In Situ Polimerization TP For Continuous Composites

Katholieke Universiteit Leuven
Faculteit Ingenieurswetenschappen
Departement Metaalkunde en
Toegepaste Materiaalkunde
Kasteelpark Arenberg 44 - B-3001 Leuven
Characterisation of the in-situ

polymerisation production process
for continuous fibre reinforced
thermoplastics
Promotor: Proefschrift ingediend tot

Prof. Ignaas Verpoest het behalen van het doctoraat
Co-Promotor in de ingenieurswetenschappen
Prof. Paula Moldenaers door
Hilde Parton
Februari 2006
Faculteit Ingenieurswetenschappen
Departement Metaalkunde en
Toegepaste Materiaalkunde
Kasteelpark Arenberg 44 - B-3001 Leuven
Characterisation of the in-situ

polymerisation production process
for continuous fibre reinforced
thermoplastics
Jury: Proefschrift voorgedragen tot

Prof. Guido De Roeck: Voorzitter het behalen van het doctoraat
Prof. Ignaas Verpoest: Promotor in de ingenieurswetenschappen
Prof. Paula Moldenaers : Co-Promotor door
Prof. Ludo Froyen
Prof. Stepan Lomov
Prof. Jacques Devaux (U.C.L.)
Prof. Bart Goderis Hilde Parton
Dr. Véronique Michaud (EPFL)
UDC: 66.095.26
Februari 2006
c
°Katholieke Universiteit Leuven – Faculteit Ingenieurswetenschappen
Arenbergkasteel, B-3001 Heverlee (Belgium)
Alle rechten voorbehouden. Niets uit deze uitgave mag worden vermenig-
vuldigd en/of openbaar gemaakt worden door middel van druk, fotokopie,
microfilm, elektronisch of op welke andere wijze ook zonder voorafgaande
schriftelijke toestemming van de uitgever.
All rights reserved. No part of the publication may be reproduced in any

form by print, photoprint, microfilm or any other means without written
permission from the publisher.
D/2006/7515/5
ISBN 90–5682–666–2
Preface/Voorwoord
Na een zalige periode van drie maanden waarin ik niet aan mijn doctoraat
moest denken, zit ik hier terug achter mijn computer om de laatste hand
aan mijn thesis te leggen. Dit voorwoord is dan ook één van de laatste
dingen die ik schrijf, maar het is zeker niet het gemakkelijkste deel. Er
zijn namelijk zoveel mensen die mij deze laatste jaren geholpen hebben
om dit doctoraat tot een goed einde te brengen... Sorry dan ook als ik
iemand vergeet, maar dat betekent niet dat ik jullie minder dankbaar ben.
Eerst en vooral zou ik de techniekers willen bedanken want zonder hun
kennis en hulp is een experimenteel doctoraat onmogelijk. Jo en Manuël,
bedankt om mijn mal lekvrij te maken. Ik weet dat er gevloekt is, maar
uiteindelijk is alles toch in orde gekomen, zelfs met de bladveer! Bart,
Kris, Johan en Louis, bedankt om altijd klaar te staan als ik een vraagje
had of iets niet kon optillen. Ik ben altijd met veel plezier naar de hal
gekomen.
Vervolgens mag ik natuurlijk prof Ignaas Verpoest niet vergeten. Zijn
enthousiasme en liefde voor zijn vak zijn ongeëvenaard en werkten aansteke-
lijk. Bovendien heeft hij mij de kans gegeven om zelf een richting te kiezen
voor mijn doctoraat ook al betekende dit dat ook hij moest bijleren.
Then I would like to thank prof Jacques Devaux for giving me the
opportunity to do the GPC measurements and Pascale Lipnik for making
the TEM images. These have both become crucial elements of my thesis.
Een ander cruciaal deel van mijn thesis heb ik te danken aan prof
Bart Goderis. Door de metingen in Grenoble heeft mijn thesis een extra
dimensie gekregen. Bovendien heb je altijd tijd vrij gemaakt om mij te
helpen en dat was nodig aangezien ik weinig tot niets van de technieken
kende. Thanks also to Monika Basiura for doing the first steps of the data
processing.
Natuurlijk wil ik ook mijn leescomité, prof Paula Moldenaers, prof
Stepan Lomov en prof Ludo Froyen, bedanken voor het lezen van de tekst
en het geven van opmerkingen.
Thanks also to dr Véronique Michaud for willing to come from Switzer-
land to be a member of my jury. Moreover, we have had several interesting
discussions during conferences and it sometimes felt comforting to know
i
ii Preface/Voorwoord
that things are very similar wherever you go.

The Amiterm project partners are also acknowledged for their input in
my work and the interesting discussions. I would also like to express my
gratitude to Cyclics Corporation for giving me the opportunity to work
with their interesting material.
I think that everybody would agree that a job needs to be interesting,
however, the people you work with are equally important. Therefore, I
would like to thank all my colleagues from the CMG and from MTM,
for ensuring a pleasant atmosphere. Special thanks go to my office mates
Isabel, Darren, Jochen and Joris. It was great to be in an office where
sometimes composites were not the most important thing on earth! En
Isabel, sommigen denken misschien dat we na vier jaar wel uitgebabbeld
zijn, maar niets is minder waar. Ik zal het missen om een vriendin te
hebben op het werk!
En dan zijn er natuurlijk nog de familie en vrienden. Het was absoluut
noodzakelijk om af en toe stoom af te laten en te zagen en te klagen,
dat lucht op! En jullie hebben allemaal je deel gehad, bedankt om te
luisteren...
Mama en papa, zonder jullie was ik natuurlijk niet geworden wie ik
nu ben. Bedankt voor alles!!!
En dan resten er mij nog de twee belangrijkste mannen in mijn leven.
Wouter, bedankt dat je mijn planning gerespecteerd hebt en niet nog een
weekje vroeger bent geboren. Maar vooral bedankt voor je lachjes, ze
maken keer op keer mijn dag goed!
En Bert, jou kan ik nooit genoeg bedanken. Je bent werkelijk mijn
steun en toeverlaat... Keer op keer heb je geluisterd en me opnieuw moed
ingesproken als het tegen zat. Of frietjes gaan halen omdat dat me vreemd
genoeg troost. Zonder jou had ik het zeker niet gekund, je geeft me rust
en dat heb ik nodig.
Bedankt-Thanks
^
¨
Abstract/Samenvatting
Abstract The advantages of continuously reinforced thermoplastic com-

posites are overshadowed by the high melt viscosity of the matrix which
seriously hampers impregnation. In-situ polymerisation of cyclic oligo-
mers, such as CBTr resin, however, simplifies impregnation and even al-
lows for the use of thermoset production techniques such as Resin Transfer
Moulding.
The feasibility of using CBTr resin to isothermally produce composites
was demonstrated and the resulting matrix and composite properties were
characterised. The isothermal nature of the production process leads to
the development of well-oriented, thick lamellae with few interlamellar
tie-molecules. This results in brittle matrix behaviour that decreases the
impact resistance of the resulting composites.
A more fundamental study of the crystallisation process revealed a three
phase structure, consisting of semi-dense stacks in which liquid amorphous
and dense lamellae are alternated. The dense lamellae, in turn, consist
out of crystalline grains and a dens amorphous fraction.
Samenvatting De voordelen van continu vezelversterkte thermoplasten

worden overschaduwd door de hoge smelviscositeit van de matrix die de
impregnatie bemoeilijkt. In-situ polymerisatie van cyclische oligomeren,
zoals CBTr hars, vereenvoudigt de impregnatie echter en laat boven-
dien toe om productiemethodes voor thermohardende composieten, zoals
RTM, te gebruiken.
De mogelijkheid om composieten isotherm te produceren met CBTr hars
wordt aangetoond. Bovendien worden ook de resulterende matrix- en
composieteigenschappen gekarakteriseerd. Het isotherm karakter van het
productieproces leidt tot de vorming van georiënteerde, dikke lamellen met
weinig verbindingsmoleculen. Dit resulteert in een bros matrixgedrag en
een verlaagde impactweerstand van de composieten.
Een meer fundamentele studie van het kristallisatieproces leidde tot een
drie-fasen structuur bestaande uit semi-dense stapels. Deze stapels bestaan
uit vloeibaar amorf afgewisseld met dense lamellen, die op hun beurt zijn
opgebouwd uit kristallijne korrels en een dense, amorfe fractie.
iii
Nomenclature
List of Symbols
◦
2θ Bragg angle
α Degree of conversion −
αs Fraction of semi-crystalline stacks −
β Ratio of overall over local crystallinity −
βhkl Integral breadth of hkl-diffraction peak Å
χc Degree of crystallinity −
∆Hm Melting enthalpy J/g
δ Crack opening m
δ Polydispersity −
² Porosity −
² Strain −
²∗ Yield strain/Strain to failure −
η Stress partitioning factor −
η Viscosity Pa · s
γ Fraction rigid amorphous within semi-crystalline stack −
γLV Surface tension Pa
κ Fraction dense amorphous within the dense pools −
l1 , l2 Characteristic lengths in semi-crystalline stacks Å
λ Wavelength Å
ν Poisson coefficient −
v
vi Nomenclature
φ, (1 − φ) Crystalline or amorphous local volume fraction −
ρ Mass density kg/m3
σ Stress MPa
σ∗ Strength MPa
◦
θ Contact angle
ϕ, (1 − ϕ) Crystalline or amorphous overall volume fraction −
% Electron density e− /cm3
a amorphous
c crystalline
fl flexural
t tensile
A Area m2
a, K Mark-Houwink constants −
B Constant background factor a.u.
b Sample width m
C Constant −
Ca∗ Modified capillary number −
D Lateral dimensions of semi-crystalline stacks Å
d Midspan deflection m
Dhkl Crystallite size in hkl-direction Å
E Young’s modulus GP a
e Crack length m
Eabs Absorbed energy J
Eim Impact energy J
F Force N
G Shear modulus GP a
GIc Corrected crack propagation energy J/m2

Nomenclature vii
GI Crack propagation energy J/m2
h Sample thickness m
I Intensity a.u.
K Correction factor −
K Permeability m2
L Flow distance m
L Span length m
Lp Long period Å
m Weight g
Mm Monomer molecular weight g/mole
Mn Number average molecular weight kg/mole
Mw Weight average molecular weight kg/mole
Os Specific surface area of the phase boundary Å2
P Porod constant a.u.
P Pressure Pa
Q Heat flow W/g
Q Volumetric flow rate m3 /s
R Ideal gas constant, 8.31 J/(mole · K)
s Modulus of scattering angle Å

◦
T Temperature C
t Time s
◦
Tm Melt temperature C
tpol Polymerisation time min
tstir Stirring time s
v Velocity m/s
Vf Fibre volume fraction %
Vm Matrix volume fraction %

viii Nomenclature
vs Superficial velocity m/s

Vi Vacuum level bar
List of Abbreviations
A-UD please refer to page 53
APLC Anionic polymerisation of lactam
B-GF Glass fibres from which the sizing is burnt off
CBT Cyclic butylene terephthalate
CET Cyclic ethylene terephthalate
DAF Dense amorphous fraction
DSC Differential scanning calorimetry
ECA-UD please refer to page 53
GF Glass fibres
GPC Gel permeation chromatography
IDF Interface distribution function
IFF Interface interference function
iso Processed isothermally
LA Liquid activator
LRAT Linear regression of the autocorrelation triangle
PA Polyamide
PBT Poly(butylene terephthalate)
pCBT Polymerised cyclic butylene terephthalate
PET Poly(ethylene terephthalate)
RAF Rigid amorphous fraction
RIM Reaction injection moulding
ROP Ring opening polymerisation
RRIM Reinforced reaction injection moulding
RTM Resin transfer moulding
Nomenclature ix
S-B please refer to page 53

S-UD please refer to page 53
SALLS Small angle laser light scattering
SAXS Small angle X-ray scattering
semi Processed semi-isothermally

SRIM Structural reaction injection moulding
TEM Transmission electron microscopy
TGA Thermogravimetrical analysis
TP Thermoplastic
TP-RTM Thermoplastic resin transfer moulding
TS Thermoset
W-R580 please refer to page 53
WAXD Wide angle X-ray diffraction
-Q Quenched
-S Cooled down slowly

c-pCBT Macrocycle of polymerised cyclic butylene terephthalate
CDP Cyclo-depolymerisation
ED- Entropy driven

IM- Injection moulded
RP- Reprocessed
x Nomenclature
Contents
Preface/Voorwoord i
Abstract/Samenvatting iii
Nomenclature v
Contents xi
1 Introduction 1
2 Literature review 5
2.1 Manufacturing of continuous fibre reinforced thermoplastics 5
2.1.1 Impregnation of fibrous reinforcement . . . . . . . 5
2.1.2 Processing routes . . . . . . . . . . . . . . . . . . . 6
2.1.3 Facilitating impregnation . . . . . . . . . . . . . . 7
2.1.4 Consolidation of preforms . . . . . . . . . . . . . . 9
2.2 Reactive processing of thermoplastic composites . . . . . . 10
2.2.1 Reactive polymer systems . . . . . . . . . . . . . . 10
2.2.2 Processing window . . . . . . . . . . . . . . . . . . 13
2.2.3 Processing methods . . . . . . . . . . . . . . . . . 15
2.3 Liquid moulding processes . . . . . . . . . . . . . . . . . . 16
2.3.1 Resin transfer moulding . . . . . . . . . . . . . . . 16
2.3.2 Structural reaction injection moulding . . . . . . . 18
2.3.3 Thermoplastic resin transfer moulding . . . . . . . 19
2.4 Cyclic butylene terephthalate . . . . . . . . . . . . . . . . 23
2.4.1 Cyclic oligomers and their applications . . . . . . . 23
2.4.2 Production of cyclic oligoesters and their polymeri-
sation . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.3 Properties of CBTr resin and (reinforced) pCBT . 32
2.4.4 Production methods for reinforced pCBT . . . . . 36
2.5 Morphology of poly(butylene terephthalate) . . . . . . . . 38
2.5.1 Semi-crystalline polymers . . . . . . . . . . . . . . 38
2.5.2 Crystalline structure of poly(butylene terephthalate) 42
2.5.3 Influence of fibrous reinforcement . . . . . . . . . . 45
xi
xii CONTENTS
2.6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . 47
3 Problem statement 49
4 Materials and Methods 51

4.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.1.1 Polymers . . . . . . . . . . . . . . . . . . . . . . . 51
4.1.2 Reinforcements . . . . . . . . . . . . . . . . . . . . 52
4.2 Production set-up for thermoplastic RTM . . . . . . . . . 55
4.2.1 Mould set-up . . . . . . . . . . . . . . . . . . . . . 55
4.2.2 Injection set-up . . . . . . . . . . . . . . . . . . . . 55
4.2.3 Processing parameters . . . . . . . . . . . . . . . . 57
4.3 Production methods differing from thermoplastic RTM . . 58
4.4 Mechanical properties . . . . . . . . . . . . . . . . . . . . 60
4.4.1 Tensile and three point bending tests . . . . . . . . 60
4.4.2 Impact tests and damage evaluation . . . . . . . . 60
4.4.3 Interlaminar fracture toughness . . . . . . . . . . . 62
4.4.4 Charpy . . . . . . . . . . . . . . . . . . . . . . . . 63
4.5 Matrix characterisation . . . . . . . . . . . . . . . . . . . 63
4.5.1 Gel permeation chromatography . . . . . . . . . . 63
4.5.2 Thermal Analysis . . . . . . . . . . . . . . . . . . . 66
4.5.3 Transmission Electron Microscopy . . . . . . . . . 67
4.5.4 Light and X-ray scattering . . . . . . . . . . . . . 68
5 Composites: Production and Properties 71

5.1 Production . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.1.1 Production of biaxial composites . . . . . . . . . . 71
5.2 Matrix properties . . . . . . . . . . . . . . . . . . . . . . . 73
5.2.1 Composites produced by TP-RTM . . . . . . . . . 73
5.2.2 Composites produced in the hot press . . . . . . . 78
5.2.3 Overview of matrix properties . . . . . . . . . . . . 81
5.3 Tensile and flexural properties . . . . . . . . . . . . . . . . 82
5.3.1 Theoretical stiffness . . . . . . . . . . . . . . . . . 82
5.3.2 Unidirectional composites . . . . . . . . . . . . . . 84
5.3.3 Effect of interface and matrix properties on flexural
properties of unidirectional composites . . . . . . . 90
5.3.4 Cross-ply composites . . . . . . . . . . . . . . . . . 91
5.3.5 Overview of tensile and flexural composite properties 95
5.4 Drop weight impact and interlaminar fracture toughness . 97
5.4.1 Interlaminar fracture toughness . . . . . . . . . . . 97
5.4.2 Drop weight impact properties . . . . . . . . . . . 98
5.4.3 Overview of impact properties . . . . . . . . . . . 102
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 103
CONTENTS xiii
6 Properties and morphology of pCBT 105

6.1 Crystallisation kinetics of pCBT . . . . . . . . . . . . . . 105
6.2 Processing routes for pCBT . . . . . . . . . . . . . . . . . 108
6.3 Mechanical properties of pCBT . . . . . . . . . . . . . . . 109
6.3.1 Flexural tests . . . . . . . . . . . . . . . . . . . . . 109
6.3.2 Charpy impact tests . . . . . . . . . . . . . . . . . 110
6.3.3 Tensile tests of reprocessed pCBT . . . . . . . . . 111
6.4 Physical properties of pCBT . . . . . . . . . . . . . . . . . 112
6.4.1 Degree of conversion and molecular weight . . . . 112
6.4.2 Degree of crystallinity . . . . . . . . . . . . . . . . 113
6.5 Morphology of pCBT . . . . . . . . . . . . . . . . . . . . . 114
6.5.1 Unit cell level . . . . . . . . . . . . . . . . . . . . . 115
6.5.2 Lamellar structure as revealed by TEM . . . . . . 117
6.5.3 Spherulitic superstructure . . . . . . . . . . . . . . 119
6.6 Matrix modification . . . . . . . . . . . . . . . . . . . . . 121
6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 123
7 Time-resolved X-ray measurements 125

7.1 Synchrotron radiation . . . . . . . . . . . . . . . . . . . . 125
7.2 Materials and Experimental set-up . . . . . . . . . . . . . 127
7.2.1 Materials . . . . . . . . . . . . . . . . . . . . . . . 127
7.2.2 Experimental set-up . . . . . . . . . . . . . . . . . 127
7.3 WAXD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
7.3.1 Data processing . . . . . . . . . . . . . . . . . . . . 128
7.3.2 Determination of the degree of crystallinity . . . . 131
7.3.3 Crystallisation kinetics . . . . . . . . . . . . . . . . 134
7.4 SAXS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.4.1 Data processing . . . . . . . . . . . . . . . . . . . . 138
7.4.2 Assignment of characteristic lengths . . . . . . . . 142
7.4.3 Characteristic lengths and local degree of crytallinity144
7.5 Comparison of SAXS and WAXD results . . . . . . . . . 149
7.5.1 The invariant . . . . . . . . . . . . . . . . . . . . . 149
7.5.2 Crystallisation via a mesomorphic phase . . . . . . 152
7.5.3 Rigid amorphous fraction . . . . . . . . . . . . . . 154
7.5.4 Effect of a third phase on the local crystallinity . . 155
7.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8 Conclusions and Outlook 167
A Production of a leaf spring prototype 171

A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 171
A.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
A.3 Production set-up . . . . . . . . . . . . . . . . . . . . . . 172
A.4 Production . . . . . . . . . . . . . . . . . . . . . . . . . . 172
A.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 175
xiv CONTENTS
B Principles of X-ray scattering 177

B.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 177
B.2 X-ray Scattering . . . . . . . . . . . . . . . . . . . . . . . 177
B.3 Wide-angle X-ray Diffraction . . . . . . . . . . . . . . . . 179
B.4 Small-angle X-ray Scattering . . . . . . . . . . . . . . . . 179
B.4.1 Correlation function analysis . . . . . . . . . . . . 180
B.4.2 Interface distribution function . . . . . . . . . . . . 184
C Thermoplastic Polyurethanes 187

C.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . 187
C.2 Thermoplastic polyuretane (TPU) . . . . . . . . . . . . . 187
C.3 Reversible (de)-polymerisation . . . . . . . . . . . . . . . 188
C.4 Characterisation . . . . . . . . . . . . . . . . . . . . . . . 189
C.4.1 Molecular Structure . . . . . . . . . . . . . . . . . 189
C.4.2 Thermal Analysis . . . . . . . . . . . . . . . . . . . 191
C.4.3 Viscosity . . . . . . . . . . . . . . . . . . . . . . . 192
C.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Nederlandstalige Samenvatting 195
References 211
Curriculum Vitae 227
Publications 228
Chapter 1
Introduction
Polymer composites are an interesting material class since they com-

bine good mechanical properties with a low density resulting in excellent
specific properties. Today, there is a wide variety in polymer compos-
ites, ranging from short fibre reinforced thermoplastics for use in e.g. car
bumpers and household tools to continuously reinforced thermoset com-
posites for high-end aerospace applications.
Independent of the reinforcement type, there are two main classes
of matrices, namely thermoplastic and thermoset matrices. These two
classes differ in polymer structure resulting in different properties and
processing techniques. Thermoplastics consist of entangled long molecu-
lar chains, built-up with primary bonds, whereas only secondary bonds
between the chains exist. These polymers can either by amorphous or
semi-crystalline. Thermoset polymers, on the other hand, consist of a
three dimensional network of primary bonds.
As a result of this structure difference, thermoplastics are able to flow
once heated above their flow temperature, whereas thermosets would,
upon raising temperature, decompose before starting to flow. This leads to
major differences in processing techniques of thermoplastic and thermoset
polymers and their composites.
Thermoplastic polymers are most often first polymerised and then
processed in the molten state to the final product. Thermosets on the
other hand, are polymerised during the production since, once completely
cured, their shape and properties cannot be altered anymore. One of the
major advantages of thermoplastics compared to thermosets is therefore
the possibility to post-shape and reprocess them, consequently opening
a new window of recycling opportunities. Other processing advantages
of thermoplastics are related to the absence of a chemical reaction lead-
ing to shorter cycle times and avoidance of emission of volatile organic
compounds.
Property-wise, improved toughness resulting in better impact proper-
1
2 Introduction
ties is a major advantage of thermoplastics. One of the drawbacks is the

smaller temperature range in which thermoplastics can be used. More-
over, they suffer from creep, therefore, thermosets are the material of
choice for high-temperature applications.
In the current composite market, thermoplastic composites have a
share of only 30%1 . From this 30%, the major part encompasses short
fibre reinforced thermoplastics. The annual growth of the thermoplastic
composite market is however larger than the growth of the thermoset
market. This growth can be partially explained by the overall increasing
interest in continuously reinforced thermoplastics.
The arrear of continuously reinforced thermoplastics compared to ther-
mosets is partially caused by the difficulty in impregnating the fibrous
reinforcement with the highly viscous thermoplastic melt. Compared to
thermoset monomers, the viscosity of a thermoplastic melt is several or-
ders of magnitude higher. This high viscosity is less of an issue in short
fibre composites. In these composites, high pressure and shear stresses
are used to blend fibres and matrix. This is however not possible for con-
tinuously reinforced thermoplastics since the fibre orientation would be
distorted and fibre breakage may occur.
Due to the large difference in viscosity, the manufacturing routes for
continuously reinforced thermoplastics and thermosets are completely dif-
ferent. One way to deal with the high matrix viscosity is to reduce the
matrix flow length. This can be achieved by intimate mingling of ma-
trix and fibres before processing. This mingling process often results in
a rather expensive intermediate product. Moreover, since the subsequent
processing is completely different from thermoset composites, the transi-
tion from one material class to another remains difficult.
Another option to deal with the high matrix viscosity is simply to
reduce this viscosity during processing by introducing the polymerisation
step into the production process. Low viscous prepolymers hence im-
pregnate the fibrous reinforcement, after which they are polymerised to
obtain adequate matrix properties. If the reduction of the viscosity is
large enough, traditional thermoset production techniques, such as Resin
Transfer Moulding, can be used to produce thermoplastic composites.
Not all thermoplastic polymer systems are suitable for an in-situ poly-
merisation process and recently, interesting developments have renewed
the interest in in-situ polymerisation for the production of structural com-
posites. One of these developments is the (upcoming) commercialisation
of cyclic oligomers such as CBTr resin. CBTr resin consists of low
viscous cyclic oligomers of the engineering thermoplastic poly(butylene
terephthalate). This resin is a suitable candidate for thermoplastic resin
transfer moulding and will be used throughout this study.
1 source: JEC-Composites, January 2005
Introduction 3
Even though the potential of CBTr resins for use in traditional ther-
moset composites’ production techniques is often stated, little is known
about the final composite properties and the influence of the production
process on the matrix properties. For these reasons, this study will char-
acterise the in-situ polymerisation production process for continuously
reinforced thermoplastics with cyclic butylene terephthalate oligomers.
First, a literature review is presented dealing with the different aspects
of the current state of the art, ranging from manufacturing of continuously
reinforced thermoplastics, liquid moulding processes to cyclic oligomers
and the morphology of poly(butylene terephthalate). From this literature
review, a problem statement, which further outlines this study, will be
formulated.
The main part of this work consists out of the manufacturing of con-
tinuously reinforced CBTr resin composites using thermoplastic RTM
followed by the characterisation of the composite properties with empha-
sis on the properties of the in-situ polymerised matrix. Subsequently, the
effect of the processing route on the properties and morphology of the ma-
trix (without reinforcement) will be studied. Finally, the crystalline struc-
ture of the matrix during its formation is investigated by time-resolved
X-ray measurements.
4 Introduction
Chapter 2
Literature review
2.1 Manufacturing of continuous fibre rein-

forced thermoplastics
Thermoplastic composites have property-wise significant advantages over
their thermoset counterparts, including improved toughness and advanced
recycling options. The major drawback, however, is their melt viscosity
which hampers fibre impregnation. At conventional processing conditions,
the melt viscosity of thermoplastics is around 50-2000 Pa·s whereas the
viscosity of thermosets typically does not exceed 50 Pa·s [2]. Since the
viscosity has a profound influence on the impregnation of the fibrous rein-
forcement, manufacturing of thermoplastic composites differs from ther-
moset composite processing.
2.1.1 Impregnation of fibrous reinforcement

In order to successfully produce composites, it is crucial to fully impreg-
nate the fibrous reinforcement, not only to obtain a void-free part but
also to ensure stress transfer between fibres and matrix. There are two
levels of matrix flow in composite production, the macro-flow, which is
the flow in between the fibre bundles and the micro-flow, the flow inside
the fibre bundles. Although both flows are crucial, only the macro-flow is
often considered in a first approximation.
The macro-flow of the matrix through the fibrous reinforcement can
be described by Darcy’s law. The three-dimensional form of this law is
given in Equation 2.1, with − →
v the velocity, K the permeability tensor, η
→
−
the viscosity and ∇ P the pressure gradient. The brackets indicate the
volume averaging operation.
5
6 Literature review
KD − →E
h−
→
vi=− ∇P (2.1)
η
Q dL K1D ∂P
=² =− (2.2)
A dt η ∂x
ηL2
timp = ² (2.3)
∆P K1D
The primary flow in composites’ processing is mostly one-dimensional,
Equation 2.2 with Q the volumetric flow rate, A the cross-sectional area,
² the porosity, L the flow distance in the x-direction, t the time, K1D
the one-dimensional permeability and ∂P ∂x the pressure gradient in the x-
direction. Equation 2.3, with timp the total impregnation time, results
from the integration of Equation 2.2.
It is clear that the impregnation time depends on the fibre preform,
more specifically on its permeability, on the final part properties, namely
the porosity (or fibre volume fraction) and the part’s dimensions, on the
applied pressure gradient and on the matrix viscosity.
For thermoplastic matrices with a high melt viscosity, impregnation
time is increased substantially if the processing techniques for low viscous
thermosets would be applied. In order to limit the impregnation time and
hence the total cycle time, the pressure gradient can be increased. This
increase is however limited since large pressure gradients can distort the
fibre orientation and even cause fibre breakage, which can be the case with
hot melt impregnation. Therefore, there are two main processing routes
for the production of continuous fibre reinforced thermoplastic composites
at intermediate pressures [3, 4], reducing the required flow distance or
reducing the matrix viscosity during impregnation.
Reinforcing thermoplastics with discontinuous (or short) fibres poses
less problems and short fibre reinforced thermoplastics are hence more
commonly used. The high melt viscosity of the matrix can be overcome
by ‘mixing’ fibres and matrix in e.g. an extruder, introducing high shear
forces that allow the fibres to be distributed evenly into the matrix. For
these composites, fibre length and orientation are often of lesser impor-
tance, resulting, of course, in inferior mechanical performance compared
to continuous fibre reinforced composites.
2.1.2 Processing routes

Figure 2.1 shows a schematic overview of the manufacturing options at
intermediate pressures for thermoplastic composites reinforced with con-
tinuous fibres. The reduction of the flow distance is usually obtained by an
intimate mingling process whereas the viscosity reduction can be achieved
by either solvent impregnation or in-situ polymerisation.
Thermoplastic precursors
Polymerised thermoplastics
Reinforcement
Intimate Viscosity
mingling reduction
Film Hybrid Powder Solvent In-situ

stacking yarns impregnation impregnation polymerisation
Preforms
Final composite part
Figure 2.1: Processing routes for thermoplastic composites with

a continuous reinforcement
Facilitating impregnation usually results in the production of preforms

which subsequently need to be consolidated to form the final composite
part. For economic reasons, it is however preferable not to introduce this
intermediate step and directly produce the final part from its constituents.
2.1.3 Facilitating impregnation

Intimate mingling processes
The aim of intimate mingling processes is to reduce the distance that the
resin is required to flow in order to fully impregnate the fibrous reinforce-
ment. Three main techniques are employed [3], namely film stacking, yarn
and fabric hybridisation, and powder impregnation.
Film stacking In the film stacking process, alternate layers of (thin)

polymer sheets and reinforcement are stacked together. Pressure and heat
are then applied to allow the matrix to flow transverse to the fibre bundles
hence impregnating them and, depending on the pressure-temperature
cycle, fully consolidating the part. A very similar process exists for ther-
moset composites with the difference that the matrix sheets are not fully
cured.
8 Literature review
a b c d
Figure 2.2: Different types of hybrid yarns (a) commingled, (b)

co-wrapped, (c) core-spun, (d) stretch-broken, adapted from [6]
This technique is quite simple and less expensive than the hybrid al-
ternative [5] and therefore often used to produce flat panels for research
purposes. Complex shapes are however difficult to achieve since the stacks
are not very drapeable. Moreover, the mingling intimacy is rather low and
therefore this technique is not suited for very dense fabrics and high fibre
volume fractions.
Hybrid yarns and fabrics The second method of intimate mingling

of fibres and matrix involves the spinning of thermoplastic fibres. These
thermoplastic fibres can then be blended with the reinforcing fibres at
the fabric or filament level. Although the use of hybrid fabrics, where
matrix en reinforcing yarns are placed next to each other, is economically
attractive, the flow length is increased and this technique is therefore less
frequently employed [3].
Various techniques exist to produce hybrid yarns (blending at the fila-
ment level). These all aim to give a uniform distribution of the matrix and
the reinforcing fibres without inflicting damage on the latter. Svensson
et al. [4] have reviewed the manufacturing of thermoplastic composites
from hybrid yarns. Figure 2.2 schematically presents the different types
of hybrid yarns.
The commingled yarns, Figure 2.2a, provide the most uniform distri-
bution leading to a required matrix flow distance of around 20-40 µm.
Co-wrapping, Figure 2.2b, on the other hand, results in a better protec-
tion of the reinforcing fibres during manufacturing and further (textile)
processing. The distribution quality is however diminished.
Core-spinning, Figure 2.2c, leads to yarn properties very comparable
to those of the co-wrapped yarns with the additional advantage of being
extremely supple. In the last technique, stretch-breaking, Figure 2.2d,
the reinforcing fibres are stretch-broken to a predetermined length and
are then twisted into a yarn. Although the twist gives rise to an improved
flexibility, the fibres are no longer continuous and have an imperfect align-
ment leading to a reduction in stiffness and strength.
When using hybrid yarns, fibre wet-out is postponed until the con-
solidation step, which then of course requires more efforts compared to

preconsolidated preforms [5].
Powder impregnation During powder impregnation, the fibres are

initially spread in order to facilitate the deposition of the impregnating
thermoplastic powder inside the yarn. The dry powder particles adhere
electrostatically to the fibres but since these forces are relatively weak,
the powder is subsequently stabilised to the fibre surface by heating the
system to sinter the powder particles to the fibres [3, 6].
Powder impregnation is a dry process, avoiding solvents and water,
which is a major incentive for its exploration. Other advantages are the
high extent of mixing if the fibre tow is spread sufficiently, flexible tows
that facilitate further processing and the low stress history, minimising
fibre damage [5]. Grinding of polymers is however not always easy and
may limit the obtainable particle size.
Reduction of the viscosity

Solvent impregnation Solvent impregnation requires the polymer to
be dissolved in a solvent to reduce the viscosity during impregnation.
There are however a number of disadvantages to this process. First of all,
semi-crystalline polymers are often very difficult to dissolve and therefore
require aggressive and often toxic solvents. As a result, solvent impreg-
nation is primarily being used with amorphous thermoplastics [3, 7].
Moreover, the complete removal of the solvent after impregnation can
be very difficult. Remaining solvent can not only cause voids when it
evaporates during subsequent processing, but can also compromise further
processing and the in-service performance.
In-situ polymerisation Contrary to the processing routes previously

described, not the thermoplastic polymer but its low-viscous precursors
impregnate the fibrous reinforcement. Similar to the production of ther-
moset composites, a chemical reaction takes place during and after im-
pregnation to fully develop the matrix’ structure and properties. The re-
active processing of thermoplastics and their composites will be discussed
in detail in Section 2.2.
Although preforms can be produced with this process, one of its major
advantages is the possibility to exclude this intermediate step from the
process and directly manufacture the final part.
2.1.4 Consolidation of preforms

In order to make the final composite part from preforms, three major
steps are necessary, heating the preform thereby melting the matrix, con-
solidation and compaction in order to obtain a void-free composite part
10 Literature review
and solidification of the final part [5]. During consolidation, an external

pressure is applied to the part to squeeze out entrapped air, suppress voids
and uniformly disperse the fibres. Unless it is amorphous, a thermoplastic
polymer crystallises upon solidification. The matrix properties are hence
strongly dependent on the cooling history.
An overview of the models for consolidation of the different preforms
is beyond the scope of this literature review, therefore, the interested
reader is referred to following references and the references therein [4, 8–
10]. There are a number of processes that are used to consolidate ther-
moplastic preforms including pultrusion [11–13], filament winding [11, 13]
and compression moulding [7, 11].
2.2 Reactive processing of thermoplastic com-

posites
The main aim of reactive processing of thermoplastic composites is the di-
rect impregnation of the fibrous reinforcement without intermediate steps.
There are three conditions that have to be fulfilled in order to apply direct
impregnation of a fibre bed by a liquid matrix [14, 15]:
• The matrix viscosity during the impregnation stage is very low,

<1 Pa·s.
• Once the fibre bed is fully impregnated, the matrix can be solidified
chemically or physically in a sufficiently short time.
• The final matrix has high enough physical properties to transmit

good mechanical stability to the composite part.
The viscosity requirement can be reached by introducing a chemical

reaction into the production process and hence impregnating the rein-
forcement by low viscous monomers or oligomers. Most research efforts
have focussed on reactive pultrusion and liquid moulding processes with
a limited amount of available polymer systems.
2.2.1 Reactive polymer systems

