experimental information must be collected before bio-sourced compos-
ites can be developed with this kind of resin. Thus, some perspectives can be formulated
as "guideline" for following works:

• Finding a curing process without leading to defects remains still challenging. Espe-
cially the exact origin of the bubbles for highly cured bubbles could not be explained
in the frame of this work. One possibility would be to cure the epoxy inside a FTIR,
so that a possible emission of CO2 can be proved. Another way to avoid this prob-
lem is the use of flatter geometries. For the ease of specimen preparation, glass
tubes have been used so far. They have the disadvantage that their height is rather
elevated for efficient degasing. Especially the fact that the resin should be used as
matrix for FRPs would justify flatter molds. Potentially, this would be sufficient to
avoid the formation of bubbles.

• Samples cured by T amb + PC or 45◦ C + PC presented the less defects. Unfortunately,

they require much time. Other hardeners can be used to turn the kinetics and to
accelerate network formation.

• To validate some results, more batches are necessary. As batch-to-batch variability

hardly avoidable, testing only a single batch provides insufficient statistical mean-
ings.

• Testing was only performed at room temperature. Other testing temperatures have
to be imposed to have a better understanding of the resin. Especially for creep tests,
data at different temperatures can be useful to develop TTS-shifts.

Only the future will show if bio-sourced composite materials will be competitive with
conventional ones. Even if the behavior of bio-sourced resins and fibers are understood,
the elaboration of fabrication processes for hybrid materials and their characterization
will take some time before those materials will be employed in high-performance appli-
cations.
 72

Appendix A

Physico-chemical results of
InfuGreen810-SD4771 system

This section regroups some main findings of the work of De Halleux who did a physico-
chemical characterization of the same epoxy-hardener system used in this work.

First, the figure A.1 shows the linear dependency between conversion and Tg .

F IGURE A.1: Linear dependency between Tg and α. From: [67]

Ballout studied the same epoxy system using DSC. The resin is cured at higher tempera-
tures than it is the case for this work. It can be seen that the endothermic peak is slightly
shifted to the right for higher curing temperatures, meaning that the Tg is higher. The
thermograms are shown in fig. A.2.
Appendix A. Physico-chemical results of InfuGreen810-SD4771 system 73

F IGURE A.2: Thermograms of the InfuGreen810 resin cured at higher tem-

peratures

The curve in fig. A.3 shows the result of a isothermal model-free kinetic method giving
the curing degree as a function of time. The red dots on the graph show values taken
for the further analysis. The graph can be seen as a validation of the model free kinetic
method for computing Tg . The two outer dots are obtained thanks to a DSC analysis. The
inner dots are spaced by the time that samples spent in the oven at 80◦ C from the first dot
(60min and 100min). Their Tg is taken as the value yielded on the curve. The model-free
method overpredicts α by about 1%.

F IGURE A.3: Isothermal evolution of curing degree with time

 74

Appendix B

Mechanical properties of the RTM6

resin

The following two figures show the stress-strain response of the RTM6 resin and a frac-
ture surface surface of a tensile specimen. Both figures are the result of tests performed at
room temperature.

F IGURE B.1: Stress-strain response of RTM6 resin. From [41]

F IGURE B.2: Fracture surface of RTM6 resin. From [46]

 75

Bibliography

[1] Projet MACOBIO, accessed: 30/05/2020. URL: http://www.materianova.be/index.