Polyamides are most commonly used in reactive thermoplastic processing
[16]. Their semi-crystalline nature is responsible for mechanical proper-
ties which are sufficient for use in structural applications. Thermoplastic
polyesters on the other hand have properties very similar to those of poly-
amides and the recent development of the low viscous CBTr resin (cyclic
butylene terephthalate) has launched a renewed interest in the reactive
processing of thermoplastics.
2.2 Reactive processing of thermoplastic composites 11
Polyamides
Polyamides can be synthesised by two major production routes, polycon-
densation, Equation 2.4 and anionic ring-opening polymerisation, Equa-
tion 2.5. The major disadvantage of the polycondensation process is the
release of water which excludes the use of closed mould processes.
O O
NH2 CH2 x
NH2 + OH C CH2 C OH
x-2
diamine diacid H O O
NH2 CH2 N C CH2 C OH + H2O
x-2
(2.4)
O O H O O
C N + C N C N C CH2 x NH
CH2 x CH2 x CH2 x
O H
catalyst monomer C N
+ n+1
CH2 x
O O O H O
C N + C N C CH2 x
N C CH2 x
NH2
n
CH2 x CH2 x
(2.5)
Different polyamides, PA6 and 12, starting from respectively capro-

lactam and laurolactam, and copolyamides (PA6,6) can be produced ac-
cording to these reaction schemes. In order for the anionic ring-opening
polymerisation of lactams (APLC) to proceed relatively fast, both an ac-
tivator (or initiator) and a catalyst are necessary.
Traditionally, anionic polymerisation is initiated by mixing together
similar volumes of two batches of lactam pre-blended with an activator
and a catalyst respectively. Both compounds have a short pot-life as they
slowly polymerise in the tanks [16]. To overcome this problem, a liquid
activator system (LA), named GrinolitTM was developed by EMS-Chemie
[14, 15, 17]. The liquid activator contains both activator and catalyst dis-
solved in a solvent.
Because neither the catalyst nor the activator is in contact with the
monomer, both components (molten lactam and LA) are stable prior to
mixing and have hence a prolonged pot-life compared to the pre-blended
lactam compounds. Mixing of the liquid activator system however requires
a more advanced mixing system since the mixing ratio is no longer 50:50
but 98.5:1.5 to 95:5 depending on the preferred reaction speed [18].
Since the APLC is sensitive to both moisture and oxygen, it is very

important to produce the composites in an inert atmosphere. Given that
fibre sizings contain reactive groups, the compatibility with the reaction
needs to be investigated. Most commercially available glass fibre sizings
however inhibit the polymerisation reaction [15, 18], hence research has
initially focussed on carbon fibre reinforcements.
Comparing PA6 to PA12, it is interesting to note that although capro-
lactam is cheaper and the mechanical properties of PA6 are somewhat bet-
ter, the equilibrium conversion of PA12 is higher (2% versus 5-10%) and
its dimensional stability in moist surroundings is superior [18]. Copoly-
merisation of both to combine these properties has also been investigated
[19].
Thermoplastic polyesters
Commercial thermoplastic polyesters are industrially produced by a trans-

esterification reaction between glycols and terephthalates. During this
reaction, a limited amount of cyclic oligomers is formed which are in equi-
librium with the polymer (Section 2.4.1). These cyclic oligomers have a
viscosity which is substantially lower than the viscosity of the polymer.
Moreover, these oligomers can again undergo an entropically-driven ring-
opening polymerisation to form the polymer. Hence, they are excellent
candidates for reactive processing.
In 1999, Cyclics Corporation was founded after buying the patents con-
cerning the cyclic oligomer technology from General Electric. Although
the first patents dealt with polycarbonates, thermoplastic polyesters, more
specifically CBTr resin, i.e. cyclic butylene terephthalate, which poly-
merises to form the well-known engineering thermoplastic poly(butylene
terephthalate), was the main focus of this company. Since this resin sys-
tem will be used during this study, it is described in more detail in Sec-
tion 2.4.
Today, the main disadvantage of cyclic oligomers is their price, which is
quite high compared to other engineering thermoplastics. For this reason,
Yan et al. [20] investigated the in-situ solid-state polymerisation (IS-SSP)
of poly(ethylene terephthalate) based on the polycondensation reaction.
Their processing route is in principle applicable to all thermoplastic poly-
mers synthesised by polycondensation. A major issue for this process,
however, is the removal of the polycondensation by-products and the very
large polymerisation times.
The production of biodegradable reinforced polyesters for medical ap-
plication by reactive processing was studied by a group of researchers
at the University of Nottingham [21–23]. They used a biodegradable
poly-caprolactone (PCL) matrix reinforced with biodegradable glass fi-
bres. The viscosity of the monomer is only 1.07 mPa·s compared to
12650 Pa·s for the molten PCL [22] but its major disadvantage is the
sensitivity to moisture.
Other reactive polymer systems

Most research on reactive processing of thermoplastic composites has fo-
cussed on polyamides and thermoplastic polyesters. However, some other
polymer systems have been investigated.
Pini et al. [24] investigated the perspectives of polyphtalamide (PPA)
for reactive moulding applications. This polymer fills the gap in between
engineering polymers and high-performance polymers such as poly(ether
ether ketone) (PEEK). EMS Chemie provides the oligomers which can
be further polymerised. The main disadvantages of this polymer system
are the formation of water during the polycondensation reaction and the
rather high viscosity of the oligomers (> 50 Pa·s).
A different type of reactive processing can be achieved with certain
thermoplastic polyurethanes (TPU). This process was first described in a
U.S.patent from Edwards et al. [25]. These TPU’s undergo a reversible
depolymerisation upon heating, resulting in a significant viscosity drop,
which is discussed in more detail in Appendix C.
2.2.2 Processing window

An important aspect of reactive processing is the processing window.
Compared to traditional processing of thermoplastics which mainly con-
sists out of three steps, namely heating to melt the polymer, consolidation
and cooling to allow for solidification, the reactive process also incorpo-
rates a polymerisation step which strongly influences the processing win-
dow.
A crucial step in the manufacturing of composites is the impregnation
of the reinforcement. As mentioned above, the direct impregnation should
be accomplished before the viscosity of the reactive mixture reaches 1 Pa·s,
seriously limiting the available time window for impregnation. The im-
pregnation window is hence defined as the time for the reactive mixture
to reach 1 Pa·s. Although this limit is rather a rule of thumb than a strin-
gent condition [26], it does illustrate the dependence of the impregnation
window on the processing parameters.
Figure 2.3 shows the viscosity as function of time and temperature
for APLC12 with the liquid activator system. The lower temperature
boundary for the processing is of course the melting point of the monomers
(154◦ C). The viscosity evolution is determined by the reaction kinetics
and since the reaction speed increases with increasing temperature, the
impregnation window decreases.
The reaction kinetics are of course not only determined by the tem-
perature but also by the kind and amount of activator/catalyst system
Figure 2.3: Viscosity as function of time and temperature for

APLC12 with liquid activator [17]
used. Moreover, a compromise must be found between the impregnation

window and the overall cycle time, which depends on the time to reach
equilibrium conversion. Whereas the impregnation window is preferred
to be as large as possible (slow reaction kinetics), the overall cycle time
should be kept as short as possible (fast reaction kinetics).
van Rijswijk et al. [27, 28] very recently investigated different activa-
tor/catalyst combinations for APLC6. They found that depending on the
production process (casting versus vacuum infusion) and hence depend-
ing on the impregnation process, a different set of activator/catalyst is
optimal. For the production of continuously reinforced composites, poly-
merisation should be delayed to allow for impregnation but after its onset,
polymerisation should proceed very rapidly.
Besides the impregnation window and the polymerisation time, the
manufacturing is also dependent on the solidification of the matrix. When
the matrix is semi-crystalline, this solidification process includes crystalli-
sation.
For APLC12 with the liquid activator system, the time-temperature-
transformation diagram (TTT-diagram, Figure 2.4) illustrates the overall
processing window. As mentioned before, the melting point of the pre-
cursors provides the lower temperature boundary. The melting point of
the final polymer on the other hand defines an interesting turning point
in terms of processing.
Processing below the polymer melting point allows for solidification at
the processing temperature and hence isothermal processing. When for
example a closed mould process is used, isothermal processing is a sig-
nificant advantage since it avoids the time- and energy-consuming ther-
mal mould cycle that cannot be avoided when polymerisation takes place
Figure 2.4: Time-temperature-transformation diagram for

APLC12 with liquid activator [17]
above the polymer melting point.

For APLC, several authors have seen interesting effects of crystallisa-
tion on the polymerisation, [17, 18, 29]. Firstly, crystallisation strongly de-
creases the reaction rate, Figure 2.4, since some active centres are caught
in the crystalline regions, thus decreasing the overall reactivity. The sec-
ond effect is an increase in the final degree of conversion. The crystalline
fraction is not involved in the polymer-monomer equilibrium and as only
the polymerised fraction can crystallise, the final concentration of mono-
mer is shifted to a lower content.
The downside of the isothermal process is the relatively low reaction
temperature and hence an increase in polymerisation time. For CBTr
resin and APLC6, this increase is not detrimental, for APCL12 however
isothermal processing is excluded since it leads to unacceptable processing
times.
2.2.3 Processing methods

Reactive thermoplastic pultrusion
Pultrusion is a well-known production technique for thermoset composites
and is mostly applied for medium to high-volume applications. Tradition-
ally, the reinforcement (often unidirectional) passes through a resin bath
after which the impregnated fibres are pulled through a heated die. In-
side this die, the matrix is subjected to a temperature-pressure cycle to
allow for consolidation and curing [12]. This resin bath is not often used in
thermoplastic pultrusion since the time window for impregnation is rather
small after mixing the components.
Botelho et al. [30–32] however overcame this problem by passing the
fibres along the interface of two immiscible solvents each containing one
reaction compound (interfacial polycondensation). The impregnated tows
were later on compression moulded and compared to commercially avail-
able prepregs showing an increased void content due to volatile reaction
products and low molecular weight polyamide [31].
In order to avoid an open resin bath, the resin is sometimes injected
into the fibre preform as it enters the pultrusion die. This allows for
the in-line mixing of the matrix, ensuring a low initial viscosity and the
option to process in an inert atmosphere [12]. This process is referred to
as reaction injection pultrusion (RIP).
Reaction injection pultrusion of PA6 was studied by Ning et al. [33, 34]
and Cho et al. [35, 36], whereas the reaction injection pultrusion of PA12
with the liquid activator system is investigated by Luisier et al. [18, 37].
Each of these researchers investigated the process parameters for their
specific activator/catalyst combination, showing that the die design and
pulling speeds strongly depend on the system used. The mechanical prop-
erties of the resulting composites also strongly depend on the processing
parameters since these govern not only the degree of conversion but also
the matrix crystallinity.
CBTr resin is also an interesting material for use in thermoplastic
pultrusion [38], to our knowledge however, no systematic study on this
has been published.
One of the major advantages of the reactive pultrusion process is the
possibility to have a non-isothermal process without sacrificing the total
manufacturing time. This can be accomplished by introducing different
temperature zones into the die.
Thermoplastic liquid moulding

Although the continuous pultrusion process is very cost-effective and al-
lows for high production speeds, it is limited to rather simple shapes. In
contrast, liquid moulding processes are not continuous, but allow for com-
plex shapes and the use of fibre preforms. Thermal cycling of the mould,
on the other hand, significantly increases manufacturing times, therefore,
isothermal processing is preferred.
In the next section, (thermoplastic) liquid moulding is discussed more
elaborately.
2.3 Liquid moulding processes

2.3.1 Resin transfer moulding
Resin transfer moulding (RTM) is the most common liquid moulding tech-
nique for manufacturing structurally capable thermoset composites. The
process uses a closed mould into which dry reinforcement is placed. After
2.3 Liquid moulding processes 17
placement of
fibre preform
compaction of
fibre preform
resin
injection
resin reaction
and solidification
demoulding
Figure 2.5: Schematic overview of RTM process
mould closure, liquid resin, previously mixed with a hardener, is injected

into the mould to impregnate the reinforcement. The injection pressure
varies from 0.1 to 1 MPa and the viscosity of the injected resin should be
below 1 Pa·s and it is preferred to be below 0.5 Pa·s [11]. The RTM pro-
cess is shown schematically in Figure 2.5. The injection can be assisted
by vacuum (VARTM).
A critical aspect of thermoset RTM is to ensure that the mould is com-

pletely filled before the final stage of crosslinking starts since this crosslink-
ing involves a viscosity rise. Depending on the resin, the crosslinking oc-
curs at ambient temperature or at an elevated temperature. Figure 2.6
shows the interactions in RTM and other liquid moulding techniques be-
tween the processing variables, materials and process issues and other
moulding problems. Although these issues were set forward for thermoset
resins, they equally apply to thermoplastics if curing is changed to poly-
merisation. Thermoplastic resin transfer moulding (TP-RTM) will be
discussed in more detail in Section 2.3.3.
A production process very similar to RTM is vacuum infusion (VARI).

In this process, the resin is infused by solely applying vacuum to the outlet,
so no overpressure is used. Contrary to RTM, this processes uses a single
sided mould, relying on a vacuum bag or a flexible upper mould to provide
compaction and sealing. The tooling cost is hence reduced.
Processing Material Process Moulding

variables issues issues problems
Cure Reaction Low degree

temperature kinetics of cure
Initial Resin Nonuniform

mould/fibre curing cure
temperature
Resin Viscosity Incomplete

injection changes filling
temperature
Resin Mould Dry-spot

injection filling formation
pressure
Resin Fibre Void
injection wetting formation
technique
Preform
Fibre permeability
structure
Figure 2.6: Liquid moulding interactions among processing vari-
ables, material and process issues and moulding problems,
adapted from [41]
2.3.2 Structural reaction injection moulding

Reinforced reaction injection moulding (RRIM) and structural reaction
injection moulding (SRIM) are processes derived from reaction injection
moulding (RIM). Sweeney [42] gives a very limited definition of the RIM-
process: reaction injection moulding is a process that moulds parts by
carrying a polymerisation reaction in the mould itself.
At the start of a RIM cycle, two low viscous liquids are mixed by
high-pressure impingement. The mixing takes places at pressures around
10-40 MPa [11]. This is in contrast with RTM, where the injection pressure
lies between 0.1 and 1 MPa. The pressure within the mould of the RIM-
process is much lower, below 1 MPa. The reaction that takes place during
the process is initiated by mixing the components, not by raising the
temperature.
The RIM-process is mostly conducted with polyurethanes. At the end
of the eighties, however, NylonRIM was being developed [16, 42]. Nylon-
RIM is mostly used to make block-copolymers, but can also be used to
make a pure polyamide part, although this is not often the case.
In RRIM, fibres are injected along with the reacting components. Be-
2.3 Liquid moulding processes 19
Table 2.1: Overview of liquid moulding processes showing their

main characteristics (matrix material; matrix reactivity; method
of initiating the reaction; type of fibrous reinforcement; pressure)
matrix reactivity initiation

fibres P
TS TP high low mixing T↑
RIM • •∗ • • none 10→40 MPa∗∗
RRIM • •∗ • • short 10→40 MPa∗∗
SRIM • •∗ • • cont. 10→40 MPa∗∗
RTM • • • cont. 0.1→1 MPa
VARI • • • cont. -0.1→0 MPa
TP-RTM • • • • cont. -0.1→1 MPa
∗ these methods are often considered typical thermoset processing techniques since in
the class of thermoplastics, only polyamides are currently used
∗∗ pressure during mixing of components, mould pressure below 1 MPa
cause the viscosity of the injected mixture is limited, the fibre length does
not exceed 0.5 mm so only short fibre reinforced composites can be made
using RRIM [11]. The second alternative method is SRIM. This method is
a combination of RTM and RIM. The preform is placed inside the mould
before the liquid components are injected. Compared to RTM, the com-
ponents used have a higher reactivity so that the cycle time is shorter but
due to the rising viscosity, the parts that can be manufactured are also
smaller. The injection pressures are comparable to those used in RTM
but the injection speed is higher (seconds versus minutes), which means
that lower fibre volume fractions are used and more voids are encountered
in the final part [11, 41].
The quality and properties of the composites formed by SRIM depend
on how effectively placement of the fibres is retained in the mould cavity
during mould filling, and how efficiently the reactant mixture wets-out the
individual fibre filaments. Both factors depend crucially on the viscosity
of the reactant mixture, which increases rapidly as polymerisation takes
place even in the relatively short time required to fill the mould [43].
Even though it is feasible to use e.g. the anionic ring-opening poly-
merisation for structural reaction injection moulding, it is still considered
to be a thermoset production technique and efforts to use thermoplastics
have ceased in the mid-nineties. Research efforts on resin transfer mould-
ing of thermoplastic however have recently been revitalised, partially due
to interesting developments in the thermoplastics industry.
2.3.3 Thermoplastic resin transfer moulding

Thermoplastic resin transfer moulding (TP-RTM) has received renewed
attention due to the development of both the liquid activator system for
the APLC and the commercialisation of cyclic oligoesters. The major

differences with thermoset RTM are the higher processing temperatures
(above the prepolymer’s melting point) and the larger reactivity of the
polymer systems. Since the reaction is also initiated by the mixing of two
components, TP-RTM might also be called TP-SRIM. Table 2.1 gives an
overview of the liquid moulding processes and their main characteristics.
Besides the requirements for direct impregnation, described in Sec-
tion 2.2, the closed mould process imposes extra prerequisites.
• The reaction should proceed without the formation of any by-products
[17].
• Large exotherms should be avoided since they might create hot-spots
and through the thickness property variations [44].
The first prerequisite is essential for the success of TP-RTM since by-
products can induce voids upon vaporisation or alter the matrix and in-
terface properties. The second prerequisite, however, is less stringent but
the lack of an exotherm is nevertheless advantageous.
Most research efforts concerning TP-RTM have focussed on the charac-
terisation of the unreinforced polymer systems, focussing on the processing
windows and on the mechanical properties of the produced composites.
The processing window is of course different for each of these systems and
was discussed above for the polyamides.
Table 2.2 shows some of the mechanical properties for reinforced poly-
amides produced with thermoplastic RTM. These properties are for woven
composites and hence fibre dominated. In order to compare these proper-
ties, they were normalised to a fibre volume fraction of 50%. Looking at
the PA12 composites, it is clear that the results differ between the differ-
ent authors. Nevertheless, Connor et al. [14] and Mairtin et al. [15] obtain
results similar to the results obtained from testing compression moulded
commingled yarns.
Mairtin et al. [15], however, also noticed large variations between the
plates produced, but no attempt was made to investigate its cause. More-
over, the compression strength of the plates produced with TP-RTM was
significantly lower than the compression moulded commingled yarns. This
was attributed to poor fibre-matrix adhesion. The PA6 composites pro-
duced by Gittel et al. [45, 46] are also inferior to what is theoretically
expected, considering that PA6 has a higher stiffness than PA12. The
large variations between authors might be attributed to the sensitivity
of the polymerisation process to moisture and oxygen, leading to poor
matrix properties and possibly a relatively high void content.
Table 2.2: Mechanical properties of reinforced polyamides (woven reinforcements) produced by thermoplastic liquid
moulding, modulus and strength as cited and normalised to a fibre volume fraction of 50%
∗
reference matrix fibre Vf E (GPa) σ ∗ (MPa) E50 (GPa) σ50 (MPa)
Connor [14] APLC12 carbon 45 50 673 55.6 748
Rosso [47] APLC12 carbon 54 50.8 651 47.0 603
Rosso [47] PA12a carbon 54 51.8 512 48.0 474
Mairtin [15] APLC12 carbon 54 57 779 52.8 721
Mairtin [15] PA12b carbon 56 63 788 56.3 704
Greaney [19] APLC12 carbon 60 56 768 46.7 640
2.3 Liquid moulding processes
Greaney [19] 90/10 APLC6/12c carbon 60 55 685 45.8 571

Greaney [19] 95/5 APLC6/12c carbon 60 62 774 51.7 645
Gittel [45, 46] APLC6 carbon 37 34 430 45.9 581
Pillay [48] APLC6d carbon 50 65 822 65 822
Gittel [45, 46] APLC6 glass 60 26 390 21.7 325
a
film stacking of PA12 sheets and carbon fabric, autoclave processed
b compression moulded from commingled yarns
c matrix is copolymer of PA6 and PA12 with the given ratios
d two batch system compared to liquid activator system for other results
21
In all the papers, the sizing issue is mentioned, indicating that opti-
mising the amount and type of sizing is crucial. Moreover, only specific
types of carbon fibres could be used for APLC6 and 12 and even though
Gittel et al. [45] report data for glass fibre reinforced PA6, later reports
all mention difficulties in finding compatible glass fibres [18, 29].
Rosso et al. [47] have studied the interface properties of carbon fibre re-
inforced APLC12 and showed that fibres with a standard oxidative treat-
ment have in general a better adhesion than fibres with bot a standard
oxidative treatment combined with a PA12 compatible sizing. Moreover,
they compared the macromechanical properties to a film stacked PA12
composite and found that the interface related properties, such as the
interlaminar shear strength, were somewhat better, but the impact prop-
erties are around 20% lower, which might be related to the improved
interface.
During the course of this study (2001-2005), more researchers have
looked into TP-RTM. Pillay et al. [48] and van Rijswijk et al. [27–29, 49]
have focussed on the combination of APLC6 with the conventional two
batch system and the VARI process. For this system, their major concern
was to find a suitable catalyst/activator system to enlarge the time win-
dow for impregnation. Some attempts have been made to optimise a glass
fibre sizing, but to our knowledge, this work is still in progress. The matrix
properties within the composite were also not studied intensively although
they expect a higher degree of crystallinity than for melt-crystallised PA6
[29, 48].
The composites research group at EPFL have on the other hand con-
tinued their work on the APLC12 and the LA system with carbon fibre
reinforcements. They have studied more in detail the void formation in
this system. They found that the inert N2 -atmosphere in which the mono-
mer is kept is responsible for creating voids and therefore degassing is a
better solution [50, 51]. Moreover, the formation of macro- or micro-voids
depends on the modified capillary number, Ca∗ , defined in Equation 2.6
with η the viscosity, vs the superficial velocity, γLV the surface tension
and θ the contact angle [50, 52].
ηvs
Ca∗ = (2.6)
γLV cos θ
Below a critical capillary number, macro-voids are promoted whereas

above this critical value, micro-voids are more abundant. The lowest over-
all void content is reached when processing takes place at conditions that
coincide with this critical capillary number [51]. Furthermore, they con-
cluded that the capillary pressure cannot be ignored in TP-RTM since the
applied pressure is much smaller than in thermoset RTM and therefore,
capillary pressure can account for 15% of the driving force for impregna-
tion [51, 52].
2.4 Cyclic butylene terephthalate 23
Since the void content for TP-RTM without degassing the monomer
and applying any post-pressure was rather high (>10%), it was considered
as a preform production process, after which the preforms were stamp-
formed [53–55]. Since one of the advantages of RTM, namely the possi-
bility to produce complex parts is hence redundant, a set-up for a more
pultrusion-like continuous reactive impregnation line was envisaged [55].
For a European project named TECABS (Technologies for carbon fibre
reinforced modular automotive body structures), TP-RTM with APLC12
was considered as an option to manufacture a floor pan, which is a complex
3D part. Although the feasibility of producing such a complex part was
clearly shown [56], a cost comparison with the thermoset part indicated
the major drawback of the matrix system, namely the need to thermal
cycle the large mould. This increased the total cycle time from 600 to
800 s, leading to an increased cost of 12-25% [57].
Thermoplastic RTM with CBTr resin has been often mentioned in
general papers discussing the possibilities of this new material. Also some
properties of carbon fibre reinforced CBTr resin have been published but
these will be discussed in the following section.
2.4 Cyclic butylene terephthalate

The renewed interest in thermoplastic liquid moulding can be partially
explained by the commercialisation of thermoplastic oligoesters. These
oligoesters have a low melt viscosity and can be polymerised rather fast
into engineering thermoplastics making them ideal candidates for reactive
processing. This section gives an overview of recent research on cyclic
oligomers.
2.4.1 Cyclic oligomers and their applications

Polymerisation reactions such as polycondensations follow a step growth
reaction mechanism. Polymerisation is hence a result of intermolecular re-
actions between end-groups of monomers, oligomers or polymers. These
intermolecular reactions in polycondensations are accompanied by the re-
lease of small molecules. If the reactions are carried out at high concen-
trations, intermolecular reactions are favoured, nevertheless, for statistical
reasons, intramolecular reactions between the end-groups within the same
molecule occur to a minor extent. These intramolecular reactions result in
cyclic products [58]. Cyclics are therefore found in commercial grades of
polyester and polycarbonate at levels of 0-5% depending on the polymer
[59].
Since the properties of these cyclic oligomers differ from the polymer
properties, their presence has often been the cause of concern as they
might compromise the polymer’s performance by e.g. plasticising it. Their
High dilution High concentration
linear monomer
“end groups” “end groups”
cyclic cyclic cyclic

monomer dimer trimer, etc. tion
erisa polymer
olym
ning p
-ope ion
ring risat
po l yme
-de
cyclic cyclic cyclic c yclo
monomer dimer trimer, etc.
Figure 2.7: Processing routes resulting in cyclic oligomers,

adapted from [58]
possible applications have however been recognised recently and Otaigbe

[60], Hall and Hodge [58, 61] have reviewed the potential of these cyclic
products.
Figure 2.7 depicts some reaction schemes with cyclic oligomers. There
are three options to produce cyclic oligomers, (pseudo-)high-dilution, cyclo-
depolymerisation and polymer-supported synthesis. The applications for
cyclic oligomers range from composite manufacturing through ring-opening
polymerisation, to using their specific recognition properties or employing
them as building block for novel materials [61].
Production of cyclic oligomers
(Pseudo-)High dilution At high dilution conditions (low concentra-

tion of oligomers and polymer), intramolecular reactions are favoured over
intermolecular reactions. The final concentration of oligomers is however
usually lower than 0.1 M [62]. If the cyclic products are formed irre-
versibly, it is only necessary for the reactive species themselves to be
present at high dilution and hence solutions containing a relatively high
concentration of the final cyclic oligomers may be prepared. Such synthe-
ses are said to use pseudo-high dilution conditions. This synthesis route
is obviously more practical to use than the real high dilution [61].
Brunelle et al. [62, 63] have produced cyclic carbonates with this pro-
cess with a yield of 85% and >99% purity. Other cyclics produced by
pseudo-high dilution include cyclic polyetherketones, oligo(butylene tereph-

thalate), polyethersulphones and nylon 11. For more information one is
referred to [58, 61] and references therein.
Cyclo-depolymerisation (CDP) Cyclo-depolymerisation starts from

the polymer in which linkages between the polymeric units are repeatedly
broken and then reformed in an equilibration process. If this equilibration
is carried out at high dilution, its effect is similar to synthesis at high
dilution [58, 61]. Ester bonds are obvious linkages to be involved in such
a cyclo-depolymerisation [61].
Polymer supported synthesis The two routes discussed above tend

to produce some linear products together with the cyclic oligomers. Even
though these two groups can be often separated on the basis of solubility
differences, this is not always the case. Polymer supported synthesis starts
from monomers with two types of end-groups and achieves separations
by having one type of end groups attached to insoluble polymer beads.
Linear oligomers remain bound to the beads, whereas cyclics are released
into the solution and hence easily separated. This approach has been used
to produce nylon 11, nylon 6, ester and ester-amide cyclics [61].
Application of cyclic oligomers

Ring-opening polymerisation (ROP) At high concentrations, the
ring-chain equilibrium is shifted towards the polymer side, favouring re-
actions between oligomers over cyclisations. Hence this ring-opening poly-
merisation is basically the reverse of CDP. In the simplest situations, the
linkages are simply re-shuffled, so no volatiles are evolved. Moreover, since
the larger cyclic oligomers are usually not strained, heat release is rather
small or absent [62]. The absence of a significant reaction exotherm is a
major advantage in the polymerisation of large volumes.
The conversion from cyclics to polymer is hence entropy-driven (ED-
ROP). In the neat mixture of cyclic oligomers, the molecules have only
limited translational entropy1 and since they are cyclic, the available con-
formations are limited. When the rings open, there is only a modest
change in translational entropy, but a significant increase in conforma-
tional entropy [58].
The lack of evolution of heat and volatiles coupled with the relatively
low viscosities of mixtures of cyclic oligomers makes ED-ROP very at-
tractive for composites processing and coating processes. Moreover, since
CDP is the reverse of ED-ROP, a huge potential for recycling is present
[61, 64].
1 translational entropy is due to the number of distinguishable spatial arrangements;
conformational entropy is due to the number of distinguishable shapes of a molecule

keeping centre of mass fixed
The base material for this ring-opening polymerisation should be free

of linear oligomers since their presence would limit the ultimate achievable
molecular weight [63]. If linears are excluded, the molecular weight is
purely determined by the amount of initiator used and hence extremely
high molecular weights can be achieved [63].
Another issue are the very high melting points of the small cyclics
in the cyclic oligomer families, moreover, their solubility in the larger
cyclic oligomers is often very poor. This not only leads to high processing
temperatures at which side reactions are difficult to prevent, it might
also prevent the small cyclics to become incorporated into the final linear
polymer [61].
Recognition properties Cyclic oligomers can be used as recognition

systems. In this role, they provide a cavity into which a moiety can be
recognised. For good recognition, the sizes of the host and guest should
match well and there should also be favourable interactions between them
[58, 61]. With CDP, a macrocyclic library can be composed and this
library can then be screened to identify its most compatible member.
These screening methods have been developed to understand structure-
activity relationships in protein and pharmaceutical chemistry [58].
Building blocks Cyclic oligomers have already been used to build novel
materials and they can be incorporated into supramolecular polymeric
structures [61].
2.4.2 Production of cyclic oligoesters and their poly-

merisation
As was mentioned in Section 2.2.1, Cyclics Corporation was founded in
1999, buying out the General Electric patents concerning the cyclomer
technology. Research in General Electric first focussed on cyclic oligomeric
carbonates. This focus shifted later on to cyclic oligoesters, which are
today the main business of Cyclics Corporation. For more information
about the production and subsequent polymerisation of cyclic carbonates,
one is referred to following references and the references therein [62, 63, 65–
69].
Production of cyclic oligoesters

Cyclic oligoesters are present in commercial polyesters and have been iso-
lated by a variety of extraction techniques, e.g. [70]. The cyclic trimer of
poly(ethylene terephthalate) was first isolated in 1954 and the dimer in
1969. An extensive study reporting the incidence of cyclic polyesters in 13
types of alkylene iso- and terephthalates was detailed by Wick and Zeitler
in 1983 [63]. Until the work of Brunelle et al. [71] however, research
focussed on obtaining pure single-sized cyclics usually by high dilution

chemistry. The yields of these reactions were often very low and purifi-
cation less then simple. Moreover, the discrete cyclic oligomers had often
very high (>300◦ C) melting points whereas the preparation of mixtures
of oligomers can provide significant melting point depression [59, 71].
Pseudo-high dilution The pseudo-high dilution process was first in-

vestigated to make cyclic butylene terephthalate oligomers (CBT) in a
high yield. Using the results of model experiments as guidelines for choice
of solvent and amine, cyclisation reactions were carried out by Brunelle
et al. [71] by concurrent addition of equimolar amounts of isophtaloyl
chloride (IPC) or terephthaloyl chloride (TPC) in CH2 Cl2 and butane-
diol in dry tetrahydrofurane (THF) to a slight stoichiometric excess of
diazabicylco[2.2.2]octane (DABCO) or quinuclidine in CH2 Cl2 over 1 h
with a final product concentration of 0.2 M. This reaction yields cyclic
oligomers as the major products, with a small amount of linear oligomers
still present. Isophthalate cyclics were isolated in a 45% yield, and CBT
in a 30% yield.
Optimisation of this reaction by minimising unwanted side-reactions,
allowed CBT to be formed in a 0.25 M reaction of 1 h with yields as high
as 85%. Similar procedures were adopted to produce a variety of alkylene
phthalate cyclics via direct reaction of diols with diacid chlorides [71].
Ring-chain equilibration Although depolymerisation of polyesters has

been known to produce cyclic oligoesters through ring-chain equilibration,
the first efficient preparation of alkylene phthalate cyclic oligomers fol-
lowing this route was only patented in 1995 (U.S.patent 5.407.984). After
laboratory studies, larger-scale work was carried out in an 8-litre stain-
less steel autoclave, using o-xylene as solvent and various tin or titanate
catalysts, with the tin catalysts being more effective [63]. At about the
same time, the group of prof. Semlyen published a series of papers on the
preparation of cyclics via ring-chain equilibration for several polyesters
[72–75].
Brunelle et al. [63] carried out depolymerisation reactions on com-
mericial PBT (Valoxr 315) by dissolving it in dry ortho-dichloro-benzene
(o-DCB) at reflux, then adding the equilibration catalyst. Table 2.3 illus-
trates the yield of CBT at various conditions, showing that the dilution
does not need to be extreme in order to obtain a reasonable yield.
The reaction mechanism of depolymerisation, which is basically a
transesterification mechanism, is shown in Figure 2.8. Metal alkoxides
have proven to be the most efficient catalyst for such reactions [63]. The
metal alkoxides function by activating a carbonyl via complexation, then
transferring an alkoxide ligand. Thus a linear polymer will interact with
a metal alkoxide catalyst to transfer an alkoxide onto the polymer chain,
Table 2.3: Effect of depolymerisation concentration on the yield

of CBT (left) using 1.0 mole% of cyclic stannoxane [63] (right)
using 1.0 w% of a dibutyltinoxide catalyst [73]
Conc. (M) % CBT Dilution ratio % CBT
0.05 89.7 1/70 70
0.075 61.1 1/50 56
0.10 50 1/30 42
0.15 32.9 1/10 13
0.20 25.9
0.30 15
forming a species such as I, and releasing a linear alkoxide which is termi-

nated with the transferred ligand (II). The original polymer chain length
will be decreased by m units in such a reaction. The metal terminated
polymer I can then react either by chain-chain transfer (polymerisation
reaction) or by a back-biting reaction to form cyclics. The degree of cycli-
sation versus polymerisation can only be controlled by the concentration
of species I in solution. Cyclisation to cyclics with degree of polymerisa-
tion p+1 releases another linear polymer chain terminated with a metal
Ia, which is shorter in length by p+1 units and which can continue to
react to form cyclics. Eventually equilibrium is reached, and the linear
chains (I, Ia, II) react degenerately at the same rate as they form cyclics.
A mixture of cyclic oligomers is being formed during this reaction. An
attempt to correlate the experimental ring-size distribution to theory was
presented by Hubbard et al. [76] whereas Table 2.4 shows the experimen-
tal results from several authors. It is clear that the ring-size distribution
depends on the reaction conditions but also on the total reaction time
[63, 77]. Since the ring-size distribution has a profound effect on the melt-
ing point of the mixture, controlling this mixture is crucial for further
processing.
O
HO CH2 4 O C C O CH2 4 O C COOH
n
O O
M R
OR
R OR
(RO) xM O O
C
O
O
HO C C O CH2 4 O M(OR)x CH2 4
m
O I C
O O
+ O
O
C CH2 4
RO C C O CH2 4 O H O
n-m p
O O
II C
O
O O CH2 4
O O
C C
O
HO C C O CH2 4 O M(OR) x +
m-(p+1)
O Ia
C C O
O O CH2 O
4 p
III
Figure 2.8: Mechanism for metal alkoxide catalysed formation
of cyclics via depolymerisation of poly(butylene terephthalate),
adapted from [63]
30
Table 2.4: Ring-size distribution of cyclic oligomers produced by various methods by investigation of GPC-traces
CDP- CDP- CDP- CDP- CDP- CDP- CDP-
PBT PBT∗ PET∗ PET PET PET PET∗∗ PET∗∗∗ PPT
reference [73] [73] [74] [74] [74] [74] [59] [59] [78]
dilution 1/10 - - 1/10 1/20 1/30 - - -
dimer 25 49 - - - 0.9 - 82 -
trimer 36 25 85 45 20 17 39.6 25.0 1
tetramer 19 14 7 20 18 16 14.9 12.7 6
pentamer 11 9 5 14 17 16 22.1 15.3 4
hexamer 4 1 2 9 13 12 4.4 14.0 <7
heptamer 5 2 >1 6 12 12 12.0 - -
octamer <1 <1 >1 3 9 10 8.6 - -
nonamer - - - 2 6 8 6.6 - -
decamer - - - >1 3 5 1.8 - -
∗ commercial, as received samples, ∗∗ reaction in solution, ∗∗∗ reaction in suspension
Literature review
Initiation
O O CH2 O O OH
4
C C Cl Sn OH
O OH
R
HO C C O CH2 4 O Sn Cl
n+1
O R
O
C C O II
O CH2 4 n
O
Propagation O OH
I + II HO C C O CH2 4 O Sn Cl
2(n+1)
O R
Figure 2.9: Ring-opening polymerisation of oligomeric PBT
cyclics with Sn-containing initiator, adapted from [71]
Ring-opening polymerisation of cyclic oligoesters

Ring-opening polymerisation of cyclic oligoesters can be initiated by many
types of compounds, but certain tin and titianium initiators were most
effective [63]. The mechanism of this reaction is shown in Figure 2.9 for
CBT with the catalyst used throughout this work.
The reaction involves an initiation to form an active chain end, fol-
lowed by propagation reactions continuing until all the cyclic oligomers
are depleted and the ring-chain equilibration becomes degenerate. In this
case, the initiator becomes built into the polymer, which is not termi-
nated unless quenched. As was mentioned above, due to their size, the
cyclic oligomers are nearly strain-free and the polymerisation is almost
thermoneutral, leading to complete equilibration of ester groups. Poly-
dispersities therefore approach 2.0.
The molecular weight of the polymer is controlled by the molar ratio
of cyclic ester to linear functionalities, present in remaining linear oli-
gomers or the in the catalyst [59, 71]. When cyclic catalysts are used as
initiators, the molecular weight does not decrease with increasing catalyst
level since no linear functionalities and end-groups are introduced. Hence
a macrocyclic polymer is formed [71, 79, 80] as is shown Figure 2.10.
There are three conditions that influence the completeness of poly-
merisation [71]:
• purity of oligomers,
• complete mixing of the initiator before the onset of polymerisation,
• high enough reaction rate to complete polymerisation before the

onset of crystallisation.
Bu O O Bu
Bu Sn Sn
O O Bu
O O CH2 O O O O CH2 2
O
4 Bu
C C C
Sn
O Bu
O Bu
Sn
C C O C Bu
O O O CH2 O
O CH2 4 n
O 4 n +1
Figure 2.10: Ring-expansion polymerisation of oligomeric PBT

cyclics with a cyclic stannoxane initiator, resulting in a macrocy-
cle, adapted from [71]
Table 2.5: Effect of catalyst on the molecular weight of PET

polymerised from purified CET at 293 ◦ C for 10 min [81]
catalyst (0.5 mole%) Mn (kg/mole)
Sb2 O3 25.3
Zn(Ch3 COO)2 8.0
CH3 COONa 10.5
CH3 COOK 15.7
GeO2 6.3
MgO 5.9
Youk et al. [80, 81] studied the polymerisation of cyclic (ethylene tereph-
thalate) (CET) with different catalysts, Table 2.5. They also investigated
the effect of various CET’s prepared according to different methods and
purified differently. They obtained results similar to those of Brunelle
et al. [71], namely that the base material and the catalyst used play an
important role in the final properties of the polymer. They also investi-
gated the thermal properties and showed that the degree of crystallinity
strongly depends on the polymerisation temperature.
Besides polymerising homopolymers, cyclic oligoesters are often copoly-
merised [59, 71, 78, 82–85] to obtain better or different properties. Copoly-
mers of CBT and CET were e.g. produced in order to control the crys-
tallinity of the formed polymer, Table 2.6. Nanocomposites produced
with CBT are also a new area of interest [86].
2.4.3 Properties of CBTr resin and (reinforced) pCBT

CBTr resin from Cyclics Corporation is available in two system types, the
two-part system where catalyst and oligomers need to be mixed by the user
Table 2.6: Mixed cyclic alkylene oligomer properties and poly-

merisations (0.3 mole% of tetrakis(2-ethylhexyl)-titanate (TOT),
20 min at 190 ◦ C) [71]
CET/CBT conversion Mw Tm,polym ∆Hm,polym
molar (%) (kg/mole) (◦ C) (J/g)
0/100 96 104.1 228 72.1
1/99 100 90.4 226 68.6
2/98 99 132.7 224 67.3
3/97 98 113.0 224 64.4
5/95 98 121.1 220 59.0
7/93 99 110.0 217 57.4
10/90 94 79.9 214 47.8
Valoxr 315 98 98.0 226 45.8
and the one-part system, which contains a predefined mixture of a certain

catalyst and the oligomers. One advantage of the two-part system is the
larger flexibility concerning the catalyst and its level. Moreover, more
time is available to prepare production as polymerisation is initiated by
mixing the two components and not by raising the temperature. Hence,
the two-part system was chosen for the most part of this work. As was
mentioned above, the catalyst not only determines the reaction speed, but
also the final achievable molecular weight. Table 2.7 gives an overview of
the catalysts commonly used with CBTr resin.
Depending on the production process, a fast or slower catalyst is the
best choice. Since the aim of this work is to study liquid moulding ap-
plications with CBTr resin, a catalyst with moderate activity (Fascatr
4101) is chosen to expand the available time window for impregnation. A
one-part system named CBT-XB3 does exist containing 0.3 mole% of this
catalyst.
For reactive processing of thermoplastics, the processing window is
very important as was explained in Section 2.2.2. Moreover, for closed
mould processes, it is very advantageous to avoid thermal cycling of the
mould. CBTr resin does not have a defined melting point since it consists
of a mixture of oligomers, but a melting range which lies between 120 and
160◦ C [83]. Since PBT only melts above 220◦ C and the ring-opening poly-
merisation is thermoneutral, a temperature window exists for isothermal
processing, which offers a distinct advantage over polyamide 12, which
cannot be processed isothermally and over polyamide 6 which has a large
exotherm.
Figure 2.11 shows the viscosity evolution of the one-part system, CBT-
XB3. On the lefthand-side, the difference in viscosity evolution between
processing above and below the polymer melting point is illustrated. At
225◦ C, the starting viscosity is very low but it rises fast to a plateau near
Table 2.7: Different catalysts to polymerise CBTr resin [87],

tradenames are from Atofina Chemicals Inc.
structure name tradename activity
R OR dialkyl Fascatr 4214

Sn dialkylstannate R = butyl high
R OR
Bu O O Bu 1,1,6,6-tetrabutyl-
Sn Sn 1,6-distanna- stannoxane ∗ high
Bu O Bu 2,5,7,10-tetreaoxy-
O
cyclodecane
Cl OH
Sn
butyltinchloride
dihydroxide Fascatr 4101 moderate
Bu OH O
C7H15
O O O butyltin
Bu Sn O C C7H15 tris(2-ethylhexoate) Fascatr 4102 slow
O O C7H15
O
∗
no tradename available, group name given
500 Pa·s, the part is hence too soft for demoulding and requires a cooling
cycle to solidify [44, 88]. At 190◦ C however, the initial viscosity is slightly
higher, and the reaction is somewhat slower. More importantly, the final
viscosity rises to a level which coincides with crystallised PBT, hence the
part can be demoulded at this temperature [44, 88].
On the righthand-side of Figure 2.11, the onset of the viscosity rise is
depicted for different possible isothermal processing temperatures. It is
clear that at higher temperatures, the reaction, and hence the viscosity
rise, proceeds more rapidly, decreasing the time window for impregna-
tion (time to reach 1 Pa·s). Increasing this time window has however
consequences on the overall polymerisation time, which is illustrated in
Figure 2.12.
Since this material is relatively new, studies on the properties of the
polymerised resin and its composites are rather scarce and some of them
were not available at the start of this study. However, a short overview
is given below, Table 2.8 and 2.9. In order to distinguish polymerised
CBTr resin from commercially available PBT, the first will be referred
to as pCBT. Moreover, when a cyclic catalyst is used, the pCBT is a
macrocycle, which will be referred to as c-pCBT.
Figure 2.11: Viscosity-evolution versus time for the one-part

CBTr resin, CBT-XB3 (left) adapted from [44], (right) adapted
from [89]
Figure 2.12: (left) Degree of conversion as function of time for

different temperatures for the one-part CBTr resin, CBT-XB3,
adapted from [44], (right) Impregnation and polymerisation time
for the one-part CBTr resin, CBT-XB3, adapted from [44]
Table 2.8: Properties of unreinforced PBT, pCBT and c-pCBT

PBT pCBT pCBT c-pCBT
Reference [90] [91] [82] [92]
Tensile modulus (GPa) 1.96-3.04 1.9-3.0 - 1.93-3.0
Tensile strength (MPa) 56.5 58 53.8 56.5
Elongation at break (%) 50-300 50-300 >50 50-300
Flexural modulus (GPa) 2.31-2.79 - 2.38 2.28-2.76
Flexural strength (MPa) 84-117 - - -
Density (g/cm3 ) 1.3-1.38 1.31-1.38 1.31 1.31-1.38
Melting point (◦ C) 235 - - 220-265
Table 2.9: Mechanical properties of reinforced pCBT

Reference [91] [91] [91] [44, 94]
Fibre glass carbon carbon carbon
Orientation random (0,90,45,45)s UD
Testing direction - 0 90 0
Fibre volume fraction (%) 57 47 47 55
Tensile modulus (GPa) 13.2 51.2 47.9 115
Tensile strength (MPa) 191 635 717 1565
Elongation at break (%) 1.7 - - -
Flexural modulus (GPa) 10.9 - - -
Flexural strength (MPa) 192 - - 1310
The properties listed for pCBT and c-pCBT compare well to the lit-
erature values for PBT. The processing method (temperature profile) is
however not given for most of the entries, only Bahr et al. [82] mention
that the tensile bars were injection moulded during polymerisation. The
isothermal process can however have an effect on the properties of pCBT,
which was not yet investigated.
Moreover, Al-Zubi et al. [92] showed properties of c-pCBT in their
overview table which they did not reproduce since an elongation at break
of more than 50% was listed, whereas their tests revealed a very brittle
tensile behaviour (elongation at break < 4%). This brittle behaviour was
attributed to the rotomoulding process. They claim that the zero-shear
process of rotomoulding induces less entanglements between the polymer
chains.
The same brittle behaviour for c-pCBT was noticed by Miller et al. [79,
93]. They attributed this behaviour to the macrocyclic nature of c-
pCBT, which reduces the effective entanglements and the amount of tie-
molecules. Heat treatment of the c-pCBT efficiently breaks down the
macrocycles and increases the polymer toughness. These authors how-
ever did not investigate linear catalysts which would immediately yield
linear pCBT.
Table 2.9 shows some properties of reinforced pCBT. These composite
data only include fibre-dominated properties which are hardly affected
by the matrix properties. Moreover, no data were available on the ma-
trix properties within this composite (conversion, molecular weight, crys-
tallinity). A systematic study of composite properties with special atten-
tion for the matrix properties is hence needed.
2.4.4 Production methods for reinforced pCBT

Several options for the production of reinforced pCBT are stated in litera-
ture [44, 88, 94–96], these include compression moulding of several prepreg-
types, pultrusion, liquid moulding (RTM) and resin film infusion. Most
of these papers however only highlight the possible production methods
without specifically demonstrating their feasibility or characterising the
properties of the composite hence produced.
The production of two types of prepregs was described by Winckler et
al. [96]. They describe pastille type prepregs which consists of a regular
array of resin droplets on top of the reinforcement. More intimate min-
gling can however be achieved by a direct powder deposition process. The
prepregs produced by this powder deposition process were subsequently
used by Coll et al. [97, 98] in what they call a resin film infusion process.
They have focussed on the heat transfer problem and consolidation anal-
ysis for making large sandwich panels with a reinforced pCBT skin and a
foam core.
Bank et al. [91] have recently described the thermal spray process
for short fibre composites where a mixture of CBTr resin powder and
chopped fibres are sprayed onto a mould to manufacture prepregs. This
method can also be applied for continuous fibres when the powder is
sprayed in between subsequent plies. The hence prepared prepregs are
then compression moulded. This process is very similar to the direct
powder deposition process, differing mainly in the shape of the prepregs
(flat versus net-shape).
Thermoplastic RTM is an interesting option for the production of re-
inforced pCBT. The resin fulfills all the characteristics required for this
process. Although its potential is often recognised in literature, data on
the composite properties do not exist. During this study however, Repsch
et al. [99] and Weyrauch et al. [100] published papers on the process sim-
ulation and the automatic control of RTM processes with CBTr resin.
They nevertheless focussed on the flow characteristics without considering
the effect of the investigated process parameters on the material proper-
ties.
Rösch et al. [95] on the other hand described an injection system to
allow the use of the one-part system in RTM. This system consists of a
resin delivery cylinder driven by a separate pneumatic cylinder. Both the
cylinder and the end plug have a heater which can be controlled separately.
The resin is preheated in the cylinder but only molten in the end-plug,
prior to entering the mould. Once again, only the process was outlined
but no information was given on the parts produced with this process.
a b c
Figure 2.13: Schematic overview of organisational levels in

semi-crystalline polymers: (a) spherulitic superstructure, (b)
lamellar stacking and (c) unit cell (α-PBT, [103])
2.5 Morphology of poly(butylene terephtha-

late)
The properties of thermoplastics are strongly influenced by their mor-
phology, which can be amorphous or semi-crystalline depending on the
processing parameters. Additionally, some of the composite properties
are dominated by the matrix behaviour (e.g. transverse and shear prop-
erties, fracture toughness [101]) and hence by its morphology. Therefore,
this section will shortly describe the morphology of semi-crystalline ther-
moplastics and poly(butylene terephthalate) in particular.
2.5.1 Semi-crystalline polymers

Morphology
Thermoplastics can be either fully amorphous or semi-crystalline. Within
these semi-crystalline polymers, three structural levels of crystalline or-
ganisation exist, Figure 2.13. The first level consists of the spherulitic
superstructure (1-100 µm), Figure 2.13a. These spherulites grow dur-
ing crystallisation from different nucleation sites and can finally impinge
each other, resulting in a space-filling spherulitic superstructure. The
spherulites consist out of crystalline ‘arms’, with the space in between
these arms being filled by amorphous material [102].
These crystalline arms of the spherulite contain crystalline lamellae.
These lamellae, alternated by amorphous material form lamellar stacks
and are the second organisational level, Figure 2.13b. When polymers
crystallise from the melt, were they have a randomly coiled conformation,
folded chain lamellae are formed because of kinetic hindrance [104]. This
results in lamellar thicknesses in the order of 10-100 nm. Since a single
polymeric chain is often about 1000 nm long, some chains can crystallise
into two or more lamellae or even into more than one spherulite. These
2.5 Morphology of poly(butylene terephthalate) 39
extended chain polymer precursors

conformation
crystallisation during
polymerisation
kin
hin
te
sta
eti
d
n
elt
ere
ng
ca
tio
lid
m
lly
ali
isa
dp
so
ne
er
roc
m
an
ly
ess
po
formation of
metastable crystals
folded chain randomly coiled

conformation conformation
Figure 2.14: Schematic representation of chain conformations in

polymers, adapted from [104]
molecules are called tie-molecules and are crucial for the mechanical prop-
erties of a semi-crystalline polymer [102, 105, 106].
Inside the lamellar structure, the third organisational level exists,
namely the unit cell of the crystallite, Figure 2.13c. This unit cell is
similar to the unit cells found in metals, with this difference that only a
small part of the molecular chain lies in one unit cell. Moreover, the bond
types in the different crystal directions are very different leading to highly
anisotropic mechanical properties at the unit cell level. Some polymers
can crystallise in different polymorphs, each characterised by their own
unit cell.
Crystallisation during polymerisation

According to Wunderlich [104], crystallisation during polymerisation can
lead to a very different crystal morphology ranging from a fully extended
chain conformation to a folded chain conformation, which is the typical
conformation for crystallisation from the melt, Figure 2.14.
Crystallisation and polymerisation can be divided into three cate-
gories. First, polymerisation and crystallisation can be truly simultaneous.
In this case, a mobile monomer is added to the growing crystal and both
covalent and secondary bonds are set immediately. Second, it is feasi-
ble that polymerisation is followed in succession by crystallisation before
the molecule is completely polymerised. The third category is separate
polymerisation and crystallisation. In this case, the polymer molecule is
completely in the dissolved or molten state before crystallisation com-
mences. Although this is the most common process, complications can
still occur if the polymer molecules are produced in a special conformation

and crystallisation begins before randomisation.
Ring-opening polymerisations, as is the case when polymerising CBTr
resin, are an example of a chain reaction polymerisation. These reactions
have shown that polymerisation at temperatures below the melting tem-
perature of the final polymer lead to a thermodynamically more stable
state than crystallisation of the identical polymer from the polymer melt
[104]. A possible mechanism of successive polymerisation and crystalli-
sation might involve the production of polymer chains which crystallise
at a certain distance from the active site. The morphology of the re-
sulting crystal is then critically dependent upon the density of the active
sites. For a low active site concentration intramolecular crystallisation
may be the only solution, while higher concentrations might accomplish
intermolecular crystallisation [104].
The tie-molecule density can also be influenced by crystallisation dur-
ing polymerisation. Miller [79] stated that if oligomers are added to poly-
mer chains that already began to crystallise, only short amorphous seg-
ments are attached to crystal growth fronts, which in turn have a low
probability of becoming part of more than one lamellae, thus forming less
tie-molecules. Hence this type of crystallisation might have an effect on
the mechanical properties.
Effect of morphology and degree of crystallinity on mechanical

properties
The mechanical properties are of course influenced by the degree of crys-

tallinity and the morphology of the polymer. It is however difficult to sep-
arate these effects since both of them are often related [106]. Crystalline
segments act as rigid areas in comparison to the amorphous regions. In
a sense they thus act as additives, which will increase the elastic modu-
lus and the ultimate strength of the polymer. However, if the crystalline
regions do not move easily with the amorphous region on elongations,
these regions act as stress concentrators, thereby weakening the polymer
[105]. Figure 2.15 depicts the influence of the degree of crystallinity on
the mechanical properties of polyamide 6.
Moreover, the ultimate strength is determined primarily by the tie-
molecules which have to transfer stresses between adjacent lamellae and
or spherulites [108]. Usually, a high degree of crystallinity, voids, large
spherulites or/and a low amount of tie-molecules result in a brittle be-
haviour.
Figure 2.15: Influence of degree of crystallinity on elastic mod-

ulus, yield and fracture stresses and toughness of polyamide 6,
adapted from [107]
Figure 2.16: X-ray diffraction pattern of polymorphs of

poly(butylene terephthalate), constructed with data from [103]
2.5.2 Crystalline structure of poly(butylene tereph-

thalate)
Unit cell level
Two polymorphs have been reported for poly(butylene terephthalate).
The α-phase occurs under normal, quiescent crystallisation from the melt
and has a kinked chain conformation [103]. This polymorph, however,
transforms reversibly into the β-phase when the oriented chains are stressed
e.g. in PBT filaments [103, 109, 110]. Both polymorphs have a triclinic unit
cell, with different X-ray diffraction patterns, Figure 2.16. Hall et al. [109]
reviewed the unit cell parameters found in literature for both polymorphs
and made ‘best estimates’ from those showing only small differences in
their dimensions, Table 2.10.
Spherulitic level
Besides the existence of two polymorphs, Stein and Misra [111] observed
and described two types of spherulitic superstructures depending on the
crystallisation temperature. In light scattering measurements with crossed
polarisers, both exhibit the typical four-leaf-clover type pattern. However,
the lobes in the scattering patterns for crystallisation temperatures be-
tween 0 and 180◦ C are located along the polar directions (0◦ -90◦ pattern),
whereas the lobes in the 200◦ C samples are located at 45◦ to the polar di-
rection, (45◦ pattern). The 0◦ -90◦ pattern is characteristic of spherulites
whose optical axis lies at an angle of approximately 45◦ to the spherulitic
Table 2.10: Best estimate for unit cell properties of PBT accord-
ing to Hall et al. [109]
α-PBT β-PBT
a (Å) 4.86 ± 0.03 4.72 ± 0.02
b (Å) 5.96 ± 0.01 5.79 ± 0.03
c (Å) 11.65 ± 0.06 13.00 ± 0.10
α (degrees) 99.7 ± 0.6 102.7 ± 1.0
β (degrees) 116.0 ± 0.7 120.2 ± 0.7
γ (degrees) 110.8 ± 0.5 103.7 ± 2.1
Volume (Å3 ) 261.5 ± 4 272.7 ± 7.4
Density (g/cm−3 ) 1.397 1.338
radius and are referred to as unusual spherulites. On the other hand, the
45◦ pattern is similar to those observed in most synthetic polymers and
arises from spherulites with their optical axis either along or perpendic-
ular to the spherulitic radius, these spherulites are referred to as usual
spherulites.
A mixed type of spherulites, showing no azimuthal orientation in the
corresponding light scattering pattern can also be observed in some sam-
ples and the change in preferential radial orientation of usual and unusual
spherulites was confirmed by Roche et al. [112]. The switch-over between
unusual, mixed and usual spherulites depending on the crystallisation
temperature however is not consistent in literature [111, 113, 114]. Fig-
ure 2.17 and 2.18 show optical micrographs and light scattering patterns
of the different types of spherulites.
X-ray diffraction studies showed that the unit cell structures obtained
for the crystals associated with both types of spherulites are identical and
contain the α-polymorph [111].
Möginger et al. [116] observed an increase in amount of usual type
spherulites when the cooling rate decreased and the melt temperature
increased for injection moulded samples. Together with the increasing
amount of usual type spherulites, PBT became more brittle, reducing
the strain at failure from more than 30 to less than 4%. The increase
in overall crystallinity was however not discussed and since the overall
degree of crystallinity is known to be higher in samples containing usual
spherulites [115, 117], the brittle behaviour cannot be solely attributed
to the type of spherulite. Ludwig and Eyerer [115] on the other hand
observed an increase in yield stress for the usual type spherulites, but
could not find a decrease in strain at the yield point. They also do not
report on a difference in strain at failure.
Figure 2.17: Optical micrographs depicting unusual, mixed and

usuals spherulites, adapted from [115]
A P
Figure 2.18: Hv light scattering patterns for (left) unusual and

(right) usual spherulites
Multiple melting behaviour
The multiple melting behaviour was first observed for PBT by Hobbs and
Pratt [118]. This behaviour may arise from: (1) the presence of alter-
nate crystal modifications, (2) molecular weight segregation accompany-
ing crystallisation, (3) variations in morphology, (4) orientation effects, or
(5) melting, recrystallisation and annealing processes during the thermal
scan.
Stein et al. [111] have attributed the multiple melting behaviour of
PBT to the existence of the two types of spherulitic superstructures. The
usual spherulites were claimed to have a better thermal stability and hence
a higher melting point compared with the unusual spherulites. Ludwig et
al. [115] on the other hand made the opposite assignment, whereas Yeh
et al. [119] have found contributions of usual spherulites in both melting
endotherms.
Since then, many authors have investigated this phenomenon, some
of them combining thermal analysis with either, scanning electron mi-
croscopy [120], model calculations [121, 122], X-ray diffraction [123] or
polarised light microscopy [124]. The appearance of two or more melting
peaks is nowadays most often ascribed to the melting and recrystallisation
process during the heating scan and to the melting of crystals formed dur-
ing secondary crystallisation upon cooling of an isothermally crystallised
sample.
2.5.3 Influence of fibrous reinforcement

It is well known that fillers with active surfaces, such as some types of
fibres, present a large number of nucleation sites. They are however not
evenly distributed throughout the melt but lined up along the fibre surface
leading to the formation of closely spaced nuclei at the fibre surface [125].
In carbon fibres, this can lead to a truly transcrystalline zone where the
polymer crystal can only grow in one direction and a one-dimensional
oriented crystal structure develops.
Since glass fibres are treated differently than carbon fibres, a true
transcrystalline zone might not exist, nevertheless, the nucleation density
is increased by the fibre surface [113, 126] leading to a higher density of
smaller spherulites in glass fibre reinforced PBT, Figure 2.19.
The higher nucleation density along the fibres has of course a pro-
nounced effect on the crystallisation kinetics as is shown in Figure 2.20.
Park et al. [128] have studied both the isothermal and non-isothermal
crystallisation kinetics of neat and glass fibre reinforced PBT and con-
cluded that the overall crystallisation rate was increased with increasing
glass fibre content as a result from the increased nucleation density.
a c
b d
Figure 2.19: Optical micrograph of unreinforced PBT (a and b)

and glass fibre reinforced PBT (c and d) after ice water quenching
(a and c) and natural-air cooling (b and d), adapted from [113]
GF-PBT
PBT
Figure 2.20: Crystallisation kinetics at 203◦ C for unreinforced

and glass fibre reinforced PBT (50 vol%), adapted from [127]
2.6 Concluding remarks 47
2.6 Concluding remarks

In-situ polymerisation of thermoplastics provides an interesting process-
ing route for the production of continuous fibre reinforced thermoplastics
without the need for intermediate products such as preforms. Liquid
moulding technologies such as RTM, on the other hand, are well-known
production routes for the production of thermoset composites. Using
thermoplastics in traditionally thermoset production techniques might fa-
cilitate the change-over from one material class to the other.
Recent developments in the thermoplastic industry, including the pro-
duction of a liquid activator system for polyamides and the commercialisa-
tion of cyclic oligoesters, have renewed the interest in reactive processing
of thermoplastic composites. Most research efforts have up till now fo-
cussed on the polyamides whereas cyclic oligoesters seem to have some
advantages over polyamides.
This literature review has shown the state of the art concerning ther-
moplastic RTM with cyclic oligoesters. Starting from this state of the art,
the next chapter will outline the contents of this work.
Chapter 3
Problem statement
Ring-opening polymerisation of thermoplastic oligoesters e.g. CBTr resin,

to form engineering thermoplastics are clearly suitable candidates for ther-
moplastic liquid moulding processes. Although this possibility was often
cited in literature, very few reports exist concerning on one hand, the
manufacturing of composites with CBTr resin and on the other hand,
the properties of the composites produced in this way.
The first step within this study will hence consist of setting-up a lab-
scale production unit for thermoplastic resin transfer moulding (TP-RTM)
in order to manufacture continuously reinforced poly(butylene terephtha-
late) starting from cyclic butylene terephthalate. As reinforcement, glass
fibre fabrics are chosen since no mechanical property data are yet available
for continuous glass fibre composites with CBTr resin. Furthermore, the
option of using glass fibres with CBTr resin is an advantage over poly-
amides since commercial fibre sizings very often inhibit polymerisation of
the latter.
In a second phase, the properties of glass fibre reinforced pCBT are
investigated for different glass fibre fabrics. Compared to traditional ther-
moplastic processing, the matrix is polymerised in-situ and then solidified.
Since isothermal processing is very advantageous for RTM, crystallisation
will take place at the polymerisation temperature. Given that the pro-
cessing route in thermoplastic RTM differs substantially from conventional
thermoplastic processing, special attention is paid to the matrix properties
within the composites. Degree of conversion and crystallinity are hence
determined for all composite samples.
The mechanical properties of the matrix are of course related to its
morphology and this morphology is influenced by the processing route.
The isothermal TP-RTM process can not only give rise to crystallisation
during polymerisation, but crystallisation at the relatively high processing
temperature can also lead to a morphology differing from rapidly cooled
49
50 Problem statement
PBT.
In the third phase of this study, the properties of the unreinforced
pCBT will therefore be investigated. The effect of the isothermal pro-
cess will be investigated by comparing these samples to samples produced
non-isothermally. Moreover, these properties will be compared to com-
mercially available PBT, produced by a normal polycondensation reac-
tion. Within this part, the mechanical properties will be related to the
physical and morphological properties of the matrix.
The effect of the processing route on the matrix properties will be
shown to be significant. It was hence decided to dedicate the last part
of this work to a more fundamental study on the crystalline structure
during its formation. In order to study this process in-situ, everyday
equipment is not sufficient anymore. Therefore, time-resolved X-ray mea-
surements were performed using synchrotron radiation. This allowed to
study the crystalline morphology at the unit cell and lamellar level dur-
ing the isothermal processing route for both reinforced and unreinforced
CBTr resin.
This work hence consists out of four parts:
• Manufacturing of continuously reinforced CBTr resin with
TP-RTM (Chapter 4 and 5 and Appendix A).
• Investigation of composite properties with emphasis on the

properties of the in-situ polymerised matrix (Chapter 5).
• Effect of the processing route (isothermal processing) on the

properties and morphology of pCBT (Chapter 6).
• Detailed investigation of the crystalline structure during its

formation by using time-resolved X-ray measurements (Chapter 7).
Chapter 4
Materials and Methods
4.1 Materials
4.1.1 Polymers
Cyclic butylene terephthalate The prepolymers used for the pro-
duction of thermoplastic composites with resin transfer moulding are the
cyclic butylene terephthalate oligomers (CBTr resin, CBT-XB0) sup-
plied by Cyclics Corporation. As mentioned before, these oligomers are
precursors for the thermoplastic poly(butylene terephthalate) (PBT). The
number of butylgroups in the oligomer mixture varies from two to seven,
resulting in a melting range from 120-160◦ C, Figure 4.1.
Before processing, the oligomers were dried overnight at 110◦ C to re-
move residual moisture, which could interfere with the polymerisation
reaction and the formed polymer by hydrolysis, Equation 4.1. A small
amount (0.45 w%) of the tin-based transesterification catalyst (Fascatr
4101, Table 2.7), which is commercially available from Atofina Chemicals
Incorporated, is added to the molten oligomers to facilitate polymerisa-
tion.
H O C C O CH2 4 n
O C C O CH2 4 m
OH + H2O
O O O O
H O C C O CH2 4 n
OH + H O C C O CH2 OH
4 m
O O O O
(4.1)
Although not used for thermoplastic RTM in this research, a one-
part system of CBTr resin with the Fascatr 4101 catalyst, named CBT-
XB3 exists. The amount of catalyst is however somewhat lower namely
0.33 w%. This material will be used for the synchrotron experiments in
Chapter 7.
51
52 Materials and Methods
Figure 4.1: DSC trace of catalysed CBT-XB0 heated at

5◦ C/min from 100 to 250◦ C, indicating the oligomer melting
range, cold crystallisation of the formed polymer and the final
polymer melting
Epoxy A very common epoxy resin for use in RTM was chosen to pro-
duce thermoset composites. Epikoter 828 is a liquid epoxy resin produced
from bisphenol A and epichlorohydrin. Its epoxy group content lies be-
tween 5260 and 5380 mmole/kg. In order to crosslink this resin, an amine
based curing agent, Epikurer DX 6514, is used. Resin and hardener are
mixed with a mass ratio of 100:17.
PBT Commercially available PBT, tradename Ultradurr B4500 and

Crastinr , was obtained from BASF-AG and DuPont Engineering Poly-
mers respectively. They supplied small injection moulded plates which
will be from hereon be referred to as PBT to avoid confusion with pCBT.
4.1.2 Reinforcements
Five different glass fibre fabrics were used to produce and characterise
glass fibre reinforced pCBT. Unidirectional fabrics were chosen since uni-
directional ply properties are necessary design parameters for structural
parts. Moreover, their transverse properties are not fibre-dominated, giv-
ing an indication of the effect of the production proces on the matrix
properties. Next, both a weave and a non-crimp fabric were chosen to
compare pCBT composites with traditionally processed PBT composites.
An overview of the fabrics is given in Figure 4.2.
In order to reach a fibre volume fraction around 50%, the number of
4.1 Materials 53
layers needed for each fabric are calculated for a 2 mm thick composite.
The fabrics were dried overnight at 110◦ C to remove moisture.
S-UD This non-crimp fabric from Saertex Wagener GmbH (S14EB970-

00940-T1300-499000) has a total areal density of 951 g/m2 . Although
this fabric is mostly unidirectional (912 g/m2 , 2400 tex, 17 µm), a small
amount of 90◦ fibres (27 g/m2 , 68 tex, 17 µm) is added to ensure for the
fabric stability. In order to reach a fibre volume fraction of 54%, three
layers were used. The fibre lay-up for the laminate is (0,90,90,0,0,90).
This lay-up is not completely symmetric, however, a non-symmetric lay-
up was preferred over changing the fibre volume fraction to either 37% (2
layers) or 75% (four layers).
A-UD Ahlström Fibre Composites supplied this non-crimp fabric (42024-

L-50-50), which consists out of three layers, 0◦ (160 tex), 90◦ (2400 tex)
and a random mat. The areal density of these individual layers was de-
termined experimentally resulting in respectively 1217, 60 and 53 g/m2 .
The fibre type of the three layers is not identical. Two layers of these
fabric were used (0, R, 90)s to reach a fibre volume fraction of 52%.
ECA-UD Eurocarbon B.V. developed a braid for a leaf spring proto-

type (Appendix A). This tubular braid consists for 84% out of 0◦ fibres
(Vetrotex, RO99 P192, 4800 tex, 24 µm), whereas the ±45◦ fibres (PPG,
sizing 1383, 300 tex, 13 µm) only encompass 16% of the total areal weight,
which is 1200 g/m2 . Two layers of this fabric were used, resulting in a
total fibre volume fraction of 48%. The unidirectional fibres are not of
the same type as the ±45◦ fibres. In order to use this tubular braid into
the plate mould, the braid was cut open.
S-B This fabric from Saertex Wagener GmbH (S31EX020-00920-01270-

250000) is a biaxial (±45◦ ) (600 tex, 12 µm), non-crimp fabric. The
measured areal density of the fabric is 590 g/m2 . Four layers were used
to reach a fibre volume fraction of 47%.
W-R580 This fabric from Saint-Gobain Syncoglass N.V. is a plain glass

fibre weave with an areal density of 580 g/m2 and filament diameter of
21 µm. A fibre volume fraction of 46% was achieved by taking four layers
of this fabric.
Twintex For comparative reasons, PBT commingled yarns were used,

namely PBT Twintex1 . This fabric from Vetrotex Renforcement S.A.,
which is a 2 × 2 twill weave, was used as a reference material. The areal
1 This material is no longer being produced
a b
c d
e f
g h
Figure 4.2: Fabrics used for the production of composites (a)
S-UD front, (b) S-UD back, (c) A-UD front, (d) A-UD back, (e)
ECA-UD, (f) S-B, (g) W-R580, (h) Twintex
4.2 Production set-up for thermoplastic RTM 55
density is 1030 g/m2 with a fibre weight fraction of 65%. Four layers of
this fabric were used to reach a fibre volume fraction of 50%. This fabric
was compression moulded to obtain composite plates of around 2 mm
thickness.
4.2 Production set-up for thermoplastic RTM

At the start of this research, there was no set-up available for the pro-
duction of thermoplastic composites with RTM. Due to a too low max-
imum mould temperature and an inadequate resin reservoir, the set-up
for thermoset RTM is not suitable for thermoplastic RTM. Therefore, it
was decided to create a new experimental set-up to produce thermoplastic
composites with the two-part CBTr resin.
4.2.1 Mould set-up

The mould was supplied by Beiner Kunststofftechnik GmbH and was
mounted on a small rig provided by Cyclics Corporation. By temporarily
diverting the tubing from the heating unit of a prepregger, available in
the composites’ lab of the department, it was possible to heat the mould
up to temperatures of 250◦ C. The mould cavity allows for the production
of A4-sized plates (320 × 200 × 2 mm3 ).
Disposable Teflon tubes are used to connect the mould to the injection
set-up at one side and to the overflow at the outlet side. The overflow is
needed to protect the vacuum pump from the resin.
Before each production run, the mould is treated with mould release
to facilitate demoulding of the plates. One layer of Zyvax Sealer and two
layers of Zyvax Watershield are used for this purpose.
4.2.2 Injection set-up

The injection set-up has to fulfill the following requirements. First, it
has to be able to heat the CBTr resin to temperatures up to 240◦ C
with the possibility to keep the melt either under vacuum or an inert N2 -
atmosphere. Second, a small amount of catalyst has to be added prior to
the injection. The catalyst and CBTr resin should be mixed in order to
induce polymerisation. Vacuum then has to be applied to infuse the resin
into the mould. Although overpressure would be useful to simplify mould
filling, this option was not yet considered for the lab-scale set-up since it
would complicate the injection set-up.
Preliminary injection set-up

The first lab-scale set-up, Figure 4.3, was the set-up advised by Cyclics
Corporation. The CBTr resin is put in a disposable Erlenmeyer, which is
submerged into an oil bath to reach the desired temperature. The input
for the temperature control is not the temperature of CBTr resin but
the oil temperature, therefore, an external temperature sensor is used to
ensure that the CBTr resin is at the correct temperature. A magnetic
stirrer mixes the catalyst with the molten resin.
T control nitrogen
oil bath supply
T sensor
CBT
disposable
Erlenmeyer
oil bath
Figure 4.3: Oil bath injection set-up (left) overview of injection

set-up, (right) detail of heated erlenmeyer
There are a number of disadvantages to this set-up. First of all, it is

not safe to use an open oil bath. The set-up is quite unstable, especially
during injection, therefore, hot oil can be spilled, causing serious burns.
Secondly, the Erlenmeyer can only be used once as the formed polymer
sticks to the glass causing it to break upon shrinking. Even if a mould
release is used, the shape of the Erlenmeyer seriously hampers cleaning.
Another disadvantage is the temperature control. Although it is quite
easy to correlate the oil temperature to the CBTr resin temperature with
some basic measurements, a direct temperature measurement of the oli-
gomers connected to a temperature control unit is preferred. The mixing
of catalyst and CBTr resin with a magnetic stirrer might be adequate for
small amounts, a mechanical stirrer would however improve the mixing
quality, therefore allowing for shorter stirring times and hence an increase
of the time window for impregnation. All the major disadvantages were
drastically improved in the second lab-scale set-up.
Final injection set-up

The improved lab-scale set-up is shown in Figure 4.4. The CBTr resin is
molten in a cylindrical reaction vessel, which is first treated with mould
release. The mould release avoids glass breakage upon polymer shrinkage
4.2 Production set-up for thermoplastic RTM 57
whereas the shape of the reaction vessel allows for cleaning and thus reuse
of the vessels. When the CBTr resin is loaded into the vessel, it is closed
with a lid, which contains 5 inlets. There are two kinds of caps for the
inlets, a completely closed cap and an open cap with silicon sealing to
allow for continuous use of e.g. a temperature sensor. If these inlets are
properly sealed, vacuum can be applied to the vessel.
injection mechanical
tube stirrer
inlet
N2 supply
T control T sensor
unit (Pt100)
heating catalyst
mantle inlet
Figure 4.4: Final injection set-up (left) temperature control unit

and heating mantle, (right) reaction vessel inlets
In order to heat the CBTr resin, the reaction vessel is put into a heat-
ing mantle that is connected to a temperature control unit. A temperature
sensor (PT 100), also connected to the control unit, is constantly measur-
ing the CBTr resin temperature. From the measured temperature, the
energy supply to the heating mantle is controlled in order to reach the
set temperature. The four open caps with sealing are used for the tem-
perature sensor, the injection tube, the mechanical stirrer and (if needed)
the nitrogen supply. The fifth closed cap can be opened in order to add
the catalyst to the melt. Vacuum can be applied through the injection
tube, which is not yet introduced into the melt, and the mould prior to
injection. During the adding of the catalyst however, this vacuum will
be broken. Infusion can then be initiated by simply pushing the injection
tube into the melt.
4.2.3 Processing parameters