php/macobio/.
[2] MACOBIO, accessed: 30/05/2020. URL: https : / / uclouvain . be / en / research -
institutes/imcn/bsma/macobio.html.
[3] Charline van Innis. “Static and fatigue failure mechanisms of a bio-sourced epoxy
resin: towards a bio-based composite”. PhD thesis. UCLouvain, Prom.: Pardoen,
Thomas, 2020.
[4] Lebenszyklusanalyse, accessed: 12/06/2020. URL: https://www.timeout.de/pdf/SR_
GP_Greenpoxy\%20Lebenszyklus-Analyse_2017_09_04.pdf.
[5] “Encyclopedia of Biomedical Engineering”. In: vol. 1-3. Elsevier BV, 2019. Chap. Man-
ufacture of Biomaterials, pp. 116 –134.
[6] Biocomposites, accessed: 30/03/2020. URL: https://en.wikipedia.org/wiki/Biocomposite.
[7] Natural composites in automotive industry. accessed: 30/03/2020. URL: http://www.
naturalfibersforautomotive.com/?cat=3.
[8] Polylactic acid, accessed: 28/03/2020. URL: https://en.wikipedia.org/wiki/Polylactic_
acid.
[9] Hans-Peter Fink Mohini Sain Omar Faruka Andrzej K. Bledzki. “Biocomposites re-
inforced with natural fibers: 2000–2010”. In: International Journal of Polymer Science
2015 (2015), pp. 1–10. DOI: 10.1155/2015/390275.
[10] Tim Hartness, George Husman, and Joel Dyksterhouse. “The characterization of
low cost fiber reinforced thermoplastic composites produced by the DRIFT pro-
cess”. In: composites, part A: applied science and manufacturing 32 (2001), pp. 1155 –
1160.
[11] J. Verrey, M.D. Wakeman, and J.-A.E. Manson V. Michaud. “Manufacturing cost
comparison of thermoplastic and thermoset RTM for an automotive floor pan”. In:
composites, part A: applied science and manufacturing 37 (2006), pp. 9 –22.
[12] Elium resins for composites, accessed: 8/04/2020. URL: https://www.arkema.com/en/
products/product-finder/range-viewer/Elium-resins-for-composites/.
[13] A. Gerhardt. “Plastic Additive Bisphenol A: Toxicity in Surface- and Groundwater
Crustaceans”. In: Journal of Toxicology and Risk Assessment 5 (1 2019). DOI: 10.23937/
2572-4061.1510017.
[14] Susan Shelnutt, John Kind, and William Allaben. “Bisphenol A: Update on newly
developed data and how they address NTP’s 2008 finding of “Some Concern””. In:
Food and Chemical Toxicology 57 (2013), pp. 284 –295.
Bibliography 76

[15] Hideto Tsuji and Shinya Miyauchi. “Enzymatic Hydrolysis of Poly(lactide)s: Effects
of Molecular Weight, L-Lactide Content, and Enantiomeric and Diastereoisomeric
Polymer Blending”. In: Biomacromolecules 2 (2001), pp. 597 –604.
[16] E. Muñoz and Juan Garcia-Manrique. “Water Absorption Behaviour and Its Effect
on the Mechanical Properties of Flax Fibre Reinforced Bioepoxy Composites”. In:
International Journal of Polymer Science 2015 (2015), pp. 1–10. DOI: 10.1155/2015/
390275.
[17] H.N. Dhakal, Z.Y. Zhang, and M.O.W. Richardson. “Effect of water absorption on
the mechanical properties of hemp fibre reinforced unsaturated polyester compos-
ites”. In: Composites Science and Technology 67 (2007), pp. 1674 –1683.
[18] María González, Juan Carlos Cabanelas, and Juan Baselga. Applications of FTIR on
Epoxy Resins, Identification, Monitoring the Curing Process, Phase Separation and Wa-
ter Uptake Materials Science, Engineering and Technology. Ed. by Prof. Theophanides
Theophile. InTech, 2012. ISBN: 978-953-51-0537-4.
[19] Epoxide, accessed: 24/03/2020. URL: https://en.wikipedia.org/wiki/Epoxide.
[20] “Comprehensive Polymer Science and Supplements”. In: vol. 5. Pergamon, 1996.
Chap. 37.
[21] W. Fisch, W. Hofmann, and J. Koskjkallio. “The curing mechanism of epoxy resins”.
In: Journal of applied chemistry 6 (1956), pp. 429 –441.
[22] Sebastian Koltzenburg, Michael Maskos, and Oskar Nuyken. “Ring-Opening Poly-
merization”. In: Polymer Chemistry. Berlin, Heidelberg: Springer Berlin Heidelberg,
2017, pp. 321–347. ISBN: 978-3-662-49279-6. DOI: 10.1007/978-3-662-49279-6_12.
URL : https://doi.org/10.1007/978-3-662-49279-6_12.

[23] Jeff Gotro. Epoxy Cure Chemistry Part 4: Nucleophiles in Action. accessed: 25/03/2020.
2014. URL: https://polymerinnovationblog.com/epoxy-cure-chemistry-part-
4-nucleophiles-action.
[24] C.-S. Chern and G. W. Poehlein. “A Kinetic Model for Curing Reactions of Epoxides
with Amines”. In: Polymer engineering and science 27.11 (1987), pp. 788 –795.
[25] A. Yousefi, P. G. Lafleur, and R. Gauvin. “Kinetic Studies of Thermoset Cure Reac-
tions: A Review”. In: Polymer composites 18.2 (1997), pp. 157 –168.
[26] Sergey Vyazovkin. “Computational aspects of kinetic analysis. Part C. The ICTAC
Kinetics Project Ð the light at the end of the tunnel?” In: Thermochimica Acta 355.2000
(1999), pp. 155 –163.
[27] M.E. Brown et al. “Computational aspects of kinetic analysis Part A: The ICTAC ki-
netics project-data, methods and results”. In: Thermochimica Acta 355 (2000), pp. 125
–143.
[28] Ghodsieh Mashouf Roudsari, Amar K. Mohanty, and Manjusri Misra. “Study of the
Curing Kinetics of Epoxy Resins with Biobased Hardener and Epoxidized Soybean
Oil”. In: ACS sustainable chemistry and engineering 2 (2014), pp. 2111 –2116.
[29] C.F. Dickinson and G.R. Heal. “A review of the ICTAC kinetics project, 2000 Part 2.
Non-isothermal results”. In: Thermochimica Acta 494 (2009), pp. 15 –25.
Bibliography 77

[30] R. J. C. Carbas et al. “Effect of Cure Temperature on the Glass Transition Tempera-
ture and Mechanical Properties of Epoxy Adhesives”. In: The Journal of Adhesion 90
(2013), pp. 104–119. DOI: 10.1080/00218464.2013.779559.
[31] M . R. Tant and G. L. Wilkes. “An Overview of the Nonequilibrium Behavior of
Polymer Glasses”. In: Polymer engineering and science 21.14 (1981), pp. 874 –895.
[32] J. D. Menczel and R. B. Prime. Thermal Analysis of Polymers: Fundamentals and Appli-
cations. Wiley.
[33] J.B. Enns and J.K. Gillham. “Effect of the Extent of Cure on the Modulus, Glass
Transition, Water Absorption, and Density of an Amine-Cured Epoxy”. In: Journal
of Applied Polymer Science 28 (1983), pp. 2831–2846.
[34] K. P. Pang and J. K. Gillham. “Competition between Cure and Thermal Degradation
in a High Tg Epoxy System: Effect of Time and Temperature of Isothermal Cure on
the Glass Transition Temperature”. In: Journal of Applied Polymer Science 39 (1990),
pp. 909 –933.
[35] Horng-Jer Tai and Hui-Lung Chou. “Chemical shrinkage and diffusion-controlled
reaction of an epoxy molding compound”. In: European Polymer Journal 36 (2000),
pp. 2213 –2219.
[36] Omar Moussa, Anastasios P.Vassilopoulos, and Thomas Keller. “Effects of low-
temperature curing on physical behavior of cold-curing epoxy adhesives in bridge
construction”. In: International Journal of Adhesion Adhesives 32 (2012), pp. 15–22.
[37] Andrzej Bledzki et al. “Bio-Based Epoxies and Composites for Technical Applica-
tions”. In: Key Engineering Materials 559 (June 2013), pp. 1–6. DOI: 10 . 4028 / www .
scientific.net/KEM.559.1.
[38] G. M. Odegard and A. Bandyopadhyay. “Physical Aging of Epoxy Polymers and
Their Composites”. In: Journal of Polymer Science Part B: Polymer Physics 49 (2011),
pp. 1695 –1716.
[39] Aaron W. Thornton et al. “New relation between diffusion and free volume: II. Pre-
dicting vacancy diffusion”. In: Journal of Membrane Science 338 (2009), pp. 38 –42.
[40] Han E.H. Meijer and Leon E. Govaert. “Mechanical performance of polymer sys-
tems: The relation between structure and properties”. In: Progress in Polymer Science
30 (2005), pp. 915 –938.
[41] Jérémy Chevalier. “Micromechanics of an epoxy matrix for fiber reinforced compos-
ites: experiments and physics-based modelling”. PhD thesis. UCLouvain, Prom.:
Pardoen, Thomas ; Lani, Frédéric, 2018.
[42] Darshil U. Shah and Peter J. Schubel. “Evaluation of cure shrinkage measurement
techniques for thermosetting resins”. In: Polymer Testing 29 (2010), pp. 629 –639.
[43] M. Zarrelli, A. A. Skordos, and I. K. Partridge. “Investigation of cure induced shrink-
age in unreinforced epoxy resin”. In: Plastics, Rubber and Composites 31.9 (2002),
pp. 377–384. DOI: 10.1179/146580102225006350.
Bibliography 78