The processing sequence used to manufacture the thermoplastic compos-
ites with the set-up described above is depicted in Figure 4.5. Since
isothermal processing is a distinct advantage for closed mould processes
and is possible for CBTr resin, the temperature of the molten oligomers
(Tmelt ) and the mould (Tmould ) are both 190◦ C unless stated otherwise.
This means that both polymerisation and crystallisation take place at
the same temperature. In principle, the part can also be demoulded at
Vn
degree of conversion
Vi
V1
time
stirring mould filling crystallisation
tstir
tfill
vacuum
catalyst
tpol
Figure 4.5: Processing sequence for the production of thermo-
plastic composites with RTM
this temperature. However, since demoulding at such high temperatures

is rather troublesome with the current mould set-up, the part is cooled
down before demoulding.
Vacuum cannot be kept constant during the resin infusion due to the
continuously increasing viscosity. If the vacuum is too high, the low initial
viscosity will lead to too fast mould filling, leaving behind porosities in the
fibre bundles. On the other hand, if the vacuum is too low, the high final
viscosity will prevent complete mould filling. Therefore, vacuum (Vi )
is increased incrementally during the infusion between 0.4 and 0.8 bar.
A 0.2 bar increment is most often used even though this might change
slightly from one fabric to another. Once the mould is completely filled
(tfill ), in- and outlet ports are closed after which sufficient time (tpol )
should be available to complete the polymerisation reaction and to allow
for crystallisation.
4.3 Production methods differing from ther-

moplastic RTM
Production of thermoset composites For the production of ther-
moset composites, the existing thermoset resin transfer moulding set-up
was used. Plates with dimensions of 500 × 300 × 2 mm3 can be produced
by using the correct picture frame. Before placement of the reinforcement,
the mould surface is treated with a mould release, PMR90, to facilitate
demoulding.
The resin and hardener are premixed with a mechanical stirrer for
5 minutes after which they are infused into the resin reservoir, which is at
room temperature. In this reservoir, the mixture is degassed for several
minutes before injection.
The mould temperature during injection is 40◦ C and the resin is in-
jected with an overpressure of 2 bar and assisted by a vacuum of 0.6 bar.
When the mould is completely filled, the mould temperature is raised to
4.3 Production methods differing from thermoplastic RTM 59
70◦ C and kept there for 1 hour to complete the first curing step. The part
is then demoulded when the mould is cooled down to room temperature
and post-cured in an oven for 2 hours at 100◦ C.
Production of composites from commingled yarns The Twin-

tex commingled yarns are processed by compression moulding in a hot
press with hot and cold stage (Pinette press) available at the depart-
ment. A picture frame (210 × 210 × 2 mm3 ) is placed in between two
steel plates to construct a mould in which to place the commingled yarns.
The press is preheated above the melting point of PBT. Table 4.1 shows
the temperature-pressure profile for the Pinette press during compres-
sion moulding. Both the hydraulic pressure as the actual pressure on the
material are listed.
Table 4.1: Pressure profile during compression moulding of com-

mingled yarns listing both the hydraulic and material pressure
t (s) T (◦ C) Phydr. (bar) Pmat. (bar)
90 240 11 0.7
5 240 155 36
30 240 11 0.7
After the pressure cycle, the press is opened and the mould is removed
to be placed immediately in between the cold press plates. This allows
for fast cooling of the composites and reduces the total cycle time since a
new plate can be hot-pressed simultaneously.
Production of (reinforced) matrix samples with hot press In

order to produce unreinforced or small samples, a home-built hot press was
used. The pressure obtained from closing the press plates was sufficient
for these samples, hence no extra pressure was used.
For the production of small samples, an aluminum picture frame mould
with four cavities was constructed to produce samples with dimensions of
50 × 50 × 2 mm3 . The resin was not infused into this mould but a small
amount of catalysed CBTr resin was prepared and poured into the mould
after which the mould was placed inside the hot press.
A larger picture frame was used for the production of thicker pure
polymer plates (200 × 200 × 3 mm3 ) needed for Charpy tests.
Injection moulding A lab-scale injection moulding machine is avail-

able both at the Chemical Engineering Department of the Faculty of En-
gineering and at the Chemistry Department of the Faculty of Science.
Grinded pCBT and PBT are molten in a small resin reservoir at around
250◦ C after which they are injected by a plunger mechanism into a small
mould. Two mould types were used, one producing tensile specimens, the
other producing Charpy specimens both according to the ISO standards.
4.4 Mechanical properties

4.4.1 Tensile and three point bending tests
The composites are tested in three point bending according to ASTM
D790-84. A load cell of either 1 or 30 kN is used depending on the fibre
orientation and test direction. Sample dimensions are 50 × 25 × 2 mm3
with a span length of 32 mm and the test speed is 0.85 mm/min. The tests
are performed on a universal Instron 4467. The same standard applies for
bending of pure polymer plates, advising however slightly different sample
dimensions of 50 × 10 × 2 mm3 .
Tensile tests are performed according to ASTM D3039-93. A load cell
of 30 or 100 kN is used again depending on the fibre orientation and test
direction. An extensometer with a gauge length of 50 mm and a maximum
extension of 12.5 mm is attached to the sample to accurately measure the
strain. Specimen dimensions are 200 × 25 × 2 mm3 and the test speed
equals 2 mm/min. In order to avoid slipping of the specimens in the grips,
sand-paper is placed in between the specimen and the grips. The tests
are performed on Instron 4467 and 4505.
4.4.2 Impact tests and damage evaluation

Drop-weight impact tests
Drop-weight impact tests at low speed are performed on a home-made
impact machine. A specimen with dimensions 60×60×2 mm3 is clamped,
leaving a square opening of 40 × 40 mm2 . For isotropic samples, a round
opening is more common, for orthotropic samples on the other hand, a
rectangular or square opening is often preferred with the sides of the
opening parallel to the axes of orthotropy. Moreover, this geometry is
similar to the unit cell of the reinforcement. The impacter used in these
tests is a so-called alufinger with a diameter of 13 mm.
During an instrumented impact test, the displacement is measured
with a laser beam and the force during impact is recorded with a load cell.
From the displacement data recorded, the impact energy (Eim ) as well as
the absorbed energy (Eabs ) can be calculated according to Equation 4.2
and 4.3.
mvi2
Eim = (4.2)
2¡ ¢
m vi2 − ve2
Eabs = (4.3)
2
4.4 Mechanical properties 61
ultrasonic
transducer
sample
glass plate
number of pixels
void free sample number of pixels
sample with voids
intensity (grey-scale) intensity (grey-scale)

Figure 4.6: C-scan experimental set-up and resulting his-
tograms for a void free sample compared to a sample with nu-
merous voids
with m the impact mass and vi and ve respectively the initial and final ve-
locity. The absorbed energy during impact is often related to the damaged
area. In order to determine this damaged area, ultrasonic C-scanning is
used.
The measured force data are corrected by a factor K, which is de-
termined by an energy-balance, to match the results obtained from the
displacement and force measurements. A detailed description of the im-
pact apparatus and the corrections to be applied can be found in [129].
Damage investigation by C-scan

In order to determine the damage area from the impact tests, all sam-
ples were subjected to an ultrasonic C-scan before and after the impact
test. The scans prior to impact are used as a quality control of the pro-
duced plates. The specimens are placed at a small distance from a glass
plate inside a waterreservoir, top of Figure 4.6. The ultrasonic wave is
transmitted through the sample and reflects from the glass plate. The
reflected wave repasses through the sample before being detected by the
transducer.
The intensity of the reflected wave is high if there is little scattering
and absorption of the wave. The intensity of the reflected beam is cor-
related to a specific grey-scale value. These grey-scale values are then
Metal tabs
Reinforcing tabs
Al-foil
Figure 4.7: Interlaminar fracture toughness (top) schematic of

sample geometry, (bottom) test set-up
used to create an image, assigning a grey-scale value corresponding to

the reflected intensity, to each pixel. From this image, a histogram can
be constructed which represents the number of pixels for each grey-scale
value. Defects, such as voids or delaminations lower the intensity of the
ultrasonic wave and are therefore represented as lighter on the grey-scale,
bottom of Figure 4.6.
After impact, the damage area is shown as a white spot on the C-
scan image or a peak at the lighter grey-scale side. The damage area can
hence be determined by choosing a threshold value above which the grey-
scale value corresponds to damaged material. It must be noted that the
choice of this threshold value is rather arbitrary. Moreover, this procedure
does not provide through the thickness information and cannot distinguish
between delaminations in between different plies.
4.4.3 Interlaminar fracture toughness

The interlaminar fracture toughness was determined according to ASTM
D5528-01. Although this standard is intended to determine the mode I
interlaminar fracture toughness of unidirectional composites, it is applied
to composites with a woven reinforcement since no specific standard exists
for these composites.
For interlaminar fracture toughness tests, the samples need a crack ini-
tiation. Therefore, plates are produced that contain a folded aluminium
foil in the middle. According to the standard, the sample thickness should
be larger than the available 2 mm, hence reinforcing tabs of 4 mm thick-
4.5 Matrix characterisation 63
ness are glued on each side. Figure 4.7 shows the sample geometry and
test set-up. More information on the sample preparation can be found in
[130].
The tests are performed on an Instron 4505 with a constant opening
speed of 2 mm/min and the crack formation is recorded with a camera.
With this test set-up, the crack length, e, can be correlated to the cor-
responding force, F , and crack opening, δ. The modified beam theory is
used for the data reduction resulting in the following expression for the
crack propagation energy, GI :
3F δ
GI = (4.4)
2be
with b the width of the sample.
For each sample, a correction is applied to account for imperfect beam
clamping by assuming an longer crack length (e + |∆|), Equation 4.5.
3F δ
GIc = (4.5)
2b(e + |∆|)
1
∆ can be determined experimentally by a linear least-square fit of ( Fδ ) 3
in function of the crack length, e. The intersection of the resulting curve
with the X-axis equals ∆.
A so-called R-curve is obtained by plotting the corrected crack prop-
agation energy, GIc , versus the crack length, e. The first value obtained
is called the initiation value, whereas the plateau-value of the R-curve is
called the propagation value.
4.4.4 Charpy
Although not suited to determine the real fracture toughness, Charpy
tests are often employed to compare the (impact) toughness of different
materials. The test set-up is schematically shown in Figure 4.8. The
Charpy impact tests are performed at the Chemistry Department of the
Science Faculty. This set-up allows for low energy testing, a striker with
a potential energy of 1 J is employed. The tests are performed according
to ISO 179-1:200 with sample dimensions of 80 × 10 × 3 mm3 . The notch
is produced for each sample with a special milling machine.
4.5 Matrix characterisation

4.5.1 Gel permeation chromatography [1, 102]
Gel permeation chromatography (GPC) is a traditional analytical tech-
nique for the determination of polymer molecular weight averages and
distributions. It belongs to a group of separation techniques referred to
sample
support
notched
striker
specimen
Figure 4.8: Schematic of experimental set-up for Charpy impact

testing on a specimen with a V-notch
as size exclusion chromatography (SEC). GPC is a liquid column chro-

matographic technique in which a sample solution is introduced onto a
column filled with a rigid porous gel and is carried through the column
by a solvent. Ideally, size separation is achieved by differential pore per-
meation. All molecules experience a solute-to-wall exclusion effect inside
the pore. Owing to greater steric interference, larger molecules are kept
away from the wall of the pore. The volume of the pore that is effectively
accessible, is thus greater for a small molecule than for a large one, Figure
4.9. Under the influence of the solvent stream passing down the column,
larger molecules are eluted from the column earlier than smaller ones.
Solvent
Xc,1
Flow
Xc,2
Polymer chains
Pore wall
Figure 4.9: Schematic representation of size separation mecha-
nism in GPC, adapted from [1]
GPC is not an absolute method for the determination of molecular

weights and calibration is required to provide a relationship between the
time or amount of liquid eluted and molecular size. The simplest type of
calibration is peak position calibration using polymer standards of nar-
row molecular weight distribution. The separation in GPC is however
based on the hydrodynamic volume of a polymer molecule in the solu-
tion rather than on molecular weight. The product of intrinsic viscosity,
[η] and molecular weight, M , was found to be directly proportional to
the hydrodynamic volume. Therefore all polymers, regardless of chemical
Figure 4.10: Illustration of a universal calibration curve
structure or architecture, should fit on the same plot of log([η]M ) versus

elution volume of retention time. This plot is referred to as the universal
calibration curve.
The intrinsic viscosity of a polymer in a solution at a certain temper-
ature can be calculated with the Mark-Houwink equation, with a and K,
the Mark-Houwink constants, Equation 4.6.
[η] = KM a (4.6)
A universal calibration curve is thus constructed by using monodis-

perse polystyrene (PS) standards. With this calibration curve, the molec-
ular weight of the sample can be calculated, providing that the Mark-
Houwink coefficients are known, Figure 4.10 and Equation 4.7.
[ηi ]Mi a 1 +1
MP BT,i = ( ) P BT (4.7)
KP BT
The measurements were performed at the polymer lab at U.C.L.2 with
a mixture of chloroform/hexafluoro-2-propanol (98/2 CHCl3 /HFIP) as
solvent. The flow rate was 0.8 ml/min at a temperature of 20◦ C. Ta-
ble 4.2 shows the Mark-Houwink coefficients for polystyrene and PBT for
these conditions. Two Waters PL HFIP-gel columns were used in series.
The chromatograph was connected to a Waters 484 UV detector working
at 254 nm. For sample preparation, approximately 2 mg of matrix was
dissolved in 80 µl of HFIP. After total dissolution, the solution was diluted
by 4 ml of chloroform. If fibres are present in the sample, the solution is
filtered before passing through the column in order to remove all fibres.
2 Unité de chimie et de physique des hauts polymères, Université Catholique de
Louvain
Table 4.2: Mark-Houwink coefficients for PS and PBT in (98/2

CHCl3 /HFIP) solution at 20◦ C
a K
PS 0.693 0.000191
PBT 0.731 0.000191
The degree of conversion was also determined from these GPC mea-
surements by comparing the amount of remaining oligomers to the amount
of polymer and is calculated according to
Aoli
α=1− (4.8)
Atot
with Aoli the area under the oligomer peaks of the retention time
curve and Atot the total area under the retention time curve. A typical
GPC chromatogram is shown in Figure 4.11. Both the number, Mn , and
weight, Mw , average molecular weight, as defined in Equations 4.9 and
4.10, were determined for the samples. From this, the polydispersity can
be calculated, Equation 4.11.
P
ni M i
i
Mn = P (4.9)
ni
i
P
ni Mi2
i
Mw = P (4.10)
ni Mi
i
Mw
δ= (4.11)
Mn
4.5.2 Thermal Analysis

A T.A. Instruments 2920 DSC was used to determine the degree of crys-
tallinity in the produced samples. Open aluminum pans were chosen in
order to easily recover the samples after the measurements. The samples
were kept under a nitrogen flow of 30 ml/min and the DSC was calibrated
using indium and tin standards.
Melting endotherms were recorded at 10◦ C/min and the melting en-
thalpy of the polymer was determined by integrating the area under the
normalised melting peak after subtraction of an arbitrary baseline. The
degree of crystallinity, χc , was calculated according to the definition in
Equation 4.12.
4Hm
χc,dsc = (4.12)
4H∞
Figure 4.11: Typical GPC chromatogram of unreinforced pCBT

showing the polymer peak and the peak of the remaining oligo-
mers
where 4Hm is the melting enthalpy of the polymer, and 4H∞ is the
melting enthalpy of the fully perfect crystal of PBT, which is found in
literature to be 142 J/g and 134 J/g for the α phase and β phase re-
spectively [127]. Since the β phase is usually only obtained by drawing
of PBT, the crystallinity of polymerised CBT (pCBT) is calculated by
assuming to consist of 100% α form.
In order to account for the fibres in the composite sample, the fibre
weight fraction is determined by Thermogravimetrical Analysis (TGA)
with a T.A. Instruments 951 TGA. The sample is heated from room tem-
perature to 650◦ C with a heating rate of 10◦ C/min. Platinum holders
were used and the samples were heated under an air flow.
4.5.3 Transmission Electron Microscopy

Transmission electron microscopy (TEM) is often employed to investigate
the lamellar structure of polymers. Unfortunately, contrast between crys-
talline and amorphous parts is often very low, therefore, for each polymer,
a suitable staining method needs to be developed.
Because of their experience with PBT, the samples were prepared at
the polymer lab of U.C.L.3 . First, the sample surfaces were prepared
for contrast colouring. Next, the samples were stained with ruthenium
tetroxide by exposing them for 48 hours to a RuCl3 × NaClO vapour after
3 Unité de chimie et de physique des hauts polymères, Université Catholique de
Louvain
Figure 4.12: Typical WAXD scattering pattern of pCBT show-

ing the fitted amorphous halo and the principal crystalline reflec-
tions
which they were cut for the first time with a diamond knife. The staining
procedure was repeated before recutting the samples and depositing them
on a TEM grid. TEM micrographs were obtained on a LEO 922 trans-
mission electron microscope operating at 200 kV. The samples obtained
for TEM were also examined by polarised optical microscopy.
4.5.4 Light and X-ray scattering

X-ray scattering The principles of X-ray scattering are explained in
more detail in Appendix B. Scattering patterns of the unreinforced pCBT
were obtained in transmission mode with a Rigaku Rotaflex RTP 30RC
goniometer at the Chemistry Department of the Science Faculty of the
K.U.Leuven. The experiments used Cu Kα radiation (40 kV-100 mA)
and the angular range covers 5◦ < 2θ < 60◦ , with 2θ the scattering angle.
The resulting patterns were corrected by subtracting both a scaled empty
cell measurement and a linear background. The shape of the amorphous
halo was determined by quenching CBTr resin and recording the scat-
tering pattern. This halo was then scaled to fit underneath the scattering
pattern of the samples to determine the integrated intensity of the amor-
phous halo (Ia ) and of the crystalline diffraction peaks (Ic ). The degree
of crystallinity was calculated according to Equation 4.13, comparing the
integrated intensity of the crystalline diffraction peaks to the overall in-
tegrated intensity (Ia + Ic ). A typical scattering pattern of polymerised

CBT is shown in Figure 4.12.
Ic
χc,waxd = (4.13)
Ia + Ic
Scattering patterns of fiber reinforced pCBT were obtained in reflec-
tion mode with a Siemens D500 goniometer. All experiments used Cu Kα
radiation (40 kV-40 mA) and the angular range is 5◦ < 2θ < 50◦ .
Small angle laser light scattering (SALLS) SALLS was measured

on an home-made vertical apparatus consisting of a polarised 1 mW
Spectra-Physics 117A type He/Ne laser (λ = 632.8 nm), a polariser set
parallel to the polarisation of the laser, the sample in a Mettler FP-82HT
hot stage, a second polariser (analyser) oriented with its polarisation per-
pendicular (Hv SALLS) to that of the first one, a screen with a beam
catcher on which the scattering patterns are projected and a Photometrix
ATC200L cooled CCD detector. The scattering angle is calibrated with
a 100 lines/mm grid. This apparatus was available at the Chemistry
Department of the Science Faculty of the K.U.Leuven.
Samples with a thickness of only a few times the spherulitic radius
are needed for these tests. Therefore, for the CBTr resin, the specimens
were molten in between two glass plates and subsequently placed in a
Mettler hot stage to be subjected to a specific temperature profile. The
PBT samples were first molten in between the glass plates after which
they were squeezed in order to obtain a thin film. These samples were
then replaced into the Mettler hot stage to be subjected to a specific
temperature profile.
Chapter 5
Composites: Production
and Properties
Resin transfer moulding with thermoplastics is a new production method

for continuously reinforced composites, which has recently received in-
creased attention due to some developments in the polymer industry. Data
on the characterisation of these composites are however scarce in litera-
ture, as was pointed out in the literature review. Therefore, this chapter
describes the production and characterisation of these composites, paying
special attention to the properties of the in-situ polymerised matrix.
5.1 Production
The processing window of CBTr resin and more specifically, the time
window for impregnation are crucial aspects in thermoplastic RTM (Sec-
tion 2.2.2). The impregnation time was defined as the time for the reactive
polymer system to reach a viscosity of 1 Pa·s. Depending on the poly-
merisation and thus mould temperature, the impregnation time varies
from seven (180◦ C) to two minutes (220◦ C) [88]. Hence, the time win-
dow for impregnation is quite narrow, leading to the necessity to carefully
control the processing parameters as small changes can drastically affect
the final part quality.
5.1.1 Production of biaxial composites

The general production parameters were discussed in Section 4.2.3 and
Figure 4.5. Table 5.1 provides the mould filling times for the processing
of biaxially reinforced pCBT (S-B). It is clear that the mould filling time
(tP3 ) is depending on both the melt temperature of the oligomer/catalyst
mixture and the stirring time. Increasing the stirring time (plate 1/2
71
72 Composites: Production and Properties
Table 5.1: Processing parameters for biaxial (±45◦ ) composites,

tP1 is the time at which the vacuum was raised for the first time,
the total mould filling time is tP3 , the volume fractions are calcu-
lated from the measured masses and known mould volume
plate 1 plate 2 plate 3 plate 4 plate 5
mfabric (g) 152.0 151.2 156.2 154.9 156.2
mCBT (g) 200 200 200 200 155
mcat (g) 0.91 0.92 0.91 0.90 0.70
TCBT (◦ C) 190 191 190 186 191
Tmould (◦ C) 190 189 190 190 190
tstir (s) 30 30 12 11 12
tP1 (s) 60 60 42 42 42
tP2 (s) 90 90 60 60 60
tP3 (s) 160 210 180 160 180
tpol (min) 30 30 30 30 30
mplate (g) 217.4 227.9 240.6 239.4 242.3
Vf (%) 47 47 48 48 48
Vm (%) 39 46 51 51 52
Vunfilled (%) 14 7 1 1 0
versus plate 3/5), increases the initial conversion and thus the viscosity,
which leads to a smaller effective time window for impregnation. If the
time window is too small for the vacuum profile applied, the mould is
not completely filled as can be seen in Figure 5.1. This is also obvious
from the calculated unfilled volume fraction, which includes both voids
and incomplete filling.
Decreasing the melt temperature on the other hand (plate 4 versus
plate 3/5), decreases the reaction speed inside the reaction vessel, resulting
in a lower degree of conversion at tstir . A lower reaction speed results in a
slower viscosity build-up and hence a larger time window. If the vacuum
profile is not altered, a lower reaction speed inside the reaction vessel
leads to a shorter filling time. Lowering the temperature in the reaction
vessel can hence be considered to increase the available time window for
impregnation without introducing a thermal mould cycle [131].
When the process parameters are kept constant, mould filling times
are reproducible as can be seen from comparison of plate 3 and 5.
Filling times depend on a number of parameters. First of all, the
viscosity profile of the resin strongly influences the resin flow and the
available time window for impregnation. As important as the available
time window for impregnation is the needed time window for impregna-
tion. This not only depends on the preform permeability but also on the
used injection strategy, including the injection and vent locations and the
pressure profile. In order to optimise the production process, each case
5.2 Matrix properties 73
inlet
plate 2 : 30s plate 5: 12s
incomplete fill
outlet
Figure 5.1: Effect of stirring time on filling process of biaxially

reinforced pCBT
needs to be considered separately. Moreover, a suitable injection strat-

egy needs to be developed by using mould filling simulations taking the
changing viscosity into account [132] and preferably the capillary effects
which were shown to be important for very low viscous resins and small
injection pressures [50]. With such simulations, the available time win-
dow can be matched with the needed time window for impregnation. To
perform these kind of simulations was however not the goal of this study.
Emphasis was therefore placed on studying the properties of the resulting
composites.
5.2 Matrix properties

The processing route of thermoplastic RTM differs significantly from con-
ventional thermoplastic processing since the matrix is polymerised in-situ.
Therefore, special attention is paid to the properties of the matrix within
the composite and the influence of the glass fibres present during the poly-
merisation and crystallisation. An overview of the matrix properties for
all investigated specimens is given in Section 5.2.3.
5.2.1 Composites produced by TP-RTM

Degree of conversion and molecular weight
GPC (Section 4.5.1) is used to measure the degree of conversion and the
molecular weight, the results are depicted in Figure 5.2. The degree of
conversion, calculated by Equation 4.8, seems to be lower when glass fibres
are present during polymerisation (S-UD to ECA-UD).
As glass fibres can be considered inert, this effect should be related to
the fibre sizing. In order to assess the influence of the glass fibre sizing on
the GPC data, pure glass fibre samples were prepared for GPC-analysis in
Figure 5.2: GPC results for plates produced with TP-RTM,

(left) degree of conversion, (right) weight average molecular
weight. Abbreviations for the different materials are explained
in Section 4.1.2, p 52
the same way as the pCBT samples, hence investigating the GPC response
of the soluble part of the sizing. The results are presented and compared to
the response of reinforced and unreinforced pCBT samples in Figure 5.3.
It is clear that the glass fibre sizing signal overlaps with the oligomer
peaks, situated in the high retention time region.
Therefore, part of the peaks in this region which are assumed to be
CBT oligomer peaks, actually originate from the glass fibre sizing. More-
over, the shape of the GPC trace of the reinforced sample in the high
retention time region resembles the shape of the fibre sizing response,
only slightly showing the two oligomer peaks. It is virtually impossible
to separate the fibre sizing response from the oligomer response. It would
be possible to carefully determine the fibre fraction of each sample by
weighing the remaining fibres after matrix dissolution. Nevertheless, the
fibre sizing loading can differ significantly along the fibre length. More-
over, samples containing different amounts of sizing should be prepared
to calibrate the detector response of the sizing. Hence, it is more practi-
cal to consider the high retention time peaks as only being related to the
oligomers, which leads to an overestimation of the fraction of remaining
oligomers. In this way, the degree of conversion for the composite samples
must be seen as a lower boundary for the actual degree of conversion.
Classically polymerised PBT normally contains 1-3% oligomers [71].
Considering that the determined degree of conversion is the lower bound-
ary of the actual degree of conversion and taking into account the equi-
librium amount of oligomers, the degree of conversion for most glass fibre
reinforced samples is satisfactory.
Besides the degree of conversion, also the molecular weight is deter-
mined with GPC measurements, Figure 5.2. Both the number and weight
average molecular weight of pCBT decrease when glass fibres are present.
Since the molecular weight is controlled by the amount of linear oligomers
and the catalyst level [59, 71] and because it can be assumed that the glass
Figure 5.3: Influence of the fibre sizing signal on the GPC re-
sults, indicating an overlap between the fibre sizing and the oli-
gomer peaks
fibre sizing introduces extra end-groups, the final molecular weight will
be reduced by the presence of the sized fibres. Similarly, the introduction
of extra PBT end-groups during processing of reinforced PBT has been
observed previously for non-reactive processing [133].
Even though the maximum molecular weight is not reached in the
composite matrix, the final molecular weight is still comparable to that of
compression moulded Twintexr . Moreover, for most samples, the weight
average molecular weight exceeds the critical molecular weight for entan-
glement for PBT, which is found to be around 50 kg/mole [79]. Only for
the ECA-UD composites, questions rise if the fibres, and more specifically
their sizing, used in this fabric can still be considered compatible with the
reactive matrix system since both the molecular weight and degree of
conversion are rather low. Moreover, the appearance of these composites
was more yellowish. Non-crystalline, low molecular weight oligomers also
have this yellowish colour. Therefore, this colour equally indicates that
the fibre sizing interferes with the polymerisation reaction, leading to poor
sizing-matrix compatibility.
Figure 5.3 showed a typical chromatogram of fibre reinforced pCBT.
An unexpected peak arises at a very low retention time, suggesting the
presence of a very high molecular weight substance, which is believed to
be partially crosslinked PBT. This peak is not present in non-reinforced
pCBT, also shown in Figure 5.3, indicating that reactive functions in the
fibre sizing promote these crosslinks.
In order to further investigate the origin of this high molecular weight
fraction, the glass fibre sizing was removed by burning off the fibre sizing
at 600◦ C for 12 hours. After this treatment, the naked fibres were used to
Figure 5.4: Crosslinking of pCBT due to side-reaction with glass

fibre sizing
reinforce pCBT (S-UD burnt). The resulting chromatogram is depicted in

Figure 5.4 and compared to pCBT reinforced with sized fibres. When the
sizing is removed, the low retention time peak is absent, clearly demon-
strating that the fibre sizing is responsible for the high molecular weight
fraction of pCBT.
Glass fibre sizings consist of a film former, a surfactant and a silane
coupling agent. The film formers often contain epoxide functional groups,
which are known to react with PBT (Equation 5.1) since they are used in
the compatibilisation of PBT blends [134–136].
O O O OH
PBT C O H + CH2 CH PBT C O CH CH2
carboxyl reaction
O OH
PBT O H + CH2 CH PBT O CH CH2
hydroxyl reaction (5.1)
Depending on the functionality of the epoxide groups and the presence

of initiators, it has been shown that various crosslinking effects may oc-
cur [134, 137]. Bergeret et al. [133] studied the interphase in glass fibre-
reinforced PBT composites. They clearly observed a reaction of the PBT
with the epoxide functional groups of the fibre sizing, improving the fi-
bre/matrix adhesion. Moreover, the presence of the catalyst might also
enhance this (side) reaction. The analysis of the sizing components is
however beyond the scope of this work but in order to improve or tailor
the composites’ properties, it is crucial to develop suitable and specific
fibre sizings.
Figure 5.5: Comparison of GPC results for isothermally and

semi-isothermally produced samples (left) degree of conversion,
(right) number and weight average molecular weight
Semi-isothermal processing One option to improve the time window

for impregnation is semi-isothermal processing. In a semi-isothermal pro-
cess, the mould is kept at a constant temperature (no thermal cycle), but
the oligomer melt is at a lower temperature in order to retard the reaction
inside the vessel and hence lower the initial viscosity. Glass fibre compos-
ites (A-UD semi) were produced by keeping the oligomer melt at 170◦ C
with a mould temperature of 190◦ C.
Figure 5.5 depicts the results of the GPC measurements. It is clear
that for semi-isothermal processing, both the degree of conversion and
the final molecular weight are lower than with the isothermal process.
This is most probably due to incomplete melting of some of the smaller
cyclics which therefore cannot polymerise. One option would be to first
heat the oligomer melt to around 200◦ C and then cool down to 170◦ C
before adding the catalyst. This will would most probably improve the
conversion properties.
Degree of crystallinity
DSC and TGA measurements, as described in Section 4.5.2, are used to
determine the degree of crystallinity of the composites, Figure 5.6. The
degree of crystallinity is lower in reinforced pCBT. This is most probably
due to the partially crosslinked pCBT which cannot crystallise and might
even hinder the nucleating effect of the fibres. Compared to the Twintexr
composites, which are crystallised upon cooling from the melt, the degree
of crystallinity of all pCBT samples, isothermally produced, is higher.
The righthand-side of Figure 5.6 shows a typical DSC curve of re-
inforced pCBT compared to a Twintexr sample. No multiple melting
behaviour is observed in these samples. There is, however, a noticeable
difference. Firstly, in the Twintexr sample, a small bump is noticed prior
to melting indicating cold crystallisation of the specimen. Moreover, the
melting peak is significantly broader indicating poorer crystal quality com-
Figure 5.6: Results from thermal analysis for plates produced

with RTM (left) degree of crystallinity, (right) typical (shifted)
DSC traces upon heating at 10◦ C/min
pared to the pCBT composites. The effect of the starting material (PBT
versus CBTr resin) and processing route on the crystalline structure will
however be discussed more in detail in Chapter 6.
5.2.2 Composites produced in the hot press

Even though the effect of the glass fibres and sizing on the matrix prop-
erties was discussed in the previous section, one might argue that, due to
reproducibility issues, the thermal history of the reinforced and unrein-
forced samples can differ. Therefore, samples containing (1) no fibres, (2)
glass fibres and (3) burnt-off, naked glass fibres were produced simultane-
ously in the home-built hot press (Section 4.3). Compared to TP-RTM,
the reactive mixture is not infused into the mould, but catalysed CBTr
resin is poured onto the fabric, impregnating the fibre bundles. Therefore,
almost no flow is induced during the actual polymerisation reaction.
Two layers of glass fibre reinforcement (S-UD) were used, resulting
in a somewhat lower fibre volume fraction of 37% compared to the TP-
RTM samples (54%). Two temperature profiles were used, isothermal
processing at 190◦ C for 40 minutes and non-isothermal processing where
the hot press was kept at 240◦ C for 20 minutes after which the samples
were cooled down slowly inside the hot press.
Influence of glass fibres on matrix properties

Figure 5.7 shows the matrix properties of the samples produced with the
hot press. For both temperature profiles, the molecular weight decreases
in the presence of glass fibres but increases again when the fibre sizing
is removed. At 190◦ C, the temperature profile is comparable to that in
TP-RTM, however, the decrease in molecular weight is less pronounced
compared to the observed decrease when processing with TP-RTM, Fig-
ure 5.2. The reason for this is twofold. First, the fibre volume fraction and
Figure 5.7: Matrix properties of hot press samples, produced

without fibres (pCBT), with 2 layers of S-UD (+ GF) and with
2 layers of S-UD from which the sizing was removed by burning
(+B-GF)
hence the amount of fibre sizing is lower in the hot press samples. Second,
less flow along the fibres is induced in the hot press during the polymeri-
sation reaction, limiting the contact between the fibres and matrix. The
same effect was observed by comparing the relative area of crosslinked
material in the GPC chromatograms, which showed a decreased effect of
the glass fibres in the samples produced with the hot press, Figure 5.8.
The molecular weight recovery after burning off the sizing (+B-GF)
is complete when processing at 190◦ C, clearly confirming the hypothesis
that the sizing introduces end-groups, limiting the achievable molecular
weight. At 240◦ C, however, the decrease in molecular weight was not
only more pronounced, the recovery upon removing the sizing is only
partial. Thermal degradation effects [138, 139] at this higher temperature
are the most probable cause of these effects since glass fibres and their
sizing might intensify the thermal degradation. Moreover, unsized glass
fibres are more prone to water absorption, also influencing the molecular
weight. An increase in time in between the sizing burn-off and the actual
processing, might increase the water content and have a negative effect
on the molecular weight.
The decrease in degree of crystallinity with the presence of glass fibres,
present in the TP-RTM samples, was not observed. In contrast, there is
even a small increase. Again this can be attributed to the lower fibre
content and hence a smaller influence of the crosslinked pCBT.
More pronounced is however the difference in degree of crystallinity
between isothermal and non-isothermal processing. Not only is, in the
isothermally produced samples, the overall degree of crystallinity higher
but also the melting peak is better defined without traces of multiple
melting endotherms, Figure 5.9. As was expected (Section 2.5.1), the
possibility for crystallisation during polymerisation in the isothermal pro-
cess influences the crystalline structure as will be discussed in Chapter 6.
Figure 5.8: Relative amount of crosslinked pCBT, as deter-

mined from the GPC measurements, comparing pCBT reinforced
with S-UD produced with TP-RTM (Vf = 54%) and in the hot
press (Vf = 37%)
Figure 5.9: Melting behaviour of pCBT after isothermal pro-

cessing at 190◦ C and non-isothermal processing with polymerisa-
tion at 240◦ C and slow cooling
5.2.3 Overview of matrix properties
Table 5.2: Overview of matrix properties determined by GPC and thermal analysis, symbols are defined in Chapter 4
α (%) Mn (kg/mole) Mw (kg/mole) δ (-) χc (%) Tm (◦ C)
S-UD 93.3 ± 1.9 25.4 ± 2.1 66.9 ± 3.8 2.64 ± 0.10 31 ± 2 227 ± 1
S-UDburnt 93.1 ± 0.3 22.1 ± 0.2 44.9 ± 0.6 2.03 ± 0.00 34 ± 1 223 ± 1
R A-UD iso 96.4 ± 3.8 31.6 ± 2.8 76.7 ± 4.2 2.44 ± 0.10 30 ± 1 229 ± 1
T A-UD semi 89.1 ± 2.2 22.8 ± 3.0 56.7 ± 5.8 2.49 ± 0.11 n.a. n.a.
5.2 Matrix properties
M ECA-UD 91.8 ± 2.0 22.7 ± 1.3 47.1 ± 2.1 2.08 ± 0.03 29 ± 4 222 ± 1
S-B 94.1 ± 2.0 22.8 ± 2.8 56.9 ± 6.7 2.50 ± 0.08 33 ± 2 223 ± 1
pCBT 98.3 ± 1 38.6 ± 4.5 85.1 ± 12.4 2.26 ± 0.05 39 ± 1 226 ± 1
pCBT-190 98.1 ± 0.8 33.9 ± 0.8 78.4 ± 2.1 2.31 ± 0.02 38 ± 0 227 ± 1
P + GF 98.2 ± 0.3 29.3 ± 2.4 70.2 ± 3.1 2.40 ± 0.10 41 ± 1 228 ± 1
R + B-GF 98.1 ± 0.4 34.7 ± 0.3 78.4 ± 0.2 2.26 ± 0.01 37 ± 1 229 ± 1
E pCBT-240 99.2 ± 0.0 35.5 ± 0.9 79.1 ± 2.1 2.23 ± 0.00 30 ± 0 227 ± 2
S + GF 96.4 ± 2.7 24.5 ± 1.7 59.0 ± 2.5 2.41 ± 0.06 33 ± 2 224 ± 1
S + B-GF 97.3 ± 0.6 32.3 ± 0.4 70.5 ± 1.0 2.19 ± 0.01 33 ± 2 226 ± 1
Twintex 98.8 ± 0.0 28.6 ± 0.4 59.8 ± 0.9 2.09 ± 0.01 24 ± 1 227 ± 1
PBT 98.8 ± 0.0 33.8 ± 0.4 69.3 ± 1.0 2.05 ± 0.03 35 ± 1 228 ± 1
81
5.3 Tensile and flexural properties