[44] S.R. White and H.T Hahn. “Process modeling of composite materials: Residual
stress development during cure. Part 2. Experimental validation”. In: Composite
Mater 26.16 (1992), pp. 2423 –2453.
[45] Ajith K. Gopal, Sarp Adali, and Viktor E. Verijenko. “Optimal temperature pro R les
for minimum residual stress in the cure process of polymer composites”. In: Com-
posite Structures 48 (2000), pp. 99 –106.
[46] Xavier Morelle. “Mechanical characterization and physicsbased modeling of highly-
crosslinked epoxy resin”. PhD thesis. UCLouvain, Prom.: Pardoen, Thomas ; Bailly,
Christian, 2015.
[47] Shahin Shadlou, Babak Ahmadi-Moghadam, and Farid Taheri. “The effect of strain-
rate on the tensile and compressive behavior of graphene reinforced epoxy/nanocomposites”.
In: Materials and Design 59 (2014), pp. 439 –447. DOI: dx . doi . org / 10 . 1016 / j .
matdes.2014.03.020.
[48] J. Richeton, S. Ahzi, and L. Daridon. “Thermodynamic investigation of yield-stress
models for amorphous polymers”. In: Philosophical Magazine 87 (2007), pp. 3629 –
3643.
[49] A.S. Argon. “Large inelastic deformation of glassy polymers. PART I: Rate depen-
dent constitutive model”. In: Mechanics of Materials 7 (1988), pp. 15 –33.
[50] R.N. Haward. The Physics of Glassy Polymers. Springer, 1973. ISBN: 978-94-010-2355-
9.
[51] Lambert C. A. van Breemen et al. “Rate- and Temperature-Dependent Strain Soft-
ening in Solid Polymers”. In: Journal of Polymer Science 50 (2012), pp. 1757 –1771.
[52] H.G.H. van Melick, L.E. Govaert, and H.E.H. Meijer. “On the origin of strain hard-
ening in glassy polymers”. In: Polymer 44.8 (2003), pp. 2493 –2502. ISSN: 0032-3861.
DOI : https :// doi. org/ 10. 1016/ S0032- 3861(03)00112- 5. URL : http :// www.
sciencedirect.com/science/article/pii/S0032386103001125.
[53] A.T. Detwiler and A.J. Lesser. “Aspects of Network Formation in Glassy Ther-
mosets”. In: Journal of Applied Polymer Science 117 (2010), pp. 1021 –1034. DOI: 10.
1002/app.31681.
[54] R.N. Haward and G. Thackray. “The use of a mathematical model to describe isother-
mal stress-strain curves in glassy thermoplastics”. In: Proc. Roy. Soc. 302 (1968),
pp. 453–472.
[55] A.S. Argon. “A theory for the low-temperature plastic deformation of glassy poly-
mers”. In: Philosophical Magazine 28 (1973), pp. 839 –865.
[56] T. Berstad et al. “The use of a mathematical model to describe isothermal stress-
strain curves in glassy thermoplastics”. In: 9th International LS-DYNA Conference 20
(2013), pp. 16–24.
[57] J. Richeton, S. Ahzi, and L. Daridon. “Viscosity, Plasticity, and Diffusion as Exam-
ples of Absolute Reaction Rates”. In: The Journal of Chemical Physics 4 (1936), pp. 283
–291.
Bibliography 79