Because the processing route followed in thermoplastic RTM does influ-
ence the conversion and molecular weight build-up, it is important to
characterise the matrix properties carefully. Nevertheless, the ultimate
goal is to achieve satisfactory composite properties since these properties
are the final design parameters. In this section, the tensile and flexural
properties are determined experimentally and the moduli are compared
to theoretical predictions. An overview of all the results is given in Sec-
tion 5.3.5.
5.3.1 Theoretical stiffness

The experimental mechanical properties are better assessed when com-
pared to the predicted mechanical properties. In order to predict the the-
oretical stiffness of the produced laminates, the classical laminate theory
is used, which is available in the commercial software, Composite StarTM
and KolibriTM .
Ply properties
Within the available version of the software, ply properties are not cal-
culated, therefore, these properties need to be calculated using following
simple formulas [140]. The employed fabrics all contain plies with contin-
uous unidirectional fibres. One of the non-crimp fabrics, namely A-UD,
however also consists of a random mat of discontinuous fibres, which re-
quires a different set of equations.
Properties of continuous unidirectional plies The simple rule of

mixtures based on the iso-strain criterion is used to calculate the longitu-
dinal modulus, E11 , Equation 5.2. A similar formula can also be applied
to determine the Poisson coefficient, ν12 , of the ply. The transverse mod-
ulus, E22 , is however seriously underestimated when using a simple rule
of mixtures based on the iso-stress criterion. There are two well known
corrections to this rule of mixtures. Firstly, a stress partitioning factor,
η, can be introduced, Equation 5.3, and this equation also holds for the
shear modulus, G12 . Chamis [140] on the other hand suggested a modified
formula based on his experimental results, Equation 5.4. The transverse
modulus calculated according to Equation 5.4 is somewhat higher than
when using a stress partitioning factor of 0.5. Therefore, both calculated
transverse moduli were used to calculate the composite properties that
were then compared to the experimental data.
E11 = Vf Ef + Vm Em (5.2)
µ ¶
1 1 Vf ηVm
= + (5.3)
E22 Vf + ηVm E2,f Em
5.3 Tensile and flexural properties 83
Em
E22 = p ³ Em
´ (5.4)
1 − Vf 1 − E2,f
Properties of a short fibre unidirectional ply The equations to cal-

culate the longitudinal, transverse and shear modulus are given in respec-
tively Equations 5.5, 5.6 and 5.7 with lf and df the length and diameter
of the fibre bundles.
lf Ef
1+2 ηL Vf
df Em −1
E11 = 1−ηL Vf Em ηL = Ef lf (5.5)
Em +2 df
Ef
1+2ηT Vf −1
E22 = 1−ηT Vf Em ηT = Em
Ef (5.6)
Em +2
Gf
1+ηG Vf −1
G12 = 1−ηG Vf Gm ηG = Gm
Gf (5.7)
Gm +1
Properties of a randomly oriented discontinuous ply Using the

unidirectional, short fibre ply properties, the properties of the random
mat can be calculated using Equations 5.8-5.10.
3 5
Erandom = E11 + E22 (5.8)
8 8
1 1
Grandom = E11 + E22 (5.9)
8 4
Erandom
νrandom = −1 (5.10)
2Grandom
The material properties of both reinforcement and matrix are given in
Table 5.3.
Table 5.3: Properties of fibres and matrix

E (GPa) G (GPa) ν (-) ρ (g/cm3 )
PBT 2.4 0.9 0.38 1.3
E-glass fibres 72 30 0.28 2.54
Composite properties
The final composite properties are calculated with the classical laminate
theory. The fibre lay-up and the distance of each layer to the mid-plane is
very important for the flexural properties. For woven fabrics (Twintexr )
and braids (ECA-UD), the composite does not consist of unidirectional
Figure 5.10: Flexural moduli of unidirectional, glass fibre rein-

forced pCBT (left) 0◦ (right) 90◦
plies since the fibres are interlaced. Therefore, to compare the experimen-
tal results to the theoretical predictions, especially in three point bending,
the modulus was calculated for the different possible stacking sequences
(e.g. (0,90)4,s and (90,0)4,s ) and an average over the different lay-ups was
considered. Even when considering an average lay-up, fibre crimp is not
taken into account in the laminate theory and this crimp is known to
lower the mechanical properties.
5.3.2 Unidirectional composites

Figure 5.10 depicts the flexural moduli of the unidirectional composites.
The theoretical longitudinal stiffness of S-UD and A-UD compares very
well to the experimental value, for ECA-UD however the difference is
almost 10%. Since the longitudinal stiffness is almost insensitive to the
matrix properties, this difference cannot by caused by the lower matrix
properties. Taking a closer look at the fabric, which resulted from cutting
open a tubular and widening braid1 , discloses the cause of this difference,
namely fibre misorientations, Figure 5.11. The influence of small changes
in the fibre orientation is given in Table 5.4 in which the longitudinal and
transverse ply properties are calculated with respectively Equation 5.2 and
5.4. It is clear that even a small deviation of 5◦ can result in a decrease of
the longitudinal modulus of 8%. The discrepancy between the theoretical
and experimental longitudinal modulus of ECA-UD is therefore attributed
to fibre misalignment.
The transverse modulus on the other hand is a matrix-dominated prop-
erty as can be seen by simplifying Equations 5.3 and 5.4. Therefore, poor
matrix properties will be translated in larger discrepancies in the trans-
verse direction. It is however well known that matrix and interface dom-
inated properties are not predicted as well as fibre dominated properties,
1 This braid was developed for a leaf spring prototype and had to be cut open to
produce flat plates, the widening shape results from the shape of the leaf spring as
shown in Appendix A
-5°
0°
+5°
Figure 5.11: Cut-open braid resulting in the unidirectional fab-

ric ECA-UD revealing the deviation from the 0◦ orientation
Table 5.4: Effect of small changes in fibre orientation on the flex-

ural modulus of ECA-UD composites, deviation from 0◦
0 2 -2 5 -5
Ef l,0◦ 29.2 28.6 29.4 26.9 28.7
Ef l,90◦ 7.85 7.91 7.77 7.96 7.63
therefore, the discrepancy between theory and experiments is always ex-

pected to be larger perpendicular to the fibre direction. Even though
the laminate theory does not take fibre crimp into account and hence the
model is not correct for braids, the longitudinal properties are predicted
quite well taking the fibre misalignment into account. The discrepancy in
transverse modulus of more than 40%, however, indicates that ECA-UD
does not obtain acceptable transverse properties, Figure 5.10. This result
is consistent with the poorer matrix characteristics of ECA-UD compared
to S-UD and A-UD, Table 5.2.
In order to investigate the effect of lower matrix properties on the
modulus, the composite properties were recalculated for different matrix
moduli using Equations 5.2 and 5.3. As expected, the modulus in the
fibre direction was hardly affected whereas the transverse modulus of the
laminates shows a quasi linear relationship2 from which a matrix modulus
can be backcalculated, Figure 5.12. The experimental transverse modulus
is hence used to calculate the matrix modulus. Even though this matrix
modulus is lower than expected for all composites (Em < 2.4), it is clear
that the overall poorer matrix characteristics, such as degree of conversion
and molecular weight, of the ECA-UD composites (Table 5.2), lead to a
much lower predicted matrix modulus compared to the S-UD and A-UD
composites.
Tensile tests were only performed for S-UD and A-UD composites.
These results are included in the overview table in Section 5.3.5. The
2 Using the classical laminate theory, it can be shown that the relationship between
Ef l and Em is not linear. However, in the small Em -range (1-3 GPa) considered here,
an (empirical) linear relationship is an acceptable approximation
Ef l,90◦ Em
S-UD 3.0 × Em + 2.0 2.23
A-UD 2.8 × Em + 0.5 2.25
ECA-UD 2.5 × Em + 1.1 1.24
Figure 5.12: Influence of matrix modulus on 90◦ laminate mod-

ulus, transverse ply properties are calculated with a stress parti-
tioning factor of 0.5. From the calculated values, a linear relation
between the transverse modulus and matrix modulus was deter-
mined. By using these equations and the experimental transverse
modulus from Figure 5.10, the matrix modulus was calculated
theoretical predictions of the modulus compare well to the experimental

results as was expected from the flexural tests.
The tensile and flexural strength in fibre direction and perpendicular
to the fibre direction are shown in Figure 5.13. Generally, the tensile
strength is lower than the flexural strength due to the larger area exposed
to the maximum stress, assuming that the main failure mode in flexural
testing was tensile failure. For the S-UD composites, the tensile strength is
higher than the flexural strength, which can be explained by the fibre lay-
up. Since the outer layer at the tensile side of the flexural test specimen
consists of a small amount of 90◦ fibres (3%), transverse failure of this
layer will be initiated at low bending force. Thus, the strength is lower
than when the (non-symmetric) samples would be turned upside down,
putting this transverse layer in compression.
The effect of the small amount of 90◦ fibres strongly increases when the
UD-samples are tested in transverse direction. This will be discussed in
the case of the S-UD composites. For these samples tested in transverse
direction, the 0◦ -layers (97% of each ply) form a thick but weak layer.
Figure 5.14 shows a typical flexural stress-strain curve. Despite the almost
unidirectional fibre orientation, this stress-strain curve exhibits the typical
‘cross-ply’ behaviour, namely a knee point.
The knee in this curve at 0.5% strain and at a stress of around 40 MPa,
indicates failure of the outer layer. Failure of this layer does not induce
total failure since the 90◦ -layer, although very thin, but oriented parallel
to the applied stress, can withstand the bending force. The micrograph
shown in Figure 5.14, taken at 1% strain, confirms that the crack initiated
at the tensile side is halted by this thin layer.
Due to the nature of the fabric, containing not only 0◦ but also a
small amount of 90◦ oriented fibres, the ultimate transverse strength of the
Figure 5.13: Tensile and flexural strength of pCBT, reinforced

with almost unidirectional fabrics, in longitudinal and transverse
direction
F
ply thickness ply orientation
0.65 90°
0.02 0°
0.65 90°
0.02 0°
0.02 0°
0.65 90°
Figure 5.14: Typical transverse flexural stress-strain behaviour

for S-UD composites (left) showing a knee which corresponds to
the failure of the outer layer (right) schematic representation and
micrograph of composite at 1% strain
laminate is actually an overestimation of the real unidirectional transverse

strength. Similar behaviour was observed for the A-UD composites when
tested in transverse direction. The ‘real’ transverse flexural strength is
determined as the stress at the knee point in the stress-strain curve and
is listed in Table 5.5. Compared to the flexural strength of PBT, which is
known to be in the range of 80-115 MPa [90], the real transverse strength
of the unidirectional composites is hence quite low.
Table 5.5: Real flexural transverse strength of unidirectional

composites, which is determined as the stress at the knee point
of the transverse flexural stress-strain curve
σf l,90◦ (MPa) σf l,90◦ ,real (MPa)
S-UD 66 ± 6 38 ± 8
A-UD 85 ± 16 36 ± 2
The transverse strength is a non-fibre dominated property and there-

fore influenced by both the matrix and interface properties. In glass fibre
composites, large local stress concentrations around the fibres exist be-
cause of the large stiffness mismatch between glass fibres and the polymer
matrix. This initiates cracks normal to the loading direction either at
the matrix-fibre interface or in the matrix [140]. These stress concen-
trations are responsible for lowering the transverse strength of thermoset
composites below the strength of the matrix.
Thermoplastic or toughened thermoset matrices, on the other hand,
should be able to compensate for these stress concentrations by local plas-
tic deformation of the matrix thus increasing the transverse strength to
above the matrix strength [141, 142]. A low transverse strength is there-
fore an indication of poor fibre-matrix interface properties, matrix brit-
tleness or a combination of both.
Bahr [83] investigated glass fibre reinforced pCBT prepared with com-
mercially available ‘epoxy-compatible’ sized fibres, which should have ad-
equate interface properties. Compared to pCBT composites made from
unsized glass fibers, the interlaminar shear strength improved substan-
tially (63 versus 24 MPa). Compared to glass fiber reinforced epoxy on
the other hand, this ILSS (63 MPa) is somewhat lower.
Since the reinforcements used in this study were epoxy-compatible,
the interface properties should also be adequate. Therefore, the low real
transverse strength is believed to indicate brittle matrix behaviour. The
mechanical properties of the matrix will be discussed in Chapter 6.
Influence of the fibre sizing Figure 5.15 compares the mechanical

properties of the S-UD composites with and without fibre sizing tested
in three point bending. There is a small increase in longitudinal stiffness
(7%), but a large drop in longitudinal strength (39%), transverse modulus
Figure 5.15: Influence of fibre sizing on the flexural properties

of S-UD composites (left) modulus (right) strength
(86 %) and transverse strength (74 %). The increase in longitudinal stiff-
ness is caused by a slightly different fibre lay-up. Whereas for the original
samples, the outer layer at the tensile side consisted out of 90◦ fibres, this
outer layer of the samples with the burnt fabric consisted out of 0◦ fibres.
The original lay-up could not be made since the stitching was removed
by burning. The decrease in longitudinal strength, however, is probably
caused by thermal degradation of the fibres due to the extended exposure
to high temperatures [143].
A more important effect of the removal of the sizing is the large reduc-
tion in transverse modulus and strength. When the composite is loaded in
transverse direction, the interface plays an important role in stress trans-
fer between fibres and matrix. Because the sizing contains components
specifically intended to improve this interface, the removal of the sizing
leads to very poor interface properties leading to a decrease in both the
transverse modulus and strength of more than 70%3 . From these results,
it is confirmed that developing a specific fibre sizing compatible with the
in-situ polymerisation process is crucial for the success of this production
process.
Semi-isothermal processing In Section 5.2.1, it was shown that the

degree of conversion and molecular weight of semi-isothermally manufac-
tured composites was lower than for isothermally produced composites.
Figure 5.16 shows the influence of this semi-isothermal process and hence
of the diminished matrix properties on the flexural modulus and strength.
As the molecular weight of the semi-isothermal specimens is still quite
high, the matrix modulus is most probably unaffected, the matrix strength
however might suffer from the rather high amount of unconverted oligo-
mers and lowered molecular weight. The decrease in longitudinal modu-
lus is quite unexpected but considering the large standard deviations, not
3 When the overall strength is considered, the real transverse strength of the S-UD
composite is lower, whereas that of the S-UD burnt composite remains 17 MPa
Figure 5.16: Influence of semi-isothermal process on the flexural

properties of A-UD composites, (left) modulus (right) strength
Figure 5.17: Relation between matrix characteristics and me-

chanical properties of unidirectional composites (left) transverse
flexural modulus, (right) ‘real’ transverse flexural strength (at
knee point of stress-strain curve)
really significant. The largest procentual decrease (20%) is seen in the

transverse strength.
5.3.3 Effect of interface and matrix properties on flex-

ural properties of unidirectional composites
As was indicated in the previous section the mechanical properties of the
composites change when the fibre sizing is removed and when the process
is semi-isothermal, leading to poor matrix properties. In this section,
the effect of interface and matrix properties is clarified. The longitudinal
modulus and strength are fibre dominated properties and show little or no
effect when changing the matrix or interface properties. The transverse
modulus and strength, on the other hand, are known to be matrix and
interface dominated.
Figure 5.17 depicts the effect of the matrix properties on the transverse
modulus and strength. Since two types of reinforcement (S-UD and A-
UD) were used, the experimental value of the modulus is normalised with
the theoretical value and the real experimental transverse strength, taken
at the knee point of the stress-strain curve, is normalised by the maximum

strength measured. The real transverse strength is not expected to change
much with the fabric used since the fibre volume fraction is comparable
for both types of composites.
The matrix properties are characterised by the product of degree of
conversion (polymer fraction) and molecular weight. The degree of crys-
tallinity is not taken into account since there are no significant differences
between the composites.
It is clear that the influence of the degree of conversion and molecular
weight on the transverse modulus is limited. Once the matrix molecular
weight and the degree of conversion reach a critical value, their effect on
the matrix modulus is limited. The effect of the interface properties, even
on the transverse modulus, can, however, not be neglected. Since the
modulus is measured at low strains, the interfacial strength should not
influence the modulus unless it is almost zero. The strong decrease in
modulus therefore suggests that most glass fibres are not bonded to the
matrix when the sizing is removed, and hence do not contribute to the
transverse composite stiffness. The measured modulus is even lower than
the matrix modulus since this matrix is filled with ‘holes’.
The matrix strength on the other hand is affected more by the degree
of conversion and molecular weight. Since the matrix strength dominates
the transverse strength, poor matrix properties can lower the transverse
strength significantly (> 20%). If, on top of that, the interface is weak-
ened, the transverse strength further decreases to less than 50% of its
original value. The effect of the interface is however not as pronounced
as for the transverse modulus. Since the fibres are not bonded to the
matrix, the matrix can deform more easily and no stress concentrations,
which lower the effective transverse composite strength, arise.
5.3.4 Cross-ply composites

Both the Twintexr and the biaxial pCBT composites (S-B) were tested
in three point bending. Two test sample orientations were considered,
namely (90,0)2,s and (±45)2,s . The experimental moduli are compared to
the theoretical predictions in Figure 5.18. It is obvious that for both test
orientations, theory and experiments do not coincide.
First, the (90,0)2,s test samples are discussed. Two reasons can be
stated for the discrepancy between experimental and theoretical flexural
modulus. First for the S-B composites, the fibre orientation is distorted
by the stitching pattern as can be seen on the lefthand-side of Figure 5.19,
whereas for the Twintexr composites, fibre crimp is not taken into ac-
count. Moreover, in three-point bending mode, both normal and shear
stresses are present throughout the beam span. Since the E/G ratio for
fibre reinforced composites is often rather large, the shear deflection can
Figure 5.18: Flexural modulus of cross-ply laminates (left)

outer layer 90◦ for S-B, (right) bias properties
45°
40°
Figure 5.19: Factors influencing mechanical performance (left)

deviation from fibre orientation for S-B, (right) effect of increasing
the span length for the woven Twintexr samples
be quite significant, Equation 5.11 [140].

" µ ¶ µ ¶2 #
F L3 3F L F L3 12 E h
d= 3
+ = 3
1+ (5.11)
4Ebh 10Gbh 4Ebh 10 G L
with d the midspan deflection, F the force, L the span length, E the
tensile modulus, G the shear modulus and b and h the width and thickness
respectively.
Even though the ASTM standard advises a span to thickness ratio of
16:1, increasing this ratio to 32:1 (span of 64 mm) has a significant effect
on the flexural modulus as illustrated on the righthand side of Figure 5.19.
The modulus increases with more than 20% by doubling the span length.
In order to avoid this last issue, some tensile tests were also performed.
The results of the (90,0)2,s -tensile tests are depicted in Figure 5.20,
clearly showing that the experimental tensile modulus compares well to
the predicted values. The strength and strain to failure differ substantially
from the non-crimp S-B to the woven Twintexr composites even though
the moduli are quite similar. Moreover, one would expect lower properties
for woven than for non-crimp composites.
Figure 5.20: Tensile properties of cross-ply laminates (left)

modulus, (right) strength and strain
Twintex
deviation from 0° orientation
S-B
Figure 5.21: Broken tensile samples indicating the difference in

failure mode and the fibre misalignment, (left) top view, (right)
side view
Figure 5.21 depicts both the top and side view of the failed tensile
samples. The lower tensile strength of the S-B composites is caused by
fibre misalignment which was clearly seen in the failed samples (top view).
The side view of the failed samples, reveals another difference in failure.
In the Twintexr samples, the damaged zone is rather small and mostly
consists out of broken fibres. The S-B composites, on the other hand,
have a larger affected delaminated area with less localised fibre breakage
showing that, due to the fibre misalignment, shear stresses became more
important resulting in a lower strength.
The flexural strength, on the other hand, is higher for the non-crimp
composites than for the woven composites (343 ± 60 MPa versus 283 ±
26 MPa). It must be noted that these test samples all failed in compression
mode. Failing in compression during a three point bending tests is typ-
ical for most thermoplastic composites. The longitudinal and transverse
compression properties are dominated by the matrix properties (longi-
tudinal matrix modulus, shear matrix modulus and matrix compressive
strength). Since these properties are generally lower for thermoplastic
matrices compared to thermoset matrices, thermoplastic composites of-
ten fail at the compression side of the flexural test specimens. The crimp
introduced in a woven fabric makes the composites more prone to fibre
Figure 5.22: Tensile properties in bias direction of cross-ply

laminate (left) top view showing 45◦ cracking, (right) typical
stress-strain curve
buckling compared to non-crimp fabrics, hence the higher flexural strength

(compression strength) is no surprise.
The (±45)2,s -specimens were also tested in three point bending and
the experimental modulus is compared to the theoretical predictions on
the right-hand side of Figure 5.18. The bias modulus of the Twintexr
composites compares well to the theoretical predictions, whereas theory
seems to underestimate the modulus of the S-B composites. As was al-
ready mentioned before, the stitches in the non-crimp fabric cause fibre
misalignment. Changing the fibre orientation from (±45)2,s to (±40)2,s
in the calculations, increases the theoretical flexural modulus to 9 GPa
compared to 7.4 GPa. Therefore, the underestimation of the bias modulus
of the S-B composites is believed to be related to fibre misorientations.
The S-B composites were also tested in tension in the bias direction.
This test is actually standardised to determine the in-plane shear proper-
ties of composites when the strains are measured in two directions (ASTM
D3518M). Even though the reliability of shear strength measurements
with this test is questionable since the final fracture also involves inter-
laminar shear fracture on top of intralaminar shear fracture, it can still
give an indication towards the matrix-dominated shear properties [144].
The measured modulus of 7.1 ± 0.5 GPa compares well to the calcu-
lated value of 7.4 GPa. The failure mode consists of cracks growing across
the width parallel to the fibre direction and delamination at the ±45 in-
terfaces, Figure 5.22. The strength (28 ± 1 MPa) is however quite low
indicating either a poor fibre-matrix interface or poor matrix properties
resulting in reduced shear properties. Looking at the matrix characteris-
tics of S-B composites characterised previously, Table 5.2, the molecular
weight is more in-line with that of the semi-isothermally produced A-UD
and the ECA-UD composites, both showing reduced matrix-dominated
mechanical properties.
5.3.5 Overview of tensile and flexural composite properties
Table 5.6: Overview of tensile and flexural composite modulus compared to the theoretical predictions, (s: semi-
isothermal process; b: naked glass fibres)
Tensile modulus (GPa)

Vf 0◦ 90◦ 45◦
lay-up
(%) exp. Cham. η exp. Cham. η exp. Cham. η
S-UD (0,90)3 54 42.2 ± 1.4 40.1 40.1 n.a. 9.9 9.2 n.a. n.a. n.a.
A-UD (0,R,90)s 52 36.4 ± 1.6 36.8 36.8 n.a. 10.0 9.2 n.a. n.a. n.a.
S-B∗ (45,-45)2,s 47 19.8 ± 0.7 20.7 20.3 n.a. 20.7 20.3 7.1 ± 0.5 7.4 7.4
Twintex twill weave 50 18.0 ± 0.7 20.8 20.4 n.a. 20.8 20.4 n.a. 7.5 7.4
Flexural modulus (GPa)

5.3 Tensile and flexural properties
Vf 0◦ 90◦ 45◦
lay-up
(%) exp. Cham. η exp. Cham. η exp. Cham. η
S-UD (0,90)3 54 38.3 ± 1.2 39.4 39.3 8.7 ± 0.9 10.0 9.3 n.a. n.a. n.a.
S-UD b (0,90)3 54 40.9 ± 2.2 39.4 39.3 1.25 ± 0.24 10.0 9.3 n.a. n.a. n.a.
A-UD (0,R,90)s 52 37.8 ± 1.4 38.8 38.8 6.8 ± 1.2 8.0 7.2 n.a. n.a. n.a.
A-UD s (0,R,90)s 52 35.0 ± 2.3 38.8 38.8 6.6 ± 0.7 8.0 7.2 n.a. n.a. n.a.
ECA-UD (0,±45) braid 48 26.5 ± 1.9 29.2 29.2 4.2 ± 0.9 7.9 7.1 n.a. n.a. n.a.
S-B∗ (45,-45)2,s 47 n.a. 25.9 25.6 9.9 ± 0.9 15.6 15.0 9.6 ± 0.7 7.4 7.3
Twintex twill weave 50 13.2 ± 0.7 26.0 25.7 13.2 ± 0.7 15.6 15.0 7.1 ± 0.7 7.5 7.4
∗ 0◦ : the fibres are oriented 0◦ in the outer layer, ∗ 90◦ : the fibres are oriented 90◦ in the outer layer, ∗ 45◦ : the fibres are oriented ±45◦
95
96
Table 5.7: Overview of tensile and flexural composite failure strength and strain, (s: semi-isothermal process; b: naked
glass fibres)
Tensile strength (MPa) Tensile strain (%)

lay-up Vf (%) 0◦ 90◦ 45◦ 0◦ 90◦ 45◦
S-UD (0,90)3 54 872 ± 86 n.a. n.a. 2.38 ± 0.17 n.a. n.a.
A-UD (0,R,90)s 52 767 ± 39 n.a. n.a. 2.32 ± 0.13 n.a. n.a.
S-B∗ (45,-45)2,s 47 229 ± 32 n.a. 28 ± 1 1.90 ± 0.38 n.a. 0.90 ± 0.19
Twintex twill weave 50 399 ± 7 n.a. n.a. 2.58 ± 0.11 n.a. n.a.
Flexural strength (MPa) Flexural strain (%)

◦
lay-up Vf (%) 0◦ 90◦∗∗ 45◦ 0 90◦ 45◦
S-UD (0,90)3 54 766 ± 113 66 ± 6 n.a. 2.20 ± 0.28 1.88 ± 0.26 n.a.
S-UD b (0,90)3 54 465 ± 59 17 ± 3 n.a. 1.39 ± 0.13 1.38 ± 0.25 n.a.
A-UD (0,R,90)s 52 924 ± 89 85 ± 16 n.a. 2.66 ± 0.19 n.a. n.a.
A-UD s (0,R,90)s 52 813 ± 40 68 ± 8 n.a. 2.65 ± 0.20 n.a. n.a.
ECA-UD (0,±45) braid 48 638 ± 81 55 ± 5 n.a. 2.81 ± 0.29 n.a. n.a.
S-B∗ (45,-45)2,s 47 n.a. 343 ± 60 252 ± 20 n.a. 4.21 ± 0.63 n.a.
Twintex twill weave 50 283 ± 26 n.a. 193 ± 13 2.78 ± 0.51 n.a. n.a.
∗0◦ : the fibres are oriented 0◦ in the outer layer, ∗ 90◦ : the fibres are oriented 90◦ in the outer layer, ∗ 45◦ : the fibres are oriented ±45◦
∗∗ultimate strength given in table, the ‘real’ transverse strength according to the failure of the outer layer is S-UD: 38 ± 8; A-UD: 36 ± 2;
A-UD semi: 28 ± 4
Composites: Production and Properties
5.4 Drop weight impact and interlaminar fracture toughness 97
Figure 5.23: R-curves for woven composites showing the inter-

laminar fracture toughness for reinforced epoxy, pCBT and PBT
5.4 Drop weight impact and interlaminar frac-

ture toughness
The interlaminar fracture toughness of a material measures the delam-
ination resistance of that material and is hence an important factor for
the impact resistance of that material. The tough nature of thermoplastic
matrices is known to considerably increase both the interlaminar fracture
toughness and the impact damage tolerance [145]. This section compares
the interlaminar fracture toughness of three composites, namely epoxy
and pCBT both reinforced with the W-R580 glass fabric and produced
by resin transfer moulding and glass fibre reinforced PBT that was com-
pression moulded from PBT-Twintexr .
These materials and the production of the composites were described
in Chapter 4. The number of reinforcement layers and fibre volume frac-
tion was constant for these composites. It should however be noted that
the Twintexr fabric is a twill weave, which is different from the plain
weave W-R580.
5.4.1 Interlaminar fracture toughness

The mode I interlaminar fracture toughness is determined as described in
Section 4.4.3. The resulting R-curves are depicted in Figure 5.23 and the
toughness data are given in Table 5.8. Even though pCBT and PBT are
chemically the same material and both have a thermoplastic nature, the
interlaminar fracture toughness of the pCBT composites compares better
to the thermoset epoxy composites.
The poor interlaminar fracture toughness of pCBT composites can be
caused by poor fibre-matrix interface properties as well as brittle matrix
behaviour. As was already mentioned in Section 5.3.2, literature data
Table 5.8: Interlaminar fracture toughness of woven composites

for reinforced epoxy, pCBT and PBT
GIc,ini (J/m2 )∗ GIc (J/m2 )
Epoxy 382 ± 71 1112 ± 82
pCBT 283 ± 90 1215 ± 55
PBT 434 ± 75 3841 ± 426
∗ for comparative reasons only, too thick crack initiation foil
show that the interface of pCBT composites made with standard sized
glass fibres is adequate. Therefore, a poor fibre-matrix interface is not
considered to cause the large discrepancy between the PBT and pCBT
composites.
There is however a significant difference in processing route between
the pCBT and PBT composites, that is most probably causing the drop in
interlaminar fracture toughness. The PBT composites were compression
moulded and therefore cooled down from the melt rather fast whereas
the pCBT composites were isothermally produced at 190◦ C allowing for
simultaneous polymerisation and crystallisation. This difference in pro-
cessing route results in a higher degree of crystallinity of pCBT composites
compared to PBT composites as was shown in Figure 5.6.
In PEEK composites, Goa et al. [146] found a decrease in interlaminar
fracture toughness with decreasing cooling rate and hence with increas-
ing crystallinity. Moreover, the transverse testing of the unidirectional
composites already pointed towards brittle matrix behaviour.
Therefore, the lower interlaminar fracture toughness of the pCBT com-
posites is believed to be caused by the processing route followed, leading
to different matrix properties and hence confirms the previous indications
of brittle matrix behaviour. The decreased interlaminar fracture tough-
ness of the pCBT composites will have a profound effect on the impact
properties.
5.4.2 Drop weight impact properties

Damage mechanisms due to an impact involve delamination, matrix crack-
ing and fibre breakage. The first two damage modes are primarily depen-
dent on the matrix properties and the impact resistance increases with
resin toughness [145]. The energy absorbed during impact is usually not
affected by the matrix toughness, the delamination area however decreases
with increasing toughness.
Drop weight impact tests were performed as described in Section 5.4.2
and the damaged area was determined by ultrasonic C-scanning. Fig-
ure 5.24 shows the C-scan image of an impacted epoxy sample and the
corresponding histogram. By determining a threshold value, the dam-
Figure 5.24: C-scan results after impact (left) C-scan image,

(right) corresponding histogram of greyvalues, dark being undam-
aged and clear being damaged material
aged area can be calculated. From the example shown here, it is clear
that undamaged material has a greyscale value of 0 (very large peak) and
no real damage peak can be observed. The tail in the histogram is hence
considered to be damaged material. The threshold value is determined for
each sample individually. Since the software chooses the grey-scale val-
ues depending on the maximum and minimum intensity of the reflected
beam, which can be different for each sample because of e.g. variations in
thickness or damage due to clamping, the threshold value differs between
samples. Moreover, the damaged area due to the clamping of the samples
should not be taken into account.
Visual observation of the impacted samples already revealed some dif-
ferences. Depending on the impact energy, the PBT composites showed
a significant dent at the impact site, which was not as pronounced in
the pCBT composites and not at all observed for the epoxy composites.
Furthermore, both the pCBT and PBT composites clearly exhibit com-
pressive failure by microbuckling of the fibre bundles at the compression
side. The better shear properties of thermoset composites prevent mi-
crobuckling explaining why this failure mode is not observed in the epoxy
samples.
Figure 5.25 shows the absorbed energy and damaged area as function
of the impact energy. The matrix type and corresponding interlaminar
fracture toughness seem to hardly affect the absorbed energy. The delam-
inated area however is clearly different. Kuboki et al. [147] have observed
the same phenomena when investigating the influence of the interlaminar
fracture toughness on the impact resistance. A higher interlaminar frac-
ture toughness results in the reduction of the damage size, not affecting
the total energy absorbed during impact.
The thermoset composite, which has the lowest interlaminar fracture
toughness, has the largest damaged area. This damaged area is signifi-
Figure 5.25: Results from drop-weight impact (left) absorbed

energy, (right) damaged area determined from C-scan image
cantly larger than for the PBT composite. The lower interlaminar frac-
ture toughness of the glass fibre reinforced pCBT compared to the PBT
composites, on the other hand, leads to an increase in delaminated area,
resulting in values in between both other composites. An overview of the
impact properties is given in the next section.
If the impact damage would only consist out of delaminations, the
absorbed energy is related to the interlaminar fracture toughness and the
damaged area. A first simplistic estimation of the absorbed energy as-
sumes that the damaged area consists of three delamination at each ply
interface. Hence the absorbed energy can be calculated as three times
the projected damaged area, Figure 5.26. It is obvious that the absorbed
energy calculated in this way, only accounts for 5 to 20% of the experi-
mental absorbed energy for the brittle matrix systems, and for 20 to 50%
for the ductile matrix system.
First, the crack growth during impact is more closely related to the
mode II interlaminar fracture toughness, which can be substantially larger
than GIc . Second, during the visual inspection, it was clear that in all
samples, although not to the same extent, fibre breakage is observed. This
failure mode of course also absorbs energy and is accompanied by matrix
fracture and interface debonding.
It is interesting to note that all these damage mechanisms (delamina-
tions, fibre and matrix fracture, interface deonding) seem to interact in
such a way that they result in the same amount of absorbed energy regard-
less of matrix toughness and interlaminar fracture toughness. This has
led some authors to state that the impact resistance only depends on the
matrix toughness and interlaminar fracture toughness when the impact
resistance is defined as absorbed energy per damaged area [147, 148].
Figure 5.26: Comparison of absorbed energy during impact

and energy determined by multiplying the interlaminar fracture
toughness by three times the projected damaged area
102
5.4.3 Overview of impact properties
Table 5.9: Overview of properties related to the impact tests for woven composites with an epoxy, pCBT and PBT
matrix
Eabs J
Eim (J) Eabs (J) K (-) Fmax (kN) area (mm2 ) area ( mm 2)
E 11.14 ± 0.06 6.07 ± 0.45 1.16 ± 0.00 5.0 ± 0.1 323 ± 25 0.019 ± 0.003
P 14.38 ± 0.19 8.64 ± 0.11 1.12 ± 0.01 5.6 ± 0.2 392 ± 36 0.022 ± 0.002
O 18.72 ± 0.40 12.54 ± 0.73 1.12 ± 0.01 6.1 ± 0.1 453 ± 28 0.028 ± 0.001
X 25.06 ± 0.27 20.56 ± 1.37 1.09 ± 0.01 6.5 ± 0.1 500 ± 14 0.041 ± 0.004
Y 30.09 ± 0.10 26.25 ± 0.52 1.09 ± 0.00 6.5 ± 0.2 549 ± 11 0.048 ± 0.001
10.66 ± 0.12 6.11 ± 0.26 1.15 ± 0.01 4.5 ± 0.0 243 ± 34 0.025 ± 0.003
p
14.03 ± 0.19 8.91 ± 0.32 1.14 ± 0.02 5.3 ± 0.1 345 ± 42 0.026 ± 0.003
C
18.78 ± 0.30 13.51 ± 0.69 1.10 ± 0.01 5.9 ± 0.1 366 ± 14 0.037 ± 0.003
B
25.08 ± 0.33 21.04 ± 0.63 1.08 ± 0.01 6.3 ± 0.3 466 ± 12 0.045 ± 0.002
T
29.25 ± 0.42 25.83 ± 0.99 1.06 ± 0.00 6.0 ± 0.0 533 ± 40 0.050 ± 0.004
11.13 ± 0.01 6.01 ± 0.06 1.18 ± 0.01 4.8 ± 0.1 223 ± 17 0.027 ± 0.002
P 14.56 ± 0.03 8.58 ± 0.22 1.14 ± 0.00 5.4 ± 0.1 276 ± 28 0.031 ± 0.002
B 19.04 ± 0.04 13.22 ± 0.52 1.12 ± 0.00 5.9 ± 0.3 300 ± 19 0.044 ± 0.001
T 25.29 ± 0.16 18.93 ± 0.65 1.10 ± 0.01 6.4 ± 0.1 362 ± 63 0.053 ± 0.008
29.75 ± 0.26 26.02 ± 1.11 1.08 ± 0.01 6.4 ± 0.2 424 ± 47 0.062 ± 0.006
Composites: Production and Properties
5.5 Conclusions 103
5.5 Conclusions
Thermoplastic RTM, which is characterised by the in-situ polymerisation
of the matrix, was used to produce glass fibre reinforced composites. Be-
cause of the reactive nature of the matrix system, the time window for
impregnation is rather small and the processing parameters need to be
closely controlled. It was shown that the stirring time and melt tempera-
ture have a significant effect on the mould filling time. In order to match
the needed and available time window for impregnation, each case needs
to be considered separately. Hence, mould filling simulations taking the
changing viscosity and the capillary effects into account are needed. These
simulations are however not the topic of this study. Emphasis was placed
on the characterisation of the matrix and composite properties.
The matrix properties were determined inside the composite. Even

though the degree of conversion was lowered by the presence of the re-
inforcement, it is still sufficient to obtain good matrix properties in the
isothermally produced composites. Moreover, since the glass fibre sizing
has a GPC response overlapping the oligomer response, the determined
degree of conversion must be seen as a lower boundary. Semi-isothermal
processing has a negative effect on the degree of conversion, lowering it
by 7% compared to the isothermal process.
The glass fibre sizing has also a profound effect on the molecular
weight. First, the molecular weight decreases in glass fibre reinforced
samples but this decrease is avoided when stripping off the sizing of the
fibres indicating that the sizing introduces extra end-groups hence low-
ering the molecular weight. Moreover, a side-reaction between epoxide
functional groups and the polymerising CBTr resin is responsible for
partially crosslinking the pCBT resulting in a very high molecular weight
substance.
The degree of crystallinity in the isothermally produced composites
is significantly higher than for PBT composites which are rapidly cooled
down from the melt. Therefore, the influence of the followed processing
route, resulting in different crystallisation conditions needs to be further
studied. The influence of the glass fibres on the degree of crystallinity
on the other hand is strongly dependent on the amount of fibres (and
sizing) present and the contact between matrix and fibres during poly-
merisation since both influence the amount of partially crosslinked pCBT
which cannot crystallise.
The mechanical properties of the composites are also studied and the
moduli are compared to theoretical predictions. For composites with poor
matrix properties (ECA-UD) the effect on the transverse modulus of uni-
directional samples is obvious. The longitudinal modulus is however un-
affected.
The real transverse strength of unidirectional samples is found to be

lower than the matrix strength. Even though this is very common for ther-
moset composites, plastic deformation of the thermoplastic matrix should
be able to account for the large strain concentrations around the fibres.
The low transverse strength and the poor tensile properties of a ±45◦
laminate could therefore be an indication of brittle matrix behaviour, al-
though also the sizing could play a role as it influences the fibre-matrix
interface, which also affects the transverse and shear properties of com-
posites.
By comparing the effect of the matrix and the interface, it is clear that
the matrix only has a substantial influence on the transverse strength, at
least for the range of properties investigated here. The effect of the inter-
face, on the other hand, is more pronounced as it lowers both the trans-
verse modulus and strength. This suggests that the interfacial strength is
almost zero and that the impregnation is poor.
For the success of these composites, it is therefore crucial to develop
an optimal sizing, which is not only compatible with the polymerisation
reaction but which might also tailor and optimise the fibre-matrix inter-
face by a suitable side-reaction hence taking advantage of the reactivity
during the production.
Impact tests and interlaminar fracture tests were performed to com-
pare the behaviour of woven pCBT composites to epoxy and compression
moulded PBT composites. Both the interlaminar fracture toughness and
impact behaviour (damaged area) of the pCBT composites compare bet-
ter to the epoxy composites and are significantly poorer than the prop-
erties of the PBT composites, suggesting that pCBT is more brittle than
PBT. Since the chemical structure of both materials should be the same,
the difference in processing route and more specifically the difference in
crystallisation conditions is believed to be responsible for lowering the
composite properties.
From the results presented in this chapter, it is clear that the process-
ing route has an influence on the matrix properties and, as a consequence,
the non-fibre dominated composite properties are affected. Therefore, the
effect of the processing route on the mechanical properties of the matrix
and on the matrix characteristics, more specifically the degree of crys-
tallinity and morphology, will be discussed in more detail in the next
chapter.
Chapter 6
Properties and
morphology of pCBT
The processing route in thermoplastic RTM not only differs from tradi-
tional thermoplastic processing because of the in-situ polymerisation but
also because of its isothermal nature. Since the processing temperature is
below the polymer melting point, it is feasible that crystallisation starts
before polymerisation is completed. As was discussed in Chapter 2, crys-
tallisation during polymerisation as well as the relatively high crystalli-
sation temperature can have an effect on the crystalline morphology and
mechanical properties of a semicrystalline thermoplastic. In this chapter,
the mechanical properties and the crystalline structure of unreinforced
pCBT will be investigated.
6.1 Crystallisation kinetics of pCBT

Figure 6.1 shows the degree of crystallinity as function of time for catal-
ysed CBT oligomers. Before crystallisation can commence, these oligo-
mers need to convert to polymer of sufficiently high molecular weight. Due
to the athermal nature of this polymerisation reaction (Section 2.4.1), the
exothermic peak during the isothermal DSC scan can be solely attributed
to crystallisation.
Crystallisation curves of PBT, crystallised from the melt, usually show
a decrease in crystallinity and an increase in crystallisation speed when
the degree of supercooling increases [113, 136]. If polymerisation and crys-
tallisation were separated, meaning that molecular weight build-up was
completed before the onset of crystallisation, the crystallisation kinetics
would not differ from melt-crystallised PBT, except for a time-shift equal
to the time needed for polymerisation.
The onset of crystallisation for temperatures below 200◦ C does not
105
106 Properties and morphology of pCBT
Figure 6.1: Crystallisation kinetics of catalysed CBTr resin. At

time = 0 min, oligomers and catalyst are mixed and polymerisa-
tion can commence
Figure 6.2: Schematic representation of dependence of melting

point on number average molecular weight according to Equa-
tion 6.1
6.1 Crystallisation kinetics of pCBT 107
differ significantly even though the molecular weight build-up is slower at

lower temperatures, Figure 2.12. The dependence of the melting point on
the molecular weight was determined by Flory and is given in Equation 6.1
and represented schematically in Figure 6.2.
1 1 2RMm
− i = (6.1)
Tm Tm ∆Hu Mn
i
with Tm the melting point for infinite molecular weight, ∆Hu the heat
of fusion per monomer unit, R the ideal gas constant, Mm the mono-
mer molecular weight and Mn the number average molecular weight of
the polymer. Therefore, at a lower isothermal processing temperature,
polymer of lower molecular weight can already be below its melting point
and sufficiently supercooled. Hence, the slower polymerisation rate can
be compensated by the lower molecular weight needed for the onset of
crystallisation.
Crystallisation kinetics of isothermal crystallisation from the melt show
a typical S-shaped curve, which is not shown here at low processing tem-
peratures. Since the melting point of a polymer is dependent on its
molecular weight, the degree of supercooling increases throughout the
isothermal process, Figure 6.2. Thermodynamically, the crystallisation
rate should hence increase with increasing molecular weight, but kineti-
cally it should slow down. Since both effects are competing and depend
on the crystallisation mode (simultaneous or separated), the decreasing
crystallisation rate seen at low processing temperatures cannot solely be
explained on the basis of molecular weight build-up.
Similar to the anionic ring-opening polymerisation of polyamides be-
low the polymer melting point, crystallisation is believed to slow down
further polymerisation by lowering the molecular mobility and by entrap-
ment of active polymerisation sites [17], thus also decreasing the crystalli-
sation rate. It is this effect which is believed to be responsible of altering
the shape of the crystallisation curves.
At the composites’ processing temperature of 190◦ C, a clear maxi-
mum in degree of crystallinity is observed whereas for a temperature of
170◦ C, the final degree of crystallinity is relatively low, most probably
since further molecular weight build-up is prevented by the crystallisa-
tion, resulting in a high amount of remaining oligomers and very low
molecular weight pCBT. This low molecular weight pCBT does not seem
to crystallise at 170◦ C.
Since the crystallisation curves presented here clearly differ from melt
crystallised PBT, it can be concluded that for temperatures below 200◦ C,
where polymerisation is relatively slow, crystallisation starts before poly-
merisation is complete. As mentioned in Section 2.5.1, this simultaneity
can affect the crystal structure, the final degree of crystallinity and hence
the mechanical properties.
Figure 6.3: Different processing routes followed for pCBT
6.2 Processing routes for pCBT
In order to assess the mechanical properties of pCBT depending on the

processing route, samples were prepared both with the thermoplastic
RTM set-up but without reinforcement (Section 4.2.3) and with the hot
press (Section 4.3) with different processing routes. These routes and the
abbreviations used throughout this chapter are depicted in Figure 6.3.
The processing route similar to the one during the production of the
composites is named pCBT190 and refers to isothermal processing at
190◦ C for 30 minutes, after which the samples are slowly cooled down
to below 50◦ C before demoulding. In order to separate polymerisation
and crystallisation, but still have crystallisation at a 190◦ C, pCBT was
first polymerised at 230◦ C, after which the mould was cooled to 190◦ C to
allow for primary crystallisation (pCBT230-190).
These samples were made with the TP-RTM set-up. With this set-
up is is however impossible to cool the samples quickly, therefore, some
samples were polymerised in the hot press and either cooled down slowly
inside the press (pCBT240-S) or quenched to room temperature by re-
moving them from the press after polymerisation (pCBT240-Q).
These processing routes of pCBT are compared to the results of com-
mercially available PBT plates (Section 4.1.1) to assess a possible in-
fluence of starting material and the ring-opening polymerisation route.
These plates were produced by injection moulding and hence cooled down
rapidly from the melt and will be referred to as PBT-Q.
6.3 Mechanical properties of pCBT 109
Figure 6.4: Typical stress-strain curves recorded in three point

bending mode for pCBT at different processing conditions and
PBT
Table 6.1: Comparison of flexural properties of pCBT after var-

ious processing conditions
E (GPa) σ ∗ (MPa) ε∗ (MPa)

pCBT190 3.2 ± 0.1 54 ± 5 1.7 ± 0.2
pCBT230-190 3.1 ± 0.2 73 ± 14 2.3 ± 0.6
pCBT240-Q 2.4 ± 0.3 75 ± 6 3.5 ± 0.2
PBT-Q 2.2 ± 0.1 67 ± 9 3.3 ± 0.1
Strength and strain at failure for pCBT190 and pCBT230-190 and
yield strength and strain for pCBT240-Q and PBT-Q
6.3 Mechanical properties of pCBT

6.3.1 Flexural tests
Typical stress-strain curves for pCBT are depicted in Figure 6.4 whereas
the flexural properties are shown in Table 6.1. It is clear that pCBT190
and pCBT230-190 behave differently from quenched pCBT (pCBT240-
Q) and PBT. Apart from the higher modulus, the first break in a brittle
manner whereas the latter do not break at all in this three point bending
test, but show a yield point.
The difference between simultaneous (at 190◦ C) and separated (at
230◦ C) polymerisation and crystallisation is smaller but still obvious. Al-
though both samples are brittle, breaking at small strain levels, the sam-
average and range

Epoxy 4.3 2.7 - 6.6
pCBT190 1.5 1.4 - 2.1
pCBT240-S 2.5 1.8 - 3.5
pCBT240-Q 4.3 3.7 - 6.1
PBT-Q 8.2 4.3 - 12.2
Figure 6.5: Charpy impact toughness comparing CBTr resin

processed according to different processing routes to epoxy and
PBT
ples prepared by separated polymerisation and crystallisation do exhibit

higher strain to failure and a higher strength.
As was expected from the mechanical properties of the composites,
pCBT processed as in thermoplastic RTM does not behave like a tough
thermoplastic but is a brittle matrix. Charpy impact tests were performed
to further assess this brittleness.
6.3.2 Charpy impact tests

As was described in Section 4.4.4, specimens for Charpy impact tests
need to be 3 mm thick, therefore, the specimens were not produced in the
RTM moulds but in the hot press. Epoxy (Section 4.1.1) were also used
as reference materials to compare the different pCBT samples to.
The Charpy impact toughness is shown in Figure 6.5. As is common
for impact tests, the spread on the results is rather large, therefore, the
minimum and maximum obtained value is also given. The difference be-
tween the samples is very obvious. As expected, the PBT-Q samples have
the highest toughness whereas the isothermally produced pCBT samples
(pCBT190) have the lowest toughness. The slight improvement in tough-
ness in the pCBT240-S sample is consistent with the higher strain to
failure in the flexural test.
The pCBT rapidly cooled down from the melt (pCBT240-Q) also has a
significantly higher toughness than the other pCBT samples, however, its
value is not as high as the PBT sample. Although both samples are cooled
down rapidly from the melt, there is still a major difference in processing
route. During injection moulding, the polymer is injected into a cold
mould leading to an amorphous outer layer and an increasing crystallinity
throughout the thickness of the sample. This amorphous layer is less
pronounced in the samples prepared in the hot press. Hobbs and Pratt
6.3 Mechanical properties of pCBT 111
[149] found that for PBT the thickness of this amorphous layer has a
larger influence on the toughness of PBT than it has on the toughness of
other semicrystalline thermoplastics. Nevertheless, a clear trend is seen
for the pCBT samples. Separation of polymerisation and crystallisation
as well as increasing the cooling rate increases the impact toughness.
In order to compare pCBT to thermoset matrices, epoxy samples were
also tested. Their impact toughness is even higher than for the isothermal
pCBT sample. It must however be noted that the epoxy impact tough-
ness might be artificially high due to a misalignment between the notch
and the striker. This misalignment caused the crack to deviate from its
original path, leading to a larger fracture surface and hence a possible
overestimation of the impact toughness. No data in literature were found
for the Charpy impact toughness of this specific epoxy resin system.
Al-Zubi et al. [92] and Miller et al. [79, 93] also observed brittle fracture
and low fracture toughness values for macrocyclic poly(butylene tereph-
thalate) polymerised from CBTr resin and a cyclic catalyst (c-pCBT)1 .
They however attributed the brittle behaviour to the macrocyclic nature
of pCBT, which influenced the crystalline structure and suggested that
breaking the macrocycles would be sufficient for an increase in ductility
by heat treatment of the c-pCBT. Our results however clearly indicate
that the processing route is the key factor since linear pCBT is equally
brittle to macrocyclic pCBT.
6.3.3 Tensile tests of reprocessed pCBT

From the results presented above, it is clear that pCBT is brittle when
processed under similar conditions as it would be in RTM. Moreover, if
pCBT is polymerised at a higher temperature and cooled down sufficiently
fast, it can behave as a ductile material. Although not an option for the
production of composites, since it would increase cycle times substantially,
reprocessing might be a viable option to regain ductile behaviour.
Both pCBT190 and PBT-Q were grinded and reprocessed by injection
moulding (giving RP-pCBT and RP-PBT). The resulting tensile bars were
tested and the broken specimens are shown in Figure 6.6. Even though
it is obvious that the RP-PBT shows more necking than the RP-pCBT,
some of the latter samples clearly show neck formation and thus ductile
behaviour.
Table 6.2 shows the tensile properties of the reprocessed specimens
as well as some properties of injection moulded CBTr resin (IM-pCBT),
polymerised during moulding from the one-part system, hence having a
slightly lower catalyst level (0.33 wt% versus 0.45 wt%) [150]. The mod-
ulus of RP-pCBT decreased in comparison with pCBT, Table 6.1, but is
now quite similar to both PBT and RP-PBT as is the strength.
1G
1c = 0.23-0.54 versus 4.2-5.6 for PBT
neck formation
Figure 6.6: Tensile samples of reprocessed (left) PBT and

(right) pCBT showing some samples with neck formation
Table 6.2: Comparison of tensile properties of injection moulded

pCBT
E (GPa) σ ∗ (MPa) ε∗ (MPa)

RP-pCBT 2.3 ± 0.2 46 ± 7 3-70
RP-PBT 2.5 ± 0.2 50 ± 7 16-225
IM-pCBT [150] 2.4 55 160
6.4 Physical properties of pCBT

There are a number of reasons why a thermoplastic polymer can be brit-
tle. Apart from defects and impurities, a low molecular weight, a high
degree of crystallinity, very large spherulites or a low density of (inter-
crystalline) tie-molecules can substantially reduce ductility. Therefore,
the properties of pCBT were characterised with special emphasis on the
crystalline structure.
6.4.1 Degree of conversion and molecular weight

One of the first prerequisites to get a tough material is a high oligomer
conversion and sufficient molecular weight build-up. The results shown in
Table 6.3 show that the remaining oligomer content in the pCBT samples
is equivalent to the equilibrium oligomer content in PBT, which is known
to be 1-3% [71].
For the original samples (not reprocessed), there are slight differences
in molecular weight, nevertheless the attained molecular weight exceeds
the critical molecular weight for entanglements, Mw,c of 50 kg/mole [79].
Hence, the attained molecular weight is sufficient for ductile behaviour.
6.4 Physical properties of pCBT 113
Table 6.3: Physical properties of unreinforced pCBT and PBT

determined by GPC and DSC measurements
α Mn Mw χc,dsc χc,waxd
(%) (kg/mole) (kg/mole) (%) (%)
pCBT190 98.2 ± 0.1 29.3 ± 0.2 61.4 ± 0.5 47 ± 2 52
pCBT230-190 98.7 ± 0.1 35.0 ± 0.4 73.3 ± 0.6 42 ± 2 49
pCBT240-Q n.a. n.a. n.a. 34 ± 1 n.a.
PBT-Q 98.8 ± 0.1 33.8 ± 0.4 69.3 ± 0.2 35 ± 1 31
RP-pCBT 97.9 ± 0.1 20.5 ± 1.9 40.6 ± 4.3 35 ± 1 n.a.
RP-PBT 98.7 ± 0.1 32.3 ± 0.2 66.3 ± 0.5 34 ± 1 n.a.
The GPC-measurements revealed a large drop in molecular weight for

the RP-pCBT compared to the pCBT, which is not as pronounced in
RP-PBT. Hydrolysis of PBT in the presence of water, Equation 4.1, is
known to decrease the molecular weight and although the samples were
dried before reprocessing, the rather high level of transesterification cata-
lyst still present in the pCBT samples might have a negative effect on this
degradation reaction and induce an extra transesterification reaction, sim-
ilar to the one which initiated the the production of the original cyclics,
Figure 2.8.
The resulting molecular weight of the RP-pCBT is lower than the
critical molecular weight for molecular entanglement mentioned above.
This limit was however deduced from previous tests of PBT with varying
molecular weights [79] and must therefore not be seen as absolute. It
is nevertheless clear that the molecular weight of the RP-pCBT is close
to the critical molecular weight, therefore, not all specimens show neck-
ing. When the molecular weight of pCBT after injection molding is large
enough, ductile behaviour comparable to that of PBT is observed.
6.4.2 Degree of crystallinity

The degree of crystallinity determined with DSC and WAXD is shown
in Table 6.3 and Figure 6.7. Despite the small discrepancy between the
two measurement techniques, both show a significantly higher degree of
crystallinity for pCBT190 and pCBT230-190 compared to samples cooled
down rapidly from the melt (pCBT240-Q to RP-PBT). This large dif-
ference in degree of crystallinity is believed to be responsible for the ele-
vated modulus of the slowly crystallised pCBT specimens (pCBT190 and
pCBT230-190) compared to the quenched samples, Table 6.1. The abso-
lute value for the degree of crystallinity of more than 40% is rather high
and can partially explain the brittle behaviour.
Reprocessing of the pCBT and PBT samples leads to a degree of crys-
tallinity comparable to that of the PBT-Q samples and is around 35%.
Figure 6.7: Degree of crystallinity of pCBT depending on pro-

cessing route, (left) DSC results (right) WAXD results
Figure 6.8: Typical DSC curves for unreinforced pCBT and

PBT, (left) original samples (right) reprocessed samples
This lower crystallinity is consistent with the lowered modulus and in-
creased strain to failure, Table 6.1 compared to Table 6.2. From literature,
it is however clear that not only the degree of crystallinity is affected by
the processing route but also the crystalline structure which might further
clarify the large difference in behaviour.
6.5 Morphology of pCBT

A closer look at the DSC traces and WAXD profiles already reveals more
information about the crystalline morphology and perfection. Figure 6.8
shows the DSC-traces of the original and reprocessed samples. Even
though the degree of crystallinity does not differ substantially between
simultaneous (pCBT190) and separated (pCBT230-190) polymerisation
and crystallisation, the DSC traces are quite different. At a DSC heating
rate of 10◦ C/min, the latter sample shows a clear indication for double
melting behaviour, as was also seen in Section 5.2.2. This points towards
further perfection of the crystalline structure upon reheating of the sam-
6.5 Morphology of pCBT 115
Figure 6.9: Typical WAXD scattering patterns for unreinforced

pCBT and PBT, (left) original samples (right) composite samples
ples, which is not observed in the DSC trace of pCBT190.

Compared to the reprocessed samples and the PBT-Q sample, the
melting peak of pCBT190 is more narrow and there is no small bump
indicating cold crystallisation prior to melting. The same differences were
observed in the composite samples comparing the samples produced with
thermoplastic RTM and by compression moulding of Twintexr , indicating
similar behaviour in the composite samples, Figure 5.6, Section 5.2.1.
6.5.1 Unit cell level

The morphology of polymers at the unit cell level can be investigated with
wide angle X-ray diffraction. The resulting scattering patterns are shown
in Figure 6.9. As expected, the scattering pattern corresponds to the α
polymorph for all samples since the samples were not stretched during
crystallisation.
There is however an unmistakable difference between the original pCBT
(excluding pCBT240-Q) and the PBT-Q samples. The pCBT diffraction
peaks are better defined, showing distinct peaks whereas some of these
peaks cannot even be resolved from the background and the amorphous
halo in the PBT sample. These narrow and well-defined peaks are charac-
teristic for large and perfect crystals (Appendix B). The patterns of the
reprocessed samples, which are not shown, compare well to the original
pattern of the PBT-Q samples.
The righthand-side of Figure 6.9 depicts the scattering pattern of the
composites. Since glass fibres give rise to an increased absorption of the X-
rays, the high fibre volume content (around 50%) impedes measurements
in transmission mode. Hence, the scattering patterns were recorded in
reflection mode. Even though the scattering patterns are less clear, it can
be seen that the presence of the glass fibres does not interfere with the
formation of larger and more perfect crystals.
The difference in crystallite size can be quantified by using the Scher-
Figure 6.10: Crystallite size calculated with the Scherrer equa-

tion for three diffraction peaks
rer equation, which relates the peak width to the crystallite size, Dhkl ,
Equation 6.2.
λ
Dhkl = (6.2)
βhkl cos θhkl
with λ the wavelength, θ half the Bragg angle and βhkl the integral breadth
of the hkl diffraction peak.
Figure 6.10 shows the crystallite size in three directions as determined
with the Scherrer equation. The crystallite size in the (001)-direction
corresponds to the lamellar thickness. Although a larger difference is ex-
pected from the DSC traces, there is only a small difference in crystallite
size between the separated and simultaneous polymerisation and crystalli-
sation. When comparing these results to the PBT-Q sample however, a
large drop in crystallite size is noticed. Therefore, next to the difference
in degree of crystallinity, the crystal size also differs with the processing
route followed. Separating polymerisation and crystallisation but allowing
for slow crystallisation however only has a minor effect on the crystallite
size and perfection.
WAXD measurements compare the samples at unit cell level and an
indication of the lamellar structure is given by determining the crystallite
size with the Scherrer equation. A more detailed and graphic view at the
lamellar level can be obtained by TEM images.
100 nm 200 nm 100 nm 200 nm
Figure 6.11: Lamellar structure as revealed by TEM of (left)

PBT-Q compared to (right) pCBT190
6.5.2 Lamellar structure as revealed by TEM

Figure 6.11 shows the TEM micrographs of both PBT-Q and isother-
mally processed pCBT. In contrast to the PBT-Q sample, the lamellae in
the pCBT190 are well defined, thicker and nicely oriented. Hence these
micrographs confirm the WAXD results.
The lamellar structure of pCBT polymerised at 240◦ C and subse-
quently crystallised by slowly cooling the sample inside the hot press
(pCBT240-S), is depicted in Figure 6.12.
Especially at spherulitic boundaries, the density of lamellar stacks is
known to decrease, hence a non-space filling structure is not uncommon.
However, the TEM-images do not reveal amorphous pools alternating the
lamellar stacks, moreover loss of contrast can be a result of lamellae which
are inclined towards other lamellae.
The lamellar structure of pCBT240-S is compared to the lamellar
structure of the isothermally produced sample pCBT190 in Figure 6.13.
The transition from the crystalline lamellae to the amorphous interlamel-
lar regions is more pronounced and sharp in the isothermally produced
samples. This sharp transition is a clear indication for the reduction of
tie-molecules and a decrease in transition zone between the amorphous
and the crystalline phase.
Miller [79] stated that the tie-molecule density is influenced by simulta-
neous polymerisation and crystallisation since only short amorphous seg-
ments are attached to the crystal growth front, reducing the probability of
a polymer chain of becoming a tie-molecule. Moreover, the transesterifi-
cation catalyst remaining in the pCBT might also influence the amount of
tie-molecules. Indeed, the catalyst molecules cannot be included into the
polymer crystal, but concentrate at the surface of the growing crystals.
Owing to the mechanical tensions arising from packing density differences
Figure 6.12: Lamellar structure of slowly cooled pCBT

(pCBT240-S), as revealed by TEM
100 nm 100 nm
Figure 6.13: Comparison of lamellar structure between isother-

mally processed pCBT (pCBT190) and pCBT slowly cooled down
from the melt (pCBT240-S)
at the crystal boundaries, such a local transesterification enhancement

would drastically decrease the amount of tie-molecules.
When polymerisation and crystallisation are separated, the catalyst
can also not be included into the crystalline structure since its shape and
size would distort the regular crystalline structure. As a result of the sep-
aration of both processes, these ‘activated’ chain ends are however more
homogeneously distributed in the entangled melt, reducing concentrated
transesterification. Nevertheless, transesterification might induce local
disentanglement as was suggested for PET by Liangbin et al. [151].
The reprocessed samples were also investigated by taking TEM mi-
crographs, Figure 6.14. Although a structure very similar to PBT-Q is
present, some parts do exhibit better-defined and wider lamellae. Since
100 nm
200 nm
100 nm
Figure 6.14: Lamellar structure of reprocessed pCBT as re-

vealed by TEM, indicating a non-homogeneous structure
shearing is not so predominant in the lab-scale injection moulding unit,

some of the original structure of pCBT is believed to be restored upon
recrystallisation. Di Lorenzo et al. [124] also hypothesised the role of
residual crystal memories in PBT, stating that small chain aggregates
remaining in the liquid phase can act as precursors for crystal growth,
leading to spherulite growth at exactly the same position and with the
same appearance of the original spherulites. Insufficient homogenisation
at the molecular level is hence the cause of partially restoring the original
crystalline structure.
The TEM micrographs revealed an interesting difference between the
isothermally produced samples, in which crystallisation commences before
the polymerisation is completed and the samples in which polymerisation
and crystallisation were separated. The isothermal process leads to a
drastic reduction of tie-molecules which was observed as a sharp boundary
between the crystalline and the amorphous phase. It is believed that this
reduction in tie-molecule density as well as the overall high degree of
crystallinity causes matrix brittleness.
6.5.3 Spherulitic superstructure

It is known that two distinct spherulitic superstructures exist in PBT,
namely usual and unusual spherulites, Section 2.5.2. Moreover, poly-
mer toughness is inversely proportional to the spherulite size, hence the
spherulitic superstructure was examined. Figure 6.15 shows optical micro-
graphs of both PBT-Q and isothermally produced pCBT (pCBT190). In
both cases, the spherulites are rather small, indicating that this is not the
limiting factor for the toughness of either sample. The unusual spherulitic
50 µm 50 µm
Figure 6.15: Optical micrographs showing the superstructure of

(left) PBT-Q and (right) pCBT190
a b
c A P d
Figure 6.16: SALLS scattering patterns of pCBT (a) pCBT190,

(b) pCBT240-190, (c) pCBT240-Q, (d) pCBT240-10◦ C/min
6.6 Matrix modification 121
superstructure, which is typically formed in fast crystallising PBT can be

clearly observed in PBT-Q (compare with Figure 2.17, page 44). The
pCBT190 samples on the other hand do not clearly show one type of
spherulitic superstructure, hence SALLS patterns were recorded to ob-
tain more insight.
In order to obtain SALLS patterns, thin samples need to be prepared
with a thickness of only a few times the spherulitic diameter. Figure 6.16
shows the scattering patterns for pCBT produced with different ther-
mal histories. The isothermally produced sample, pCBT190, consists out
of usual spherulites as can be seen in Figure 6.16a. When polymerisa-
tion and crystallisation are however separated but with the crystallisation
taking place at 190◦ C (pCBT240-190),which is the same crystallisation
temperature as in the isothermal process, the spherulitic superstructure
does not change, Figure 6.16b. This result is consistent with the ob-
servations of Stein and Misra [111] even though some authors have seen
unusual or mixed type spherulites resulting from this crystallisation tem-
perature [113, 114]. More importantly, the spherulitic superstructure does
not change upon separation of polymerisation and crystallisation.
Figure 6.16c depicts the scattering pattern of pCBT polymerised at
240◦ C and subsequently quenched by removing the sample from the oven,
pCBT240-Q. As was the case for PBT-Q, unusual spherulites were formed,
indicating that pCBT behaves similarly to PBT when processed under
the same conditions. Cooling down more slowly, Figure 6.16d, resulted
in an undefined scattering pattern which can either result from small
spherulites or a mixed type of spherulites. Optical microscopy images
could not confirm either option. Therefore, no conclusion can be made
concerning this thermal history.
6.6 Matrix modification

From the results presented above, it is clear that pCBT isothermally pro-
cessed at 190◦ C is brittle, breaking at very small strains. This brittle-
ness is caused by a high overall degree of crystallinity, a perfect crys-
talline structure and a reduction in tie-molecules. A non-isothermal pro-
cess where polymerisation and crystallisation are separated and where the
polymer is cooled down fast from the melt, results in a more favourable
crystalline structure from a toughness point of view. For the production
of composites with thermoplastic RTM however, non-isothermal process-
ing would lead to a drastic increase in cycle times and is therefore not
considered to be a viable option.
In order to obtain tough pCBT without altering the isothermal nature
of the process, a first attempt has been made to chemically modify the
matrix, trying to change the resulting crystalline structure. Brunelle et
al. [71] copolymerised CBT with CET. Adding 10% of CET led to a
HO OH
HO OH
OH
HO C O
O C
C O
CH2 CH2 CH2 O O O
OH
O O
O O O C OH
O C C
HO O O O O O OH
CH CH CH
HO C O O C OH
O O C O
CH2 CH2 CH2 HO O C C O OH
HO C O O O C OH
O O
O O O O O C O O C O O
HO OH
HO C O C O O O C
O C OH
O O O C O O
O O
HO C C OH
HO O O O OH
n C C
O C OH
O O
CH3 CH3 CH3 O O O OH
C O
O C
HO C O
OH
HO OH
HO OH
Figure 6.17: Chemical structure of (left) epoxy (EponTM 164)

and (right) hyperbranched polymer (Boltronr H20)
reduction in melt enthalpy of more than 30% indicating a reduction in

degree of crystallinity. Tripathy et al. [84] on the other hand, reactively
prepared blends of c-pCBT and poly(vinyl butyral) and copolymerised
CBT with ε-caprolactone (CL) in order to tailor the physical properties
of the resulting polymer. They found that adding 20% of caprolactone
eliminates the brittle nature of solvent cast c-pCBT.
Although toughening of e.g. thermosets is most often accomplished by
phase seperation, it is believed that simply introducing branches inside the
polymer backbone might decrease the degree of crystallinity and improve
the amount of tie-molecules. When trying to modify the physical prop-
erties of pCBT chemically for use in thermoplastic RTM, care must be
taken not to increase the viscosity too much. Hence, only small amounts
of branching or copolymerisation agents can be added. In a preliminary
survey, two materials were chosen in order to investigate their influence
on the crystallinity of pCBT.
Epoxies are well known to react with PBT as was already explained
in Section 5.2.1. The first material chosen was hence a multifunctional
epoxy, EponTM 164 from Resolution. As a second modifying agent, a hy-
perbranched polymer, Boltronr H20, was chosen. The chemical structure
of both materials is depicted in Figure 6.17. The dendritic structure of
hyperbranched polymers combines excellent reactivity with a relatively
low viscosity. Moreover, it is known to promote branching of a polymer
backbone without crosslinking [152]. Toughness enhancement by adding
5-10% of these polymers to epoxies and polypropylene has already been
demonstrated [153, 154].
Samples containing 0-1-5-10% of additive were produced simultane-
ously in the hot press by adding the additive to the CBT melt. After
the catalyst was added and the melt was poured into the mould, it was
kept at 190◦ C for 40 minutes. The plates were subsequently cooled down
slowly inside the mould. WAXD measurements were performed to assess
6.7 Conclusions 123
Figure 6.18: Effect of matrix modifications on crystallinity (left)

degree of crystallinity determined by WAXD, (right) lamellar
thickness determined by peak fitting of WAXD 001 peak
the degree of crystallinity and changes in the crystalline structure.