[58] Chris S. Henkee and Edward J. Kramer. “Crazing and shear deformation in crosslinked
polystyrene”. In: Journal of Polymer Science: Polymer Physics 22 (1984), 721–737. DOI:
10.1002/pol.1984.180220414.
[59] A.A. Griffith. “The Phenomena of Rupture and Flow in solids”. In: Philosophical
Transactions of the Royal Society of London 221 (1921), pp. 163 –198.
[60] J.P. Berry. “Determination of fracture surface energies by the cleavage technique”.
In: Journal of Applied Physics 34 (1963), pp. 62 –68.
[61] W. Cantwell, A. Roulin-Moloney, and T. Kaiser. “Fractography of unfilled and particulate-
filled epoxy resins”. In: Journal of Polymer Science: Polymer Physics 23 (1988), 1615–1631.
[62] Michael Ashby, Hugh Shercliff, and David Cebon. Materials: Engineering, Science,
Processing and Design. Butterworth-Heinemann, 2007. ISBN: 978-0-7506-8391-3.
[63] C.-W. Feng et al. “Modeling of long-termcreep behavior of structural epoxy adhe-
sives”. In: International Journal of Adhesion Adhesives 25 (2005), pp. 427 –436.
[64] E. Kontou. Creep and fatigue in polymer matrix composites - Second edition. Ed. by Rui
Miranda Guedes. Woodhead Publishing, 2019. ISBN: 9780081026014.
[65] Creep and Stress Rupture Properties, accessed: 19/04/2020. URL: https://www.totalmateria.
com/page.aspx?ID=CheckArticle&site=kts&NM=296.
[66] SR InfuGreen 810, accessed: 12/06/2020. URL: http://www.sicomin.com/datasheets/
product-pdf1167.pdf.
[67] Lancelot de Halleux. “Analyse des propriétés physico-chimiques de résines époxy
biosourcées”. PhD thesis. UCLouvain, Prom.: Bailly, Christian, 2019.
[68] Differential Scanning Calorimetry (DSC), accessed: 27/04/2020. URL: https : / / www .
netzsch-thermal-analysis.com/en/contract-testing/methods/differential-
scanning-calorimetry/.
[69] Differential scanning calorimetry, accessed: 27/04/2020. URL: https://en.wikipedia.
org/wiki/Differentialscanningcalorimetry.
[70] Standard Test Method for Tensile Properties of Plastics, accessed: 27/04/2020. URL: https:
//www.astm.org/Standards/D638.
[71] Judith Moosburger-Will et al. “Influence of Partial Cross-Linking Degree on Basic
Physical Properties of RTM6 Epoxy Resin”. In: Journal of Applied Polymer Science
6.130 (2013), pp. 4338–4346. DOI: 10.1002/APP.39722.
[72] Nathan Klavzer. “Epoxy and Bio-Sourced Resins for Composite Applications: Me-
chanical Testing and Modelling”. PhD thesis. UCLouvain, Prom.: Pardoen, Thomas,
2019.
[73] J. M. Charlesworth. “Effect of Crosslink Density on Molecular Relaxations in Diepoxide-
Diamine Network Polymers. Part 2. The Rubbery Plateau Region”. In: Polymer En-
gineering and science 28 (1988), pp. 230 –236.
[74] T.D. Chang, S.H. Carr, and J.0. Brittain. “Studies of Epoxy Resin Systems: Part B:
Effect of Crosslinking on the Physical Properties of an Epoxy Resin”. In: Polymer
Engineering and Science 22.18 (1982), pp. 1213 –1220.
UNIVERSITÉ CATHOLIQUE DE LOUVAIN
École polytechnique de Louvain
Rue Archimède, 1 bte L6.11.01, 1348 Louvain-la-Neuve, Belgique | www.uclouvain.be/epl