The diffraction patterns of the modified pCBT did not differ from
unmodified pCBT. Figure 6.18 shows the results of the analysis of the
scattering patterns. The degree of crystallinity decreases substantially by
adding more than 5% of either epoxy or hyperbranched polymer, with the
epoxy being somewhat more efficient. The lamellar thickness is however
not affected significantly.
Some plates containing 5% of hyperbranched polymer were produced
with the thermoplastic RTM set-up. The quality of these plates was how-
ever insufficient to perform mechanical tests. Breaking of these plates
upon demoulding however indicated that the brittleness was not yet im-
proved. A further and more detailed study concerning the chemical mod-
ification of pCBT should be performed to asses different possibilities to
improve the toughness without sacrificing the isothermal nature of the
thermoplastic RTM process.
6.7 Conclusions
The mechanical properties of the composites which were described in
Chapter 5 indicated that the pCBT matrix resulting from the isother-
mal processing route is brittle. This chapter therefore investigated the
mechanical and physical properties of the matrix material without rein-
forcement.
Flexural and Charpy impact tests confirmed the brittleness of isother-
mally produced pCBT. Separating the polymerisation and crystallisation
as well as increasing the cooling rate from the melt significantly increased
the strain to failure and impact toughness of pCBT. The samples produced
by rapid cooling of polymerised pCBT and those which were reprocessed,
originally isothermally produced pCBT, both exhibited increased ductil-
ity. Hence it is concluded that the processing route and not the starting
material (CBTr resin) is responsible for the brittle behaviour. The ma-
trix brittleness causes a decrease in the transverse strength, interlaminar
fracture toughness and impact resistance of the composites.
In order to identify the cause of the brittle behaviour, the physical
properties of pCBT were investigated and special attention was paid to
the crystalline structure. A first prerequisite for ductile behaviour is suf-
ficient conversion of the cyclic oligomers resulting in a molecular weight
that exceeds the critical molecular weight for entanglement. For all sam-
ples except the reprocessed ones, this condition is fulfilled. From the
reprocessing of pCBT, it became clear that the catalyst, still remaining
inside the sample during reprocessing, might have a negative effect on the
molecular weight of pCBT.
Slow crystallisation, either simultaneous or separated, leads to an over-
all degree of crystallinity of more than 40%, which partially explains the
brittleness and which is consistent with the higher flexural modulus of
these samples. A more detailed study of the crystalline structure revealed
major differences between rapidly cooled PBT and isothermally produced
pCBT. Not only is the degree of crystallinity substantially higher for the
pCBT, also the size and perfection of the resulting crystallites is increased.
Moreover, the spherulitic superstructure changes from usual to unusual
spherulites when increasing the crystallisation speed.
Comparing simultaneous and separated polymerisation on the other
hand, showed that although the degree of crystallinity, the crystallite size
and spherulitic superstructure were similar, simultaneous polymerisation
and crystallisation leads to a reduction in tie-molecules explaining the
difference in mechanical behaviour. It is believed that the reduction of
tie-molecules together with the high overall degree of crystallinity is the
cause of matrix brittleness.
In order to obtain a tough pCBT matrix, polymerisation and crys-
tallisation need to be separated and pCBT should be cooled down rapidly
from the melt. For thermoplastic RTM, this means that the process can
no longer be isothermal, leading to a substantial increase in cycle time
because of thermal mould cycling. Therefore, non-isothermal processing
is not considered a viable option for thermoplastic RTM.
A first attempt was made to change the crystalline morphology with-
out sacrificing the isothermal nature of the RTM process by adding a small
amount of epoxy and hyperbranched polymer during the polymerisation
reaction. Although adding 5% of these materials already significantly re-
duced the overall degree of crystallinity, brittleness was not yet avoided.
A more detailed study is needed to further investigate the possibilities of
chemically modifying the matrix hence tailoring its physical properties.
Chapter 7
Time-resolved X-ray
measurements
The processing route for thermoplastic RTM with CBTr resin differs
substantially from conventional thermoplastic processing. As was pointed
out in the previous chapters, the crystallisation route differs significantly,
leading to a different overall semi-crystalline structure and hence mechan-
ical properties. Therefore, the structure development during isothermal
processing was investigated more in depth by using time-resolved X-ray
measurements.
7.1 Synchrotron radiation

In order to perform time-resolved X-ray measurements during e.g. crys-
tallisation, conventional X-ray sources are not suitable because of their
low intensity. Synchrotron radiation however has some specific advantages
which allow to investigate the structure development. The properties of
synchrotron radiation which are of interest for scattering experiments can
be summarised as follows [155]:
• Continuous spectrum from the infrared into the hard X-ray region.
• Very high intensity of radiation and brightness of the source as

compared with intensities of conventional sources.
• High collimation of the beam.
• Highly polarised radiation.
• Well-defined time structure which is a copy of the pulse structure

of the electron beam.
125
126 Time-resolved X-ray measurements
Figure 7.1: Schematic overview of synchrotron facility at ESRF,

Grenoble, adapted from [157], (1) linear accelerator, (2) booster
synchrotron, 300 m circumference, (3) storage ring, 844 m cir-
cumference (4) experimental hall, (5) beamline
Synchrotron radiation is generated in ring accelerators [156]. Elec-

trons are injected into a very large evacuated loop and kept circulating at
relativistic velocities by energy pumped in from powerful radio-frequency
sources. To constrain the circulating particles to the chamber, external
magnets provide an inward acceleration. A consequence of this accelera-
tion is the emission of synchrotron radiation.
The desired wavelength is usually selected by crystalline monochro-

mators, often in connection with complicated focusing and reflecting ar-
rangements to concentrate more of the available intensity on the sample.
The measurements were performed at the Dutch-Belgian beamline,

DUBBLE BM26, at the European Synchrotron Radiation Facility, ESRF,
in Grenoble, France. Figure 7.1 shows a schematic overview of this facility.
7.2 Materials and Experimental set-up 127
7.2 Materials and Experimental set-up
7.2.1 Materials
Due to the difficulty of preparing small samples of the catalysed two-part

system, it was decided to use the one-part system of the CBTr resin
(CBTXB3). This system contains 0.3 mole% of Fascat4101 catalyst,
which is slightly less then used in the composite production. Uncatal-
ysed CBTr resin (CBTXB0) was also used to determine the amorphous
scattering pattern. As a reference material, Ultradur B4500, which is
commercially available PBT, was chosen.
In order to investigate the influence of glass fibres, the unidirectional
fabric from Saertex (S-UD), described in Section 4.1.2, was cut into small
pieces and added to the CBTr resin to an amount of 30 w%.
Regular household aluminium foil (thickness ±15 µm) was folded to
obtain single-walled containers for the samples with a maximum height of
5 mm, Figure 7.2b. The samples were then dried for at least one hour at
100◦ C.
7.2.2 Experimental set-up
The experimental set-up is shown in Figure 7.2. Temperature was con-

trolled by a home-made heating device, where air first passed a heating
element after which it heats up the sample. The wavelength at which the
measurements were carried out is 1.24 Å (or 10 keV). For the WAXD, a mi-
crostrip gas chamber detector with an angular resolution of 0.03◦ has been
used. Data were collected over the angular range 5.35◦ ≤ 2θ ≤ 60.65◦ with
2θ the scattering angle, calibrated using HDPE (high density polyethy-
lene).
A 2D multiwire gas-filled SAXS detector recorded the SAXS patterns
with a sample-detector distance of 2.5 m to cover the range of 0.002 ≤
s ≤ 0.045 (Å−1 ), with s (= 2sinθ
λ ) the modulus of the scattering vector.
The SAXS detector was calibrated by the reflections of silver behenate.
The raw data need to be processed before any result can be ob-
tained. Figure 7.3 gives an overview of the processing steps needed for
both WAXD and SAXS. First the result files are separated using XO-
TOKO/BSL, a program specially developed to process this type of data,
which runs under Linux. Further processing is done with both Microsoft
Excel and OTOKO, which is similar to XOTOKO but it can be run in a
DOS-environment.
ionisation chamber 2
oven, sample holder ionisation chamber 1
thermocouple incident X-ray beam
heating element
WAXD detector
air supply
sample
a
b 5 mm
Figure 7.2: Experimental set-up for time-resolved X-ray mea-

surement (a) oven set-up, (b) sample holder
7.3 WAXD
7.3.1 Data processing
The raw WAXD data were first normalised to the intensity of the primary
beam and corrected for the detector response measured by an ionisation
chamber down-stream from the sample (chamber 2 in Figure 7.2). Then
the parasitic scattering was removed by subtracting an empty cell mea-
surement.
Subsequently, the obtained data were corrected for equipment related
problems, namely faulty detector channels and faulty frames. This was
realised by simple linear interpolation between the correct detector chan-
nels or frames. The same method was applied to remove the diffraction
peaks from the aluminium container. Figure 7.4a and b show the result of
these corrections. Next a linear background was subtracted, Figure 7.4c.
In order to calculate the degree of crystallinity from these WAXD pat-
terns, the amorphous scattering needs to be separated from the crystalline
7.3 WAXD 129
Figure 7.3: Schematic overview of processing of raw synchrotron

data both for WAXD and SAXS
scattering. For the samples containing glass fibres, however, the liquid-
like scattering pattern of these fibres first needs to be subtracted from the
total scattering.
Glass fibre scattering

The glass fibres present in some of the samples give rise to a liquid-like
scattering, which can be easily distinguished from the scattering of the
amorphous polymer, Figure 7.5. The glass fibre scattering needs to be
subtracted from the samples’ scattering pattern. Since the exact amount
of fibres in the irradiated volume is unknown, the total weight fraction
of fibres in the samples was chosen as a scaling factor before subtracting
this glass fibre scattering, Figure 7.6.
Temperature dependence of amorphous halo

In order to determine the degree of crystallinity, the amorphous scattering
needs to be subtracted from the total scattering, Figure 7.4d. Therefore,
Figure 7.4: Different steps in the data processing of WAXD re-

sults (a) raw data, (b) data corrected for faulty detector channels,
frames and aluminium peaks, (c) background corrected, (d) amor-
phous halo subtracted
Figure 7.5: Scattering of glass fibres compared to scattering of

amorphous PBT
7.3 WAXD 131
Figure 7.6: Correction for glass fibre scattering, (left) scattering

pattern after correction for linear background, (right) scattering
pattern after correction for glass fibres
the amorphous halo needs to be fitted underneath the crystalline diffrac-

tion peaks. During an isothermal measurement, the shape of the halo is
constant and can be determined from the melt pattern. When the tem-
perature rises, however, the scattering shifts to lower angles and this shift
needs to be taken into account.
Completely amorphous PBT samples at low temperatures are very dif-
ficult to obtain since PBT crystallises very fast. Moreover, upon reheating,
cold crystallisation takes place therefore limiting the temperature range
in which the shape of the amorphous halo can be determined with PBT.
Uncatalysed CBTr resin on the other hand, crystallises very slow and
from comparing the shape of the PBT melt to a CBT melt, it is clear
that the shape of the amorphous halo is very similar. For this reason,
uncatalysed CBTr resin, CBTXB0, was used to determine the temper-
ature dependent shape of the amorphous halo. The sample was molten
and cooled down from 220 to 30◦ C at 10◦ C/min. Figure 7.7 shows the
amorphous scattering versus temperature.
The amorphous scattering was then fitted by two Gaussians at se-
lected temperatures, Equation 7.1. The parameters obtained are shown
in Table 7.1.
Ã µ ¶2 !
A1 (x − xc1 )
y= p exp −2
w1 π/2 w1
Ã µ ¶2 !
A2 (x − xc2 )
+ p exp −2 (7.1)
w2 π/2 w2
7.3.2 Determination of the degree of crystallinity

The degree of crystallinity as function of time was determined by scaling
the experimental amorphous halo or the amorphous halo given by Equa-
Figure 7.7: Amorphous halo determined by the melt pattern of

CBTXB0 as function of temperature
Table 7.1: Temperature dependent parameters of fitted amor-

phous halo, with T in ◦ C
A1 0.72 A2 0.2
w1 8.4 + 0.0015T w2 11.2 + 0.02T
xc1 16.5 − 0.009T xc2 35.1 − 0.01T
tion 7.1 frame by frame to fit underneath the crystalline diffraction peaks.
The amorphous halo can then be subtracted from the scattering pattern,
Figure 7.4d. The degree of crystallinity is determined with Equation 7.2.
Ic
χc,waxd = (7.2)
Ia + Ic
with Ia and Ic the integrated intensity of the amorphous and crystalline

scattering.
Alternatively, isosbestic points could be used to determine the degree
of crystallinity during isothermal measurements. Isosbestic points are
most often encountered when measuring absorption spectra for a set of
solutions where the sum of the concentrations of the two principal com-
ponents is kept constant. The points of intersection of these spectra are
called isosbestic. The same applies for diffraction patterns of two-phase
systems.
During isothermal measurements, the peak positions remain constant,
giving rise to isosbestic points as illustrated in Figure 7.8.
7.3 WAXD 133
Figure 7.8: WAXD-patterns during isothermal crystallisation

and polymerisation showing isosbestic points, same data but dif-
ferent representation as Figure 7.4c
Since the intensity of the isosbestic points is independent from the de-
gree of crystallinity, the diffraction pattern of a 100% crystalline sample
can be obtained by scaling the crystalline scattering to pass through the
known isosbestic points. The degree of crystallinity can then be deter-
mined at points in between these isosbestic points by
Im,2θ − Ia,100%,2θ
χc,waxd = (7.3)
Ic,100%,2θ − Ia,100%,2θ
with Im,2θ , the measured intensity at an angle 2θ, Ia,100%,2θ and

Ic,100%,2θ , the intensity at that angle 2θ of respectively a 100% amorphous
and a 100% crystalline sample.
For the isothermal measurements, this method yields results almost
identical to the ones obtained by Equation 7.2. Although it is often quite
difficult to obtain an amorphous halo that covers the entire angular range,
the similarity of the two methods validates the use our experimentally
determined amorphous halo.
For non-isothermal measurements, the diffraction peaks shift to lower
angles with increasing temperature, therefore excluding the use of isos-
bestic points. It was hence decided not to use these isosbestic points but
to use the same procedure (Equation 7.2) for all the measurements.
Figure 7.9: Degree of crystallinity of CBTr resin as function

of temperature showing the crystallisation kinetics for simulta-
neous polymerisation and crystallisation at 190◦ C followed by a
cooling/heating cycle
7.3.3 Crystallisation kinetics

Figure 7.9 shows the degree of crystallinity as function of time for the
CBTr resin. The oligomers were heated rapidly to a temperature of
190◦ C and were kept at that temperature for 40 minutes. This holding
time was chosen to match the actual production cycle in thermoplastic
RTM, adding 10 minutes to allow the sample to melt. Subsequently, the
sample (now polymerised and crystallised) was cooled down to 100◦ C after
which it was heated again to above the equilibrium melting point of PBT,
which equals 245◦ C [158]. During cooling, the crystallinity still increases,
whereas upon reheating, the crystallinity decreases again. This indicates
the formation of imperfect, small crystals during cooling which melt at
temperatures below the main melting point.
From Figure 7.9 it is also clear that the time for crystallisation to start,
the induction time, is quite large. Evidently, this rather large induction
time is related to the polymerisation process. Before crystallisation can
start, the molecular weight build-up needs to be sufficiently progressed. As
discussed previously, polymerisation is not completed when crystallisation
commences. This becomes clear when comparing samples all crystallised
at 190◦ C but with different thermal histories, Figure 7.10.
As was already mentioned in Chapter 6, only a time shift would be
visible in the crystallisation curves if polymerisation would have been
completed before crystallisation starts. Instead, not only the induction
time but also the crystallisation rate differs. For the remolten pCBT as
7.3 WAXD 135
Figure 7.10: Comparison of crystallisation kinetics at 190◦ C de-

pending on thermal history
well as for the CBTr resin polymerised at 250◦ C, around 250 s are suf-
ficient to complete the major part of the crystallisation process, whereas
almost 1000 s (after the onset of crystallisation) are needed when the
sample needs to be polymerised at 190◦ C. Consequently, it is clear that
polymerisation and crystallisation proceed simultaneously at 190◦ C.
As was pointed out in Chapter 6, also the final degree of crystallinity
is larger when polymerisation and crystallisation proceed simultaneously.
The higher degree of crystallinity is attributed to the increased time frame
for crystallisation. The same effect is observed when increasing the crys-
tallisation temperature during crystallisation of PBT polymer without
molecular weight change [159].
Crystallites or local chain organisation can still remain at temperatures
above the melting point leading to nuclei that can promote recrystallisa-
tion. The samples were therefore heated to above their equilibrium melt-
ing point to remove all prior thermal history. Therefore, no significant
difference is observed between the sample crystallised from the molten
pCBT and the sample crystallised from the pCBT without prior crystalli-
sation history.
No comparison with the PBT samples could be made since the WAXD
detector failed. Park et al. [128] and Hsiao et al. [159] however investigated
the crystallisation kinetics of PBT and found induction times and overall
crystallisation times comparable to those of the pCBT samples crystallised
at 190◦ C from the melt.
Within this study, the production and characterisation of composites

Figure 7.11: Influence of glass fibres on crystallisation kinetics

at 190◦ C, (left) during simultaneous polymerisation and crystalli-
sation and (right) after polymerisation at 250◦ C
is the main issue. Therefore, the effect of adding glass fibres to the CBTr
resin was of course investigated. Figure 7.11 compares the crystallisation
kinetics for CBTr resin, with and without glass fibres, crystallised at
190◦ C. The rather large scatter in the glass fibre samples is due to the
decreased intensity of the measurements in the presence of glass fibres.
The glass fibre surface is known to act as a nucleating site, hence
favouring heterogeneous nucleation. This affects not only the induction
time for crystallisation which decreases with increasing glass fibre content
[128], but also the crystallisation rate, which increases with increasing
glass fibre content [127, 128].
On the righthand-side of Figure 7.11, no decrease in induction time is
observed. At 190◦ C, the effect on the induction time observed by Park
et al. [128] was also very minor and only became more pronounced at
higher crystallisation temperatures. Therefore, it appears that pCBT,
crystallised from the melt, behaves similar to PBT. For the sample with
simultaneous polymerisation and crystallisation on the other hand, an
increase in induction time is observed which conflicts with previous results,
Figure 7.11, left.
The onset of crystallisation is most probably delayed due to a reduction
in polymerisation rate. The glass fibre sizing contains epoxide functional
groups which cannot only react with PBT as explained in Chapter 5 but
also with the catalyst, which equally contains a OH-group. Therefore, the
catalyst, present near the glass fibre sizing, can be disfunctionalised by a
reaction with this sizing. Moreover, it promotes the crosslinking reaction
of the CBTr resin preventing the crystallisation of pCBT. Hence it is
concluded that the effective amount of catalyst is reduced by the glass
fibres sizing, decreasing the polymerisation rate and hence increasing the
induction time for crystallisation.
The effect of glass fibres as nucleating sites is nevertheless observed in
the crystallisation rate which increases upon adding glass fibres in both
cases. The overall degree of crystallinity also increases in the composite
7.3 WAXD 137
Figure 7.12: Influence of crystallisation temperature on crys-

tallisation kinetics after polymerisation at 250◦ C, (left) without
glass fibres, (right) with glass fibres
samples. This effect was not observed in the RTM samples, where the
opposite trend was seen. For the samples prepared in the hot press, on the
other hand, a similar increase was seen. As was explained in Section 5.2.2,
the amount of glass fibres in the RTM samples is much higher, 65 w%
as compared to 30 w%. Moreover, less flow is induced in the current
samples, decreasing the contact area between fibres and matrix. Hence
the production of composites in the hot press compares better to the
conditions in the current samples, leading to similar results.
Park et al. [128] and Hsiao et al. [159] also investigated the effect of
crystallisation temperature on the crystallisation kinetics. They found
that by decreasing the crystallisation temperature (or increasing the de-
gree of supercooling), the induction time and overall degree of crystallinity
decrease and the crystallisation rate increases.
Figure 7.12 shows the influence of the crystallisation temperature on
the crystallisation kinetics both for unreinforced and reinforced samples.
The induction time indeed decreases upon decreasing the crystallisation
temperature. The effect is however small in the presence of glass fibres.
The same holds for the crystallisation rate. Crystallisation temperatures
below 170◦ C could not be investigated due to the limited cooling rate
of the experimental set-up. Even at 170◦ C, crystallisation already com-
menced before the set temperature was reached, Figure 7.12.
Simultaneous polymerisation and crystallisation was also investigated
at 190 and 170◦ C. After cooling and melting of the so obtained speci-
mens, they were recrystallised at the same crystallisation temperatures.
Figure 7.13 depicts the resulting crystallisation curves. The difference
between the two temperatures is very pronounced.
At 170◦ C, the decreased polymerisation rate is translated in a some-
what longer induction time and a decreased crystallisation rate. Moreover,
the overall degree of crystallinity is substantially lower. It is believed that
the onset of crystallisation not only slows down the polymerisation but
even prevents further polymerisation and molecular weight build-up due
Figure 7.13: Influence of polymerisation temperature on crys-

tallisation kinetics during isothermal and non-isothermal process-
ing, (left) 190◦ C, (right) 170◦ C
to entrapment of active polymerisation sites. This low molecular weight

polymer and possible remaining oligomers are responsible for lowering the
melting point to around 200◦ C.
Before recrystallising the sample, it was kept at 250◦ C for 5 minutes,
this time seemed to be sufficient to complete the molecular weight build-
up. Results similar to the ones presented in Figure 7.12 are obtained and
the melting point reaches higher values.
7.4 SAXS
7.4.1 Data processing
The SAXS data were recorded with a 2D detector. Figure 7.14 shows the
patterns for both the melt and a semi-crystalline sample. One-dimensional
scattering patterns were obtained by azimuthal integration of the 2D pat-
terns. These 1D data were then further corrected before applying the
correlation function analysis, Figure 7.3.
Similar as for the WAXD data, the 1D SAXS patterns were normalised
to the intensity of the primary beam and corrected for the detector re-
sponse. A melt pattern was then subtracted to eliminate background scat-
tering. In order to perform the correlation function analysis and hence
determine the characteristic lengths of the two-phase structure, data need
to be extrapolated to both low and high angles, Section B.4.1.
Extrapolation to infinity is usually based on Porod’s law which is ap-
plicable to systems with sharp phase boundaries. Equation 7.4 assumes
a constant background, B, and a sigmoidal transition between the crys-
talline and amorphous electron densities, with σ characterising the tran-
sition layer thickness. P is the Porod constant.
P
lim I(s) = B + exp(−4π 2 σ 2 s2 ) (7.4)
s→∞ s4
7.4 SAXS 139
Figure 7.14: 2D SAXS patterns of CBTr resin at 190◦ C, (left)

prior to polymerisation and crystallisation, (right) after polymeri-
sation and crystallisation
Since the data were recorded over a relatively large s-range, the Porod
term in the tail was considered small compared to the constant background
and hence no extrapolation was performed, only the constant background
was subtracted.
Extrapolation to low angles (s → 0) was however necessary and was
performed by taking I(s) constant for s < 0.0028 Å−1 . After applying
the Lorentz correction (multiplying I(s) with s2 ), this results in a second
order polynomial passing through the origin and the data point at s =
0.0028 Å−1 . Figure 7.15 shows the Lorentz corrected pattern for the
isothermal polymerisation and crystallisation of CBTr resin.
The resulting patterns were then smoothed after which the correlation
function was calculated. From this correlation function, the structural
parameters of the two-phase structure were determined. The correlation
function analysis is described in more detail in Appendix B, Section B.4.1.
Figure 7.16 shows the correlation functions during isothermal poly-
merisation and crystallisation at 190◦ C. These functions do not exhibit a
flat minimum indicating overlapping distributions of the two characteristic
lengths. Consequently, using this minimum to calculate the local minority
volume fraction (either crystalline or amorphous) would yield an underes-
timation of the real volume fraction, leaving the quadratic expression as
only option.
Effect of finite lateral dimensions
In order to calculate the crystalline and amorphous volume fractions

within the semi-crystalline stacks, the quadratic expression is used, Equa-
tion 7.5.
2
LRATy=o = A = φ(1 − φ) = φ(1 − φ)Lp (7.5)
OS
Figure 7.15: Lorentz corrected SAXS pattern during isothermal

polymerisation and crystallisation at 190◦ C
Figure 7.16: Calculated correlation functions for the isothermal

polymerisation and crystallisation of CBTr resin at 190◦ C
7.4 SAXS 141
Figure 7.17: Real degree of crystallinity versus calculated de-

gree of crystallinity assuming infinite lateral dimensions of the
crystal for different lateral dimensions, D and Lp = 150 Å
with A the intersection of the linear regression of the autocorrelation tri-

angle with the abscissa, φ and (1 − φ) the crystalline and amorphous
volume fractions within the semi-crystalline stacks, Os the specific sur-
face area of the phase boundary and Lp the long period, which can be
determined from the subsidiary maximum of the correlation function.
Within this equation, O2S was replaced by Lp , assuming infinite lateral
dimensions of the lamellar crystals. This condition is, however, never
fulfilled and its effect is most noticeable around φ = 0.5. Goderis et
al. [160], calculated the specific surface area of the phase boundary, Os ,
as function of the lateral dimension of the crystal, D, which was taken to
be equal in both directions:
4lc + 2D 4φLp + 2D
Os = = (7.6)
Lp D Lp D
For a given degree of crystallinity, φreal , long period, Lp , and lateral
dimension, D, Os was calculated using Equation 7.6. Next, A-values
were extracted from Equation 7.5. Using these A-values, the degree of
crystallinity, φcalc , is determined assuming infinite lateral dimensions.
Figure 7.17 shows the difference in calculated and real degree of crys-
tallinity for different values of the lateral crystal dimensions. It is clear
that for very small crystals, deviations occur even at low degree of crys-
tallinity. For crystals of substantial lateral dimensions, however, devia-
tions always occur when φ is near to 0.5. Generally spoken, φ is often
under- or overestimated between 0.35 and 0.65.
7.4.2 Assignment of characteristic lengths

The correlation function analysis and the interface distribution function
yield two characteristic lengths, l1 < l2 , but neither can assign these calcu-
lated characteristic lengths to the crystalline or amorphous phase1 . This
ambiguity has led to two completely different views on the morphology
of thermoplastic polyesters such as PET and PBT, especially when the
overall degree of crystallinity is lower than 50%.
Short survey on length assignment

One school of researchers [161, 162] assigns the lower value, l1 , to the
lamellar thickness. The final semi-crystalline morphology is then com-
posed of homogeneously distributed space-filling lamellar stacks. Sec-
ondary crystallisation may hence occur by the insertion of lamellae in the
interlamellar amorphous region. This assignment is most often based on
the comparison of the product of the overall degree of crystallinity (deter-
mined by an independent method) and the long period, which often gives
values close to l1 .
Haubruge et al. [161] compared values obtained by both the correla-
tion function and the interface distribution function to those obtained by
analysis of TEM-images. They conclude that the lower value of the cor-
relation function analysis is closer to the lamellar thickness determined
in real space. It is, however, clear from their results that the correlation
function analysis underestimates the linear degree of crystallinity (30%
versus 40%) most probably due to limited lamellar dimensions as dis-
cussed above. As a consequence, both l1 and l2 differ significantly from
the TEM-value with l1 giving the best estimate. In contrast, the long pe-
riod obtained from TEM and the correlation function analysis correlate
well.
The characteristic lengths deduced from the interface distribution func-
tion on the other hand are almost equal and very close to the lamellar
thickness determined by TEM. The long period as determined by the in-
terface distribution function is, however, significantly smaller than the
values obtained by TEM-analysis.
TEM-images and their analysis need to be treated with care and the
obtained characteristic values should not be considered indisputable. A
staining procedure is necessary to have sufficient contrast between the
amorphous and crystalline zones and it is unclear how, for example, an
interface layer is stained. The preferential staining or not staining of
the interface, however, does not influence the determination of the long
period.
In contrast, if the staining procedure causes the amorphous phase to
swell, both the long period and amorphous characteristic length are over-
1 Babinet’s principle, Section B.4.1 and B.4.2
7.4 SAXS 143
estimated by the TEM image analysis, leaving the crystalline character-

istic length unaffected. Moreover, TEM-images are very local and can
therefore give rise to very different conclusions depending on the investi-
gated area whereas SAXS data correspond to a larger sample volume.
From the analysis presented by Haubruge et al. [161], it is, however,

clear that the trends as function of crystallisation temperature are very
similar and independent of the assignment of the characteristic lengths
and the analysis method.
The other school of researchers [159, 163–166] makes the opposite as-
signment, namely the lamellar thickness is the larger of both values, l2 .
The local crystalline fraction is therefore higher than the overall degree of
crystallinity leading to non-space-filling lamellar stacks. These stacks are
hence believed to alternate large amorphous pools. Secondary crystalli-
sation can consequently not only occur by lamellar insertion but also by
a stack insertion process.
Several papers aim to prove this assignment. Wang et al. [164] tried
to determine the lamellar thickness by calculating D001 from the WAXD
diffraction pattern. Since the 001 diffraction peak was not measured, they
estimated D001 from the width of the 01̄1 peak. The simple correlation
between D001 and D01̄1 applied in this publication is, however, erroneous
since it is not based on a triclinic unit cell.
Xia et al. [166] on the other hand performed a similar study to that
of Haubruge et al. [161] comparing TEM-images to the SAXS data. They
found results contradicting those from Haubruge et al., namely, the TEM
lamellar thickness corresponds more closely to the larger value determined
by the correlation function analysis.
Samples with a total volume fraction of crystals around 40%, as de-

termined by density measurements, were investigated by Santa Cruz et
al. [163]. Since the linear degree of crystallinity cannot be smaller than
the overall crystallinity, the larger value was again assigned to the lamellar
thickness.
The major weakness in this second interpretation is the assumption

of the existence of large amorphous pools alternating the lamellar stacks.
None of the researchers above have microscopic proof of the existence of
these regions. Hsiao et al. [159] calculated the volume fraction of the
lamellar stacks in the bulk sample which varied between 20% and 40%
leading to more than 60% of fully amorphous bulk material. This very
large amount of inter- or intraspherulitic amorphous material should be
visible by microscopical techniques, which is, to our experience, not the
case.
Table 7.2: Comparison of the lamellar thickness determined by

WAXD, D001 and SAXS, l1 and l2 , in Å
D001 l1 l2
pCBT 100 40 73
PBT 56 26 65
Determination of the lamellar thickness by WAXD

A rather independent method to assign one value obtained by the SAXS-
analysis to the lamellar thickness, is to calculate D001 from the width of
the 001 WAXD diffraction peak. This peak is, however, situated at rather
low diffraction angles and is often resolved in the background. Unfortu-
nately, this is also the case for our time-resolved measurements. Because
of its low intensity, this peak has, to our knowledge, not been used to
verify the lamellar thickness.
However, the room temperature measurements discussed in Chapter 6
did clearly show this peak and a number of authors have also observed the
001 diffraction peak [85, 123, 149]. Therefore, SAXS measurements were
performed and the characteristic lengths determined with the correlation
function analysis were compared to D001 determined by the Scherrer equa-
tion (Equation 6.2).
Table 7.2 shows the results of this comparison. It is clear that the
larger value from the correlation function analysis, l2 , corresponds best to
the lamellar thickness determined by WAXD. Moreover, a study on block-
copolymers of PBT [167] indicated that for a melting point of around
210◦ C, which is below the melting points observed in the current sam-
ples, the number of PBT monomeric units in an extended chain crystal
is around nine, which corresponds to a lamellar thickness2 of more than
100 Å, favouring the assignment of the largest value to the lamellar thick-
ness.
At this point, no conclusion is made about the consequences for the
overall morphology. These will be discussed later in Section 7.5.
7.4.3 Characteristic lengths and local degree of cry-

tallinity
Figure 7.18 depicts the evolution of the local volume fractions as function
of temperature and time for simultaneous polymerisation and crystalli-
sation at 190◦ C followed by a cooling/heating cycle. The corresponding
characteristic lengths are shown in Figure 7.19. As was shown above, l2
is considered to correspond to the lamellar thickness. Consequently, the
2 each unit cell of the α form crystals (c = 11.65Å, [109]) contains one monomeric
unit [103]
7.4 SAXS 145
Figure 7.18: Local volume fractions corresponding to the ideal

two-phase structure of CBTr resin during isothermal polymeri-
sation and crystallisation at 190◦ C followed by a cooling/heating
cycle
local crystalline volume fraction is around 70%. The lamellar thickness

decreases in the course of the crystallisation process. During the isother-
mal process, the degree of supercooling increases due to the molecular
weight build-up. Accordingly, at least part of the decrease observed here
might be induced by the polymerisation process and the changing degree
of supercooling [159, 163]. However, when polymerisation and crystallisa-
tion are separated, the lamellar thickness also decreases during the course
of crystallisation pointing at an additional ‘thinning’ mechanism (see be-
low). The decrease in long period is governed by the decrease in lamellar
thickness since the thickness of the interlamellar amorphous layer has only
a very minor decrease. This leads to a small decrease in local crystalline
volume fraction.
During the subsequent cooling, the long period and lamellar thickness
further decrease. Part of the dimensional decrease is caused by shrinkage
upon cooling. Shrinkage alone, however, cannot account for the decrease
since the amorphous thickness should then decrease more steeply than
the lamellar thickness as the thermal expansion coefficient is three times
larger [168]. The overall degree of crystallinity as determined by WAXD
increases during cooling (Figure 7.9), whereas the local SAXS crystallinity
slightly decreases (Figure 7.18).
Several models are proposed in literature to explain this secondary
crystallisation, the lamellar and stack insertion model and the stack thick-
ening model. Since the interlamellar amorphous layer only has minor
decrease, the stack insertion model and the stack thickening model are
Figure 7.19: Characteristic lengths corresponding to the ideal

two-phase structure of CBTr resin during isothermal polymeri-
sation and crystallisation at 190◦ C followed by a cooling/heating
cycle
Figure 7.20: Schematic illustration of space-filling spherulitic

superstructure resulting in non-space-filling lamellar stacks
favoured assuming that the stacks are not completely space-filling.

From the optical microscopy images shown in Chapter 6, it is clear
that the spherulites are space-filling. However, the density of the lamellar
stacks within a spherulite decreases towards the outside of the spherulites,
leaving more open spaces for possible stack insertion or stack thickening,
Figure 7.20. Since the average lamellar thickness decreases, these newly
formed lamellae are thinner than the existing ones. At this point, a non-
space-filling morphology is assumed, in Section 7.5, the fraction of semi-
crystalline stacks will be determined to verify if this assumption holds or
if another model needs to be proposed.
Upon reheating, the imperfect, thinner crystals, formed during cooling,
remelt before the crystals formed during primary crystallisation as was
7.4 SAXS 147
Table 7.3: Parameters of the ideal two-phase structure of CBTr

resin, pCBT and PBT after crystallisation at 190◦ C (left) and
170◦ C (right) in Å
190◦ C 170◦ C
Lp l1 l2 Lp l1 l2
CBTr resin 135 48 87 304 59 245
remolten pCBT 137 44 93 127 37 90
pCBT polymerised at 250◦ C 144 42 102 133 40 93
PBT 140 44 96 124 38 86
Figure 7.21: Schematic representation of electron densities

showing that half of the transition layer is included in the char-
acteristic lengths of the corresponding ideal two-phase system
seen in the WAXD data. An increase similar to the decrease upon cooling
in both long period and lamellar thickness is also evident from Figure 7.19.
Table 7.3 compares the characteristic lengths of all samples crystallised
at 190◦ C and 170◦ C. For the samples crystallised at 190◦ C, the differences
are relatively small. Surprisingly, the sample which was simultaneously
polymerised and crystallised exhibits the smallest lamellar thickness. A
possible explanation for the (apparent) smaller lamellar thickness can be
found in the TEM images shown in the previous Chapter (Figure 6.13,
page 118).
From these images, it was clear that the transition from crystalline
to amorphous was more sharp when polymerisation and crystallisation
were simultaneous. Figure 7.21 shows a schematic representation of the
electron densities and characteristic lengths as determined by SAXS for
a two-phase system containing a transition layer and the corresponding
Figure 7.22: Influence of glass fibres on the parameters of the

ideal two-phase structure during simultaneous polymerisation and
crystallisation at 190◦ C
ideal system. Since part of the transition layer is included into the lamellar
thickness, the core lamellar thickness is smaller. A larger transition layer
might hence result in an increased apparent lamellar thickness, whereas
the core lamellar thickness might be similar or even smaller.
At 170◦ C however, it is interesting to note the parameters of the simul-
taneously polymerised and crystallised sample. From the WAXD data, it
was clear that polymerisation is not completed and the overall degree of
crystallinity is very low. It is therefore obvious that the assignment of l1
and l2 differs from the assignment for the other samples. The lower value
is hence considered to correspond to the lamellar thickness which equals
59 Å. This value is still larger than l1 of the other samples, confirming the
assignment of the larger value to the lamellar thickness for these samples.
The characteristic lengths have only a minor decrease with crystalli-
sation temperature. This is consistent with the data obtained by Hsiao et
al. [159] for PBT. This difference, however, should increase upon increas-
ing the temperature difference, which was not investigated here.
Figure 7.22 shows the influence of the presence of glass fibres during
simultaneous polymerisation and crystallisation at 190◦ C. The decrease
in long period and lamellar thickness is more steep when glass fibres are
present during crystallisation. The effect of adding glass fibres during
isothermal crystallisation of pCBT after polymerisation at 250◦ C is de-
picted in Figure 7.23. Since the interfibre distance3 in our samples is
around 10 µm, the lamellar structure development should not be hindered
by the presence of these glass fibres.
Nevertheless, there seems to be a small decrease in characteristic lengths.
The glass fibres represent extra nucleation sites, which lead to a higher
overall degree of crystallinity and an increase in crystallisation rate. From
the experimental data, it seems that the formation of slightly smaller
3 calculated from the known sample volume and weight percentage of glass fibres,
assuming square packing of the fibres

7.5 Comparison of SAXS and WAXD results 149
Figure 7.23: Influence of glass fibres on the parameters of the

ideal two-phase structure during isothermal crystallisation of
pCBT (left) at 190◦ C and (right) at 170◦ C
lamellae is favoured. This may however simply be a time effect due to the
increased nucleation. In order to draw any firm conclusion, crystallisation
times should be increased to reach a plateau value for both cases.
7.5 Comparison of SAXS and WAXD results

7.5.1 The invariant
The total scattering power or invariant is proportional to the electron
density difference and the volume fractions of crystalline and amorphous
phases as is explained in Appendix B, Section B.4.1. Two expressions are
proposed in literature for the invariant.
Qid = C 0 ϕ(1 − ϕ)(%c − %a )2 (7.7)
Qid = Cαs φ(1 − φ)(%c − %a )2 (7.8)

with ϕ and (1 − ϕ) the overall volume fraction of the crystalline and
amorphous phase, φ and (1 − φ) the local volume fractions, (%c − %a ) the
temperature dependent electron density difference between the crystalline
and amorphous phase, αs the fraction of semi-crystalline stacks in the
irradiated volume and C and C 0 constants accounting for the relative scale
of the measured intensities. For simplicity, the models corresponding to
Equations 7.7 and 7.8 will be referred to as the overall and local model
respectively.
With these models, the results obtained from WAXD can be com-
pared to those from SAXS. In Equation 7.7, the overall volume fractions
should be equal to the volume fraction of crystallinity obtained by WAXD
whereas in Equation 7.8, the WAXD crystallinity should be compared to
αs × φ with φ the local crystalline volume fraction determined previously
with the quadratic equation.
The major difference between these models is that the overall model
assumes that the scattering of the superstructure contributes to the scat-
tering power, whereas the local model excludes this scattering based on in-
strument resolution4 . Santa Cruz et al. [163] considered the overall model
(Equation 7.7) more correct for melt-crystallised PET. They assumed,
however, that the thickness of the lamellar stacks and the distances be-
tween the lamellar stacks are so small, that the scattering arising from
this density difference can be resolved and hence contributes to the scat-
tering power (invariant). During the early stages of crystallisation, when
the spherulites and hence the semi-crystalline stacks are not at all space-
filling, this assumption cannot be valid.
In order to consider these models, the constants C and C 0 as well as
the temperature dependent electron density difference need to be known.
Temperature dependence of amorphous and crystalline density

The electron density difference should be known to correlate the degree
of crystallinity to the invariant. Moreover, in order to compare this value
to the weight fraction of crystallinity as determined by WAXD, the mass
density of both amorphous and crystalline PBT are also necessary. The
electron density, %, differs from the mass density, ρ, by a constant factor,
Equation 7.9. · − ¸
h g i 220 [g/mol] e
ρ = h −
i % (7.9)
cm3 116 e /mol cm 3
Huo et al. studied the thermal expansion of the unit cell of PBT [168].
They found a linear relation between temperature and the volume of the
unit cell. This relation was used to calculate the temperature dependence
of the crystalline density. Fakhreddine et al. on the other hand calculated
the equation of state for a PBT melt [169]. By extrapolating these data,
the amorphous density was determined. Equation 7.10 gives the calcu-
lated temperature dependencies of both the amorphous and crystalline
mass density with T in ◦ C.
ρam = 1.2715 − 6.60 · 10−4 T (7.10)

ρcrys = 1.4108 − 5.56 · 10−4 T (7.11)
Determining constants C and C 0

The constant C or C 0 in both models is introduced to account for the
relative scale of the measured intensities. This constant is independent
of temperature and should hence not differ between subsequent measure-
ments with the same sample. In order to determine this constant, some
assumptions need to be made. These assumptions can include setting αs
4 two cases described by Strobl [184], Section B.4.1
Figure 7.24: Comparison between SAXS and WAXD overall de-

gree of crystallinity calculated with the local model (left) assum-
ing αs = 1 at 100◦ C and (right) assuming αs = ϕwaxdφ
at 100◦ C
to a fixed value at a certain time and temperature or correlating the SAXS

results to the overall volume fraction of crystallinity obtained by WAXD.
Results
Since the local model is assumed to be the only correct one during the
early stages of crystallisation, this model was first considered. The most
straightforward assumption to determine C is to consider αs very close
or equal to one after cooling to 100◦ C since this coincides with the TEM-
images that do not reveal any large amorphous pools. The second option
is to determine αs so that its product with the local crystalline volume
fraction equals the WAXD overall crystallinity. Figure 7.24 shows the
results of both assumptions that were made.
It is clear that for neither case, a good correlation between WAXD
and SAXS is obtained. When the semi-crystalline stacks are considered
to be space-filling after cooling, the overall crystallinity is seriously over-
estimated. With the second assumption, fitting the WAXD crystallinity
to the overall SAXS crystallinity at the end of the cooling, this overes-
timation decreases, but the volume fraction of semi-crystalline stacks is
only 60% after simultaneous polymerisation and crystallisation at 190◦ C,
righthand-side of Figure 7.25. Keeping the TEM-images in mind, such a
large fraction of amorphous pools is impossible. Moreover, in both cases,
αs does not change during cooling and heating which is contradictory to
the variations in characteristic lengths as discussed above.
The overall model, although inaccurate during the early stages of
crystallisation, was also considered. The constant C 0 was determined
by comparing the WAXD crystallinity to the calculated overall SAXS
crystallinity at the end of the first isothermal stage at 190◦ C. The left-
hand-side of Figure 7.25 compares the results. Surprisingly, this model
seems to fit our data quite well, also in the beginning.
The righthand-side of Figure 7.25 depicts the volume fraction of semi-
Figure 7.25: (left) Comparison between SAXS and WAXD over-

all degree of crystallinity calculated with the overall model, (right)
ratio between overall and local degree of crystallinity according
to local (αs 6= 1) and overall model (β = ϕ φ
)
crystalline stacks calculated according to the local model as well as the

ratio of the overall and local SAXS crystallinity as determined with the
overall model, β, defined in Equation 7.12.
ϕ
β= (7.12)
φ
Both fractions, αs and β, demonstrate an interesting feature upon heating

and cooling. At a certain point during cooling, a maximum is reached and
although this maximum is more pronounced after simultaneous polymeri-
sation and crystallisation, it is also present in the second temperature
profile. The maximum in αs can be related to the ’closing’ of lamellar
stacks. In this case, (1 − αs ) is not only related to fully amorphous pools
but also includes a fraction of fully crystalline pools.
The maximum in β, on the other hand, could to be related to the
effect of the lateral dimensions of the crystals when φ approaches 0.5 as
described in Section 7.4.1. Due to the ‘limited’ lateral dimensions, the
calculated local degree of crystallinity is underestimated and can even
show a maximum. It is this effect that prohibits a further increase of β.
Neither model proposed at this point fits the data at hand. Even
though the overall model seems to be more correct, question remains
to which physical entity β should be attributed. Therefore, both the
crystallisation route according to Strobl and the rigid amorphous fraction
are introduced.
7.5.2 Crystallisation via a mesomorphic phase

Generally, it is assumed that lamellar crystallites grow directly from the
entangled melt. Based on experimental observations found in literature
and own experiments, Strobl [170] proposed a new route for crystallisation,
which is depicted in Figure 7.26.
Figure 7.26: Sketch of the route, proposed by Strobl, for the

formation of polymer crystallites. Stages passed through as re-
flected in the structural states found along a layer, adapted from
[170]
Crystallisation starts with the attachment of chain sequences from the

melt onto the lateral growth face of a layer with a mesomorphic, pre-
ordered, inner structure. This layer has a minimum thickness in order to
be stable in the surrounding melt. The inner mobility of this layer is high
and the layer can thicken with time by a continuous rearrangement of the
chain sequences in the zone composed of folds and loops near to the layer
surface. As the layer thickens, its mobility is lowered and finally a critical
thickness is reached where a transition into a higher-ordered, crystalline
structure occurs. This leads to a solidification of the layer and a stop of
the thickening. The resulting structure is described as a granular crystal
layer.
The last step of the route consists of the merging of the granular crys-
tals, together with an improvement of their inner perfection, resulting
in the well-known lamellar crystallites. Schick et al. [171] on the other
hand proposed a vitrification of the surroundings of the granular crys-
tals, preventing the blocks from merging because of a lack of mobility.
They attributed this intergranular vitrified material to the rigid amor-
phous fraction, described in the next section.
The existence of a pre-ordered, mesomorphic phase, proceeding the

formation of crystalline granules and lamellae affects the crystallinity as
determined by SAXS since, depending on its density, (part) of the meso-
morphic phase can be included in the ‘crystalline’ fraction. The two-phase
model as used in the SAXS correlation function analysis should therefore
be modified to include a third phase. The mesomorphic phase is a mobile
phase which is denser than the amorphous phase but not crystalline, it
will from hereon be referred to as the dense amorphous fraction, DAF.
7.5.3 Rigid amorphous fraction (RAF)
At room temperature, the existence of a third phase is already widely

acknowledged for most semi-rigid polymers such as polyesters [172–177]
and polyamides [178], namely the rigid amorphous fraction. This phase
is highly constrained because of its close proximity to crystallite surfaces.
Although its existence is recognised, no consensus has been reached con-
cerning the formation, stability, location and density of the rigid amor-
phous fraction.
For PBT, it is known that already at 5-8% crystallinity, a rigid amor-
phous fraction exists [172], which partially devitrifies between 40◦ C and
50◦ C. This temperature range is often considered the glass transition tem-
perature of PBT even though the real glass transition of the mobile amor-
phous phase is around -25◦ C.
Righetti et al. [175] indicated that for PBT, the RAF is stable until
the meting point of the crystalline PBT. This view is shared by Schick et
al. [171], although not for PBT specifically. Pyda et al. [179] on the other
hand estimated a glass transition temperature of 102◦ C for the RAF based
on heat capacity measurements. They also determined a RAF of 21% for
a crystallinity of 36%, showing that this fraction is quite substantial.
The rigid amorphous fraction of PET, its location and effect on me-
chanical and physical properties, is studied more frequently. The rigid
amorphous fraction in PET is around 25% for a degree of crystallinity of
only 30% [174, 177]. This relatively high fraction has an effect on the oxy-
gen solubility and the yield stress, which is proportional to the crystallite
size which, at its turn, depends on the RAF.
Most researchers consider the RAF to be located at the interphase
region between the crystalline lamellae and the interlamellar amorphous
regions [173, 177], having a density which varies between the crystalline
and amorphous density. Others consider the complete interlamellar amor-
phous region to be rigid [176] and a density that is below that of the mobile
amorphous fraction since the rigid amorphous fraction is frozen in at a
higher temperature [174].
Recently, however, a comparative solid-state NMR (nuclear magnetic
resonance), SAXS and WAXD study on polyamides suggests that the
RAF is situated inside the crystalline lamellae and not at the interphase
leading to a morphology consisting of crystalline grains of limited lateral
dimensions separated by the rigid amorphous phase [178]. Implicitly, this
model assumes the density of the RAF to be equal to the crystalline
density. The existence of crystalline grains is consistent with the model
of Strobl for crystallisation, assuming that the granular crystals do not
merge (completely). As mentioned above, this view was also proposed by
Schick et al. [171].
7.5.4 Effect of a third phase (dense amorphous or

rigid amorphous) on the local crystallinity
Adapted two phase models
Table 7.4 shows an overview of adapted two phase models, including the
third phase which is either a rigid amorphous or a dense amorphous (and
mobile) fraction. In the first case, the transition layer between crystalline
and amorphous is considered to consist out of RAF with a varying den-
sity, bridging the gap between the crystalline and amorphous density,
Figure 7.21. Determining the thickness of this transition layer from the
correlation function analysis is possible, but strongly depends on the ex-
trapolation to infinity and the fitting range for this extrapolation [180].
Hence, the corresponding ideal two-phase system is most often used to
determine the characteristic lengths and local volume fractions. If such a
transition layer is however present, half its fraction or one full interphase
layer is included in the local crystalline fraction. The fraction of the third
phase within the crystalline lamellae, considered by SAXS, is called γ,
note that this is not the overall fraction of the third phase and that the
definition of γ depends on the case considered.
In the second case, the RAF is also located at the crystalline-amorphous
interphase, but has a density equal to the crystalline density. Hence, its
fraction or two full interphase layers are included into the local crystalline
fraction determined by SAXS. A more correct nomenclature for the frac-
tion determined by SAXS would hence be local dense fraction.
The third case places the RAF or DAF inside the crystalline lamellae
as the material in which crystalline grains are imbedded and assumes the
density to be equal to the crystalline density. In all three cases, the local
volume fraction of crystallinity as determined with SAXS is overestimated
by a factor (1 − γ).
Therefore, the local degree of crystallinity φ as determined by the
quadratic expression in SAXS does not only differ from the overall degree
of crystallinity ϕ because the semi-crystalline stacks are not space-filling
but also because of the rigid or dense amorphous fraction attached to or
inside the lamellae.
In the models considered above, not all options are considered. First,
the density of the rigid amorphous fraction could be smaller than or equal
to the amorphous density. A density smaller than the amorphous density
seems unlikely if the RAF is located at the interphase between the crys-
talline and mobile amorphous phase since then a density dip would be
present. If the density is equal to the amorphous density, the local SAXS
model would yield correct results since SAXS would not distinguish be-
tween mobile and rigid amorphous.
Table 7.4: Schematic overview of adapted two-phase models and resulting local crystallinity, showing the crystalline 156
lamellae, the rigid amorphous fraction or dense mobile fraction (grey) and mobile amorphous fraction (white), γ is the
fraction of the third phase, RAF or DAF, that is included in the SAXS local crystallinity, φ
schematic 3rd phase local crystallinity
lRAF D 2 lc
RAF γ = lc D 2 φ = Lp
lRAF lc −lRAF
= lc φcrys = Lp
ρ varying
between = φ − Llcp lRAF
lc
ρc and ρa = φ − φγ
φcrys = φ(1 − γ)
2lRAF D 2 lc
RAF γ = lc D 2 φ = Lp
2lRAF lc −2lRAF
= lc φcrys = Lp
ρ = ρc
= φ − Llcp 2lRAF
lc
= φ − φγ
φcrys = φ(1 − γ)
n2 lc (dD+d(D+d)) lc
DAF γ = n2 lc (d+D)2 φ = Lp
2dD+d2 n2 lc D 2
ρ ≈ ρc = (d+D)2 φcrys = n2 Lp (D+d)2
lc D 2 +2dD+d2 2dD+d2
rigid = Lp ( (D+d)2 − (D+d)2 )
or φcrys = φ(1 − γ)
Time-resolved X-ray measurements
mobile
The representation of the granular model is simplified, in reality, the crystal granules and the RAF/DAF is not nicely ordered, moreover the
crystal granules can even tough eachother, creating islands of RAF/DAF
In the granular model, a large density difference between the crys-

tals and third phase (either rigid or mobile), would contribute to the
SAXS intensity profile. Moreover, the limited lateral dimensions would
seriously limit the use of the quadratic expression as was shown before.
Furthermore, since the staining procedure of TEM is based on the density
difference between the phases, the crystalline grains would appear in the
TEM-images, which is not the case. Hence, if the density of the third
phase differs from that of the crystals, this difference is believed to be mi-
nor and hence the contribution of the extra lateral surfaces can be ignored
when calculating the specific inner surface of the crystals, validating the
use of the quadratic expression.
During the early stages of crystallisation, the scattering caused by
the supermorphology cannot be resolved experimentally. Therefore, only
the local model can be valid. The quadratic expression however does not
determine the local crystalline volume fraction, but the local dense volume
fraction, which overestimates the crystalline fraction by a factor (1 − γ).
In order to determine αs and γ, it is assumed that at the lowest temper-
ature, which is 100◦ C, the lamellar stacks are approximately space-filling,
αs,100 ≈ 1. With this assumption, the constant C can be calculated
according to Equation 7.13, after which αs and γ are calculated with
Equation 7.14 and 7.15
Qid,100 = Cφ100 (1 − φ100 )(%c − %a )2 (7.13)

Qid = Cαs φ(1 − φ)(%c − %a )2 (7.14)
ϕwaxd = αs (1 − γ)φ (7.15)
The overall fraction of the third phase, either rigid amorphous or dens
amorphous (and mobile), is then determined with Equation 7.16.
RAF or DAF = αs · γ · φ (7.16)
It is interesting to note that the fraction β introduced previously to fit

the overall model to the experimental data and the WAXD crystallinity
hence incorporates both αs and (1 − γ) according to Equation 7.17.
ϕ
β= = αs · (1 − γ) (7.17)
φ
Selecting a suitable adapted two-phase model

A first step in selecting a model that fits the present data is the de-
termination of the lateral dimensions of the crystals with the Scherrer
equation. D010 and D100 were calculated for different processing condi-
tions, Table 7.5. It is clear that for the simultaneous polymerisation and
crystallisation at 190◦ C, the lateral dimensions are substantially larger
Table 7.5: Crystallite size after crystallisation at 190◦ C of

CBTr resin, pCBT and PBT
D010 D100
r
CBT resin 256 152
remolten pCBT 218 138
pCBT polymerised at 250◦ C 220 126
PBT 228 117
than for the other processing conditions. Nevertheless, the lateral dimen-
sions are small and in the same order of magnitude as the long period,
favouring the model that divides the crystalline lamellae into crystalline
grains of limited lateral dimensions imbedded in a third phase.
If the density of this third phase differs substantially from the crys-
talline density or if these crystals were imbedded into the mobile amor-
phous phase, the quadratic expression is not valid anymore and would not
yield local crystallinities in between 80 and 20% according to Figure 7.17.
It can therefore be concluded that the density of the third phase should
be close or equal to the crystalline density in order to exclude the effect
of the limited lateral dimensions. Question remains if this dense phase is
rigid or mobile.
Figure 7.27 depicts the volume fraction of semi-crystalline stacks as
well as the local fraction of the third phase for isothermal crystallisation
at 190◦ C and subsequent cooling and heating. The local dense volume
fraction (sum of DAF and crystalline) and corresponding lamellar thick-
ness are shown in Figure 7.28. The fraction of semi-crystalline stacks in-
creases quite rapidly during the early stages of crystallisation after which
it remains fairly constant5 during the second stage of the isothermal crys-
tallisation and subsequent cooling and heating.
The local fraction of DAF clearly shows an opposite trend. At the
onset of crystallisation, it has an initial value close to one, supporting
the view of Strobl, that a mesomorphic, pre-ordered (dense but not crys-
talline) structure precedes the formation of the crystalline structure. In
the first stages of crystallisation, it decreases rapidly, followed by a more
modest decrease. Upon cooling and heating, γ does not remain constant,
clearly favouring the third phase to be a mobile, instead of rigid, but dense
amorphous phase. If this third phase, situated in between the crystalline
grains, were rigid, its fraction would be either constant or increase upon
crystallisation.
Figure 7.29 shows more clearly the trends in the volume fractions and
5 The fraction of semi-crystalline stacks actually displays a maximum during cooling
and subsequent heating (arrows in Figure 7.27). The effect of not assuming αs constant
will be briefly discussed later on.
Figure 7.27: (left) Fraction of semi-crystalline stacks, αs , and

(right) fraction of RAF or DAF inside the crystalline lamellae, γ
during isothermal crystallisation at 190◦ C and subsequent cooling
and heating
Figure 7.28: (left) Local crystalline fraction determined with

the quadratic expression and (right) lamellar thickness during
isothermal crystallisation at 190◦ C and subsequent cooling and
heating
lamellar thickness during the isothermal polymerisation and crystallisa-

tion. During the first stage of crystallisation, before the semi-crystalline
stacks are space-filling, the local fraction of DAF, γ, decreases rapidly.
This decrease is accompanied by a strong increase in overall degree of
crystallinity, ϕwaxd . When αs reaches its plateau value, γ still decreases,
although less rapidly, as does the lamellar thickness. Moreover, the over-
all crystallinity increases. The change in slope in γ indicates that as long
as the stacks are not space-filling, the mesomorphic layer can grow more
freely and hence transform into crystalline grains quite rapidly. When
the stacks become space-filling, the mesomorphic phase is constrained be-
tween the crystalline and mobile amorphous phase, its conversion is hence
slowed down, but not halted completely.
Previously, it was suggested that the increase of crystallinity and de-
crease of lamellar thickness could be explained by stack thickening or
stack insertion. Since the volume fraction of semi-crystalline stacks re-
mains rather constant, this is not the case. The crystallinity should hence
increase due to a further transition from dense amorph (mesomorph) to
crystalline.
Figure 7.29: Volume fractions and lamellar thickness during

isothermal polymerisation and crystallisation at 190◦ C, αs , vol-
ume fraction of semi-crystalline stacks, γ, volume fraction of DAF
within the dense fraction, φ, local dense volume fraction, ϕwaxd ,
overall crystalline volume fraction
The latter structural transition is coupled with a decrease in lamellar

thickness because of the exclusion of entanglements and defects from the
lamellae into the mobile amorphous phase. Since the crystallisation tem-
perature is constant, question remains why stable thinner lamellae can be
formed.
The crystals formed during primary crystallisation, before the semi-
crystalline stacks become space-filling, and primary nucleation, not grow-
ing from an existing crystal, are thicker compared to crystals formed by
extension of a existing crystal. The stability of thinner crystals upon
secondary, tertiary or quaternary nucleation is obvious from the Gibbs-
Thomson equation, Equation 7.18.
µ ¶
∞ 2σe nσs
Tm = Tm 1 − − (7.18)
∆Hm lc ∆Hm D
with σe the fold surface free energy and σs the lateral free energy and
n the number of lateral surfaces to be formed. Depending on the lateral
surfaces to be formed, n can vary between four (primary nucleation) and
0 (quaternary nucleation, depicted on the righthand-side of Figure 7.30).
The last term of Equation 7.18 is often ignored since the lateral dimensions
of the crystals are believed to be large compared to their thickness. Our
results however show that this is not the case.
Therefore, the crystals formed during secondary crystallisation (αs ≈ 1),
have a smaller lamellar thickness than the original thickness. Figure 7.30
schematically represents two possible cases for the increase in crystallinity
Figure 7.30: Schematic representation of secondary crystallisa-

tion (αs ≈ 1), the mesomorphic phase transforms into the crys-
talline phase by excluding of entanglements and defects, leading
to a smaller lamellar thickness, (left) secondary nucleation, (right)
quaternary nucleation
with thinner lamellae after the semi-crystalline stacks have become space
filling.
Upon cooling, the dense mobile phase further transforms into less per-
fect, thinner crystals which upon reheating are not stable until the melting
point of the primary crystals. The melting of these less perfect crystals
does not result in a true mobile amorphous phase, but again a (slightly
thicker) dense amorphous phase is formed as can be seen from the increase
in γ, lc and φ, the local dense volume fraction, Figure 7.27 and 7.28. Since
prior to melting, the crystals pass again through the dense amorphous
phase, this phase can be stable with respect to the melt. Moreover, the
constraints posed by the adjacent crystals might prevent the formation of
the less dense true amorphous fraction.
The overall fraction of the dense amorphous, but mobile, phase (αs γφ)
is shown in Figure 7.31. It might be possible that upon cooling below a
certain glass transition temperature, this dense amorphous fraction (par-
tially) transforms into the rigid amorphous fraction by vitrification. No
experiments were however performed to verify the vitrification of the dense
amorphous fraction. If, however, all dense amorphous would vitrify, the
amount of RAF obtained would be high but in reasonable agreement with
estimates based on calorimetry [174, 177, 179].
It is clear that upon simultaneous polymerisation and crystallisation,
the overall fraction of dense amorphous is smaller as is the fraction of dense
amorphous inside the lamellae. In view of the crystallisation process when
polymerisation is not complete, namely short amorphous chain segments
at the mesomorphic growth surface, less entanglements and defects are
included into the mesomorphic phase. Moreover, crystallisation is slowed
down by the polymerisation process, therefore, increasing the overall time
frame for crystallisation. This allows for a more straightforward transition
Figure 7.31: Overall fraction of dense mobile phase during

isothermal crystallisation at 190◦ C and subsequent cooling and
heating
from mesomorphic into crystalline, leading to a smaller fraction of dense

amorph and an increase in lateral dimensions of the crystalline grains.
Fraction of semi-crystalline stacks

In the previous section, the fraction of semi-crystalline stacks, αs was
considered constant and approximately one during cooling and subse-
quent heating (before the onset of melting). As was mentioned before,
the fraction of semi-crystalline stacks actually slightly increases during
the first part of cooling and then decreases again as can be seen on the
righthand-side of Figure 7.32. An opposite trend is seen during heating.
During the early stages of crystallisation, the fraction (1 − αs ) is at-
tributed to fully amorphous pools since the stacks are not space-filling.
Once the fraction of semi-crystalline stacks has, however, reached its max-
imum the decrease upon further cooling can no longer be assigned to
fully amorphous pools. The fraction (1 − αs ) should now be considered
to consist of fully dense pools, being either fully crystalline, fully dense
amorphous or partially crystalline and partially dense amorphous.
In this case, the morphology consists of semi-dense stacks (alternating
liquid amorphous versus lamellae consisting out of crystalline and dens
amorphous), alternated with closed packed, dense lamellae, which is a
picture fully compatible with the TEM images.
The fraction of dense amorphous inside the lamellae has to be calcu-
lated differently, Equation 7.19.
ϕwaxd − (1 − αs )(1 − κ)
γ =1− (7.19)
αs φ
Figure 7.32: (left) Fraction of semi-crystalline stacks, αs , and

(right) fraction of DAF inside the crystalline lamellae considering
the dense pools either fully crystalline (κ = 0) or fully dense amor-
phous (κ = 1), γ during cooling after isothermal polymerisation
and crystallisation at 190◦ C
And the overall fraction of dense amorphous can be calculated using Equa-
tion 7.20.
DAF = αs γφ + (1 − αs )κ (7.20)
with κ the fraction of DAF inside the dense pools. Still assuming that αs
equals one at its maximum6 , the fraction of dense pools is rather small
(< 5%) and hence its effect is only minor. When polymerisation and
crystallisation are separated, the fraction of dense pools is even lower,
< 3%.
Moreover, the current data do not allow for the determination of κ. It
is, however, most likely that the dense pools contain both a crystalline and
dense amorphous fraction. To illustrate the effect of not taking αs con-
stant during heating and cooling, the lefthand-side of Figure 7.32 depicts
the result of two extreme cases, namely κ = 0 and κ = 1.
The first case considers the dense pools to be fully crystalline whereas
in the second case, the dense pools only consist out of dense amorphous.
Then, γ does not differ from the originally calculated γ, not taking the
dense pools into account, the overall DAF, however, increases by (1 − αs )κ,
Equation 7.20.
The last, and most realistic, case assumes κ = γ, or truly closed packed
lamellae. Here, the fraction of dense amorphous in the dense pools, equals
its fraction inside the lamellae of the semi-dense stacks.
ϕwaxd
γ =1− (7.21)
αs φ + (1 − αs )
DAF = αs γφ + (1 − αs )γ (7.22)
6 The assumption α = 1 at its maximum is not validated, however, based on the

s
microscopical evidence, this assumption is the most straightforward
7.6 Conclusions
Time-resolved X-ray measurements were performed to investigate the
structure development of CBTr resin at the processing conditions of ther-
moplastic RTM. For comparative reasons, isothermal crystallisation from
the melt was also studied. Since during the processing of composites, glass
fibres are present, their effect on the crystallisation is also considered.
The isothermal production process of CBTr resin results in a simul-
taneous polymerisation and crystallisation since the increase in induction
time for crystallisation is followed by a decrease crystallisation rate com-
pared to melt crystallised pCBT. This simultaneity results in a higher
overall degree of crystallinity. Upon cooling, the degree of crystallinity
increases but equally decreases upon reheating before the main melting
temperature is reached, indicating the formation of smaller, imperfect
crystals, which is confirmed by SAXS data.
No significant difference was found between the CBTr resin poly-
merised at 250◦ C and subsequently crystallised and the molten pCBT
crystallised from the melt. The results were also comparable to literature
data indicating similar behaviour as commercially available PBT.
Isothermal processing at 170◦ C, which would increase the time win-
dow for impregnation during composite production, is shown not to be
an option since, due to the relative fast onset of crystallisation and the
entrapment of active polymerisation sites, polymerisation was incomplete
as was evident from the low melting point of the final polymer.
Glass fibres are known to promote heterogenous nucleation and hence
decrease the induction time and increase the crystallisation rate. For the
melt-crystallised samples this was indeed the case. For the isothermal
process, the induction time was however increased in the presence of glass
fibres. It is believed that the effective amount of catalyst is decreased by
the glass fibre sizing reducing the polymerisation rate. After the delayed
onset of crystallisation, the crystallisation rate is nevertheless increased.
The assignment of characteristic lengths determined by the correlation
function analysis is very ambiguous in literature. The assignment here was
performed by comparing room temperature WAXD and SAXS results
which both yield the lamellar thickness, and leads to the larger length
being assigned to the lamellar thickness. The local degree of crystallinity
as determined by SAXS is hence substantially larger than the overall
degree of crystallinity.
The lamellar thickness is almost unaffected by the thermal history
of pCBT although surprisingly, for the simultaneous polymerisation and
crystallisation the lowest lamellar thickness was measured. This small
difference is believed to be related to a more pronounced transition layer
when polymerisation and crystallisation are separated, which was con-
firmed by TEM images. The lateral dimensions of the lamellae are, on
7.6 Conclusions 165
the other hand, larger for simultaneous polymerisation and crystallisation.

Only a small difference in lamellar structure was observed upon de-
creasing the crystallisation temperature. Also the glass fibres do not inter-
fere with the lamellar structure development since the interfibre distance
is quite large.
During the isothermal crystallisation of CBTr resin, the average lamel-
lar thickness decreases both when crystallised from the melt and when
crystallised during polymerisation. This decrease is continued upon cool-
ing, confirming the WAXD results. The morphology consistent with these
data as proposed in literature, consists of non-space-filling lamellar stacks
with the formation of thinner lamellae upon secondary crystallisation.
The new lamellae are then formed by stack insertion or stack thicken-
ing. This morphology is however not supported by the current data since
this would lead to a fraction of amorphous pools of more than 40% and
moreover, the fraction of semi-crystalline stacks was shown to be almost
constant during cooling.
It is known that the two phase structure, as assumed by the corre-
lation function analysis, is not correct at room temperature due to the
existence of a rigid (dense) amorphous fraction. Moreover, according to
Strobl, a mesomorphic, mobile and partially ordered phase exists during
crystallisation leading to a granular crystalline structure. The crystalline
grains can than further grow leading to the well-known lamellar struc-
ture. Others however suggested that the mesomorphic phase transforms
into the rigid amorphous phase when crystalline grains cannot merge. In
light of the existence of a third phase, several morphological models were
considered.
Because of the limited lateral dimensions of the crystals, it was con-
cluded that the third phase is present inside the crystalline lamellae, di-
viding them into grains. Since the limited dimensions of the crystals do
not limit the use of the quadratic expression, the density of this fraction
was considered to be equal or close to the crystalline density. The local
degree of crystallinity as determined by SAXS is hence not only related to
the WAXD-based overall degree of crystallinity by the volume fraction of
semi-crystalline stacks, but also depends on the fraction of a third, dense
phase inside the lamellae.
After the semi-crystalline stacks have become space-filling, the further
increase in crystallisation is related to the transition from the third phase
into the crystalline phase. The continued decrease of this third phase
upon cooling excludes the possibility of this phase being rigid. Hence, it is
concluded that the third phase is a mobile, dense amorphous phase. Since
the fraction of this dense amorphous phase inside the lamellae equals one
at the onset of crystallisation, indicating that the transition from mobile
amorphous to crystalline passes through this dense amorphous phase, the
crystallisation route proposed by Strobl is supported.
The overall fraction of dense amorphous is lower for simultaneous poly-

merisation and crystallisation compared to crystallisation from the melt.
The increased time frame for crystallisation as well as the reduced pres-
ence of entanglements and defects into the mesomorphic phase result in
this smaller fraction.
It is suggested that the dense amorphous phase can vitrify to form the
rigid amorphous phase which is encountered at room temperature.
The time-resolved X-ray measurements revealed an interesting mor-
phology and crystallisation route. At this point, no detailed investigation
on the transition layer between the dense lamellae and the interlamel-
lar amorphous layer was performed. Such a study might however shed
more light on the hypothesis that the amount of tie-molecules is reduced
during simultaneous polymerisation and crystallisation. Special attention
should then be paid to data processing as this can strongly influence the
calculated transition layer.
Moreover, the effect of the dens amorphous, which might vitrify to the
rigid amorphous fraction, on the mechanical properties and more specifi-
cally the matrix ductility would definitely be interesting.
Chapter 8
Conclusions and Outlook
Even though thermoplastics have well-known advantages over their ther-

moset counterparts for use in composites, they only encompass a small
part of the continuously reinforced composite market. The difficulty in
impregnating the continuous reinforcement with a high viscosity polymer
melt is partially responsible for this arrear. Reducing the viscosity of ther-
moplastics during impregnation would therefore not only facilitate impreg-
nation but also open up a new window of production techniques, previ-
ously reserved for thermoset composites, such as Resin Transfer Moulding
(RTM). The use of thermoset production techniques might even assist the
change-over from one material class to the other for certain applications.
In-situ polymerisation of thermoplastic prepolymers, having a water-
like viscosity, makes the production of thermoplastic composites with
RTM a real possibility. Due to the constraints of, on one side, the in-
situ polymerisation and, on the other side, the closed mould process, only
a limited amount of polymer systems can be employed. For this study,
CBTr resin was chosen since it not only fulfills all the prerequisites, it
also offers the possibility for an isothermal process, which significantly
reduces cycle times in a non-continuous process like RTM.
At the start of this study, little was known about composites pro-
duced with thermoplastic RTM and CBTr resin. Only some data on
fibre-dominated mechanical properties of carbon fibre composites were
available. During the course of this work, the feasibility of RTM with
CBTr resin was demonstrated by the production of glass fibre reinforced
composites, moreover, their properties were investigated in detail. Both
the glass fibre sizing and the isothermal nature of the production process
influence the composite properties since the matrix properties are altered.
Special attention was paid to the matrix morphology, which led to
a more fundamental study of the crystalline structure. By using time-
resolved X-ray measurements a new morphological model could be sug-
gested.
167
168 Conclusions and Outlook
Effect of isothermal processing route
Isothermal processing of CBTr resin allows for the onset of crystallisation

before polymerisation is completed. This leads to a high degree of crys-
tallinity with a crystalline morphology consisting of thick, well-oriented
lamellae. These lamellae contain a smaller fraction of dense amorphous
material and have less tie-molecules to connect them compared to those
produced by separated polymerisation and crystallisation.
Since the isothermal processing route influences the morphology, it
also affects the mechanical properties of the matrix. The high degree
of crystallinity leads to a higher modulus which is advantageous. Un-
fortunately, the combination of this high degree of crystallinity and the
perfection of the crystalline structure results in a brittle matrix, failing
at very low strain levels in three point bending and having a poor impact
toughness.
The composites produced during this isothermal process hence lose a
major advantage over thermoset composites, namely the tough nature of
the matrix. This results in a transverse strength of unidirectional com-
posites which is below the matrix strength, but more important, it signif-
icantly lowers the fracture toughness and hence impact resistance of the
composites.
Suggestions for the improvement of the matrix toughness
For the success of thermoplastic RTM with CBTr resin, it is therefore

crucial to increase the matrix toughness, preferably without altering the
isothermal nature of the process since this would increase the cycle times
intolerably. Chemical modifications of the matrix to alter the crystalline
morphology by branching of the polymer were therefore considered but
the first trials were not successful. Copolymerisation or blending are in-
teresting options but need to be further investigated. The effect of these
modifications on the initial viscosity is essential since increasing the vis-
cosity is detrimental for the impregnation of the reinforcement.
Another option to deal with the matrix brittleness would be to step
away from the isothermal process and hence from RTM. Pultrusion e.g. al-
lows for non-isothermal processing by having different temperature zones
in the die with each of these zones at a constant temperature avoiding ther-
mal cycling of the die. Since the material moves relatively fast through
these temperature zones, a cost-effective non-isothermal process can be
obtained. Pultrusion is however only complementary to RTM since the
shapes that can be produced are less complex. Another option would be to
step away from thermoset production techniques and to produce preforms
which are subsequently compression moulded. One major disadvantage
of this production route is the introduction of an intermediate product.
Conclusions and Outlook 169
Influence of the fibre sizing

The matrix morphology has a profound influence on the composite prop-
erties. The degree of conversion and molecular weight on the other hand
also influence the non-fibre dominated composite properties, although this
influence is less pronounced than the influence of the fibre-matrix inter-
face. The fibre-matrix interface properties strongly depend on the fibre
sizing.
The fibre sizing introduces extra end-groups and hence lowers the
molecular weight. Moreover, epoxide functional groups partially crosslink
the matrix. Nevertheless, the composite mechanical properties were sat-
isfactory. Removing the sizing on the other hand resulted in inferior
transverse properties clearly demonstrating the need for a specific sizing.
To our knowledge, no specific fibre sizing is developed for the in-situ
polymerisation of CBTr resin. A suitable sizing should however be devel-
oped which allows for textile processing and which is compatible with the
polymerisation reaction. If, moreover, the fibre-matrix interface could be
tailored by taking advantage of the reactive nature of the production pro-
cess, hence chemically linking the fibre sizing to the matrix, composites
produced by in-situ polymerisation would have an extra asset compared
to traditionally produced thermoplastic composites.
Fundamental crystallisation study

The development of cyclic oligomers, which can be isothermally processed
is not only interesting for the production of composites. Isothermal pro-
cessing leads to simultaneous polymerisation and crystallisation and hence
leads to a new crystallisation route for polyesters. The last part of this
study was therefore devoted to a more fundamental study of the crystalli-
sation process by using time-resolved X-ray measurements.
Although this fundamental study has less of a direct influence on the
application at hand (the production of thermoplastic composites), it did
reveal some interesting features concerning the crystallisation of polyesters
and the lamellar structure consisting not only of crystalline lamellae and
an amorphous interlaminar phase, but also a mobile, dense and amorphous
phase which is believed to form the rigid amorphous phase upon cooling.
The effect of the third phase on the matrix properties such as the duc-
tility has not yet been investigated but would certainly be an interesting
subject to study.
Overall conclusion
In conclusion, this study did not only show the feasibility of and char-
acterised the in-situ polymerisation process for the production of contin-
uously reinforced thermoplastic composites, it equally demonstrated the
170 Conclusions and Outlook
need to employ a variety of techniques and knowledge domains to fully

understand the engineering problem.
Several research strategies should therefore be developed in parallel to
ensure the future of continuously reinforced composites based on CBTr
resin. On one side, a more fundamental study of the polymer system and
crystallisation route and possible matrix modifications to toughen the
system is needed. The matrix modifications should however not increase
the initial viscosity. Such a study should be complemented by a thorough
investigation on the fibre-matrix interface in order to develop a suitable
fibre sizing.
On the other side, a more engineering approach to assess different
production routes for composites together with the development of suit-
able simulation tools for impregnation, polymerisation and crystallisation
could force a break-trough of this polymer system in the composites in-
dustry.
Appendix A
Production of a leaf
spring prototype
A.1 Introduction
One of the main goals of the European project Amiterm (Contract G1RD-
CT 2000-00455; Jan 2001-Dec 2003) was to develop a unique mould tech-
nology to allow for the efficient production of thermoplastic composite
parts. The project consisted out of two subprocesses, in which JETex
dealt with the in-situ polymerisation process. At the end of the project,
a leaf spring prototype (outer dimensions: 1280 × 113 × 17 mm3 ) for a
small three wheel city car from Cree AG (Switzerland), was produced
with CBTr resin. This appendix will describe the production of this
prototype, which was a small part of the overall project. For more infor-
mation, the interested reader is referred to [132, 181, 182].
A.2 Materials
Keeping in mind the mechanical and dimensional specifications, the leaf
spring was designed to contain a fibre volume fraction of 50%. Of these
fibres, 85% was oriented along the leaf spring axis, whereas only 15% of
the fibres were oriented in the bias direction, ±45◦ . Braiding was chosen
early on in the project as the textile production method. Figure A.1 shows
the final braided preform for the leaf spring which consisted out of 6 layers
of braiding (ECA-UD, described in Section 4.1.2), braided over a metal
template, which was removed after the braiding process. The widening
shape of the design of the leaf spring is obvious from the leaf spring
preform. Unfortunately, the template was slightly oversized causing the
preform to be too wide. This resulted in difficulties in placing the preform
171
172 Production of a leaf spring prototype
Figure A.1: Leaf spring preform showing the braided structure

as well as the widening shape of the leaf spring design
into the mould cavity. The total weight of the leaf spring preform was
2.1 kg.
Two matrix materials were used, namely CBTr resin and an epoxy
resin (Epikote 828), which were described previously, Section 4.1.1.
A.3 Production set-up

Since the leaf spring was substantially larger than the A4-sized plates pro-
duced before and hence contained a larger amount of material, a scaled-up
injection unit, Figure A.2, was manufactured which was based on the im-
proved lab-scale set-up described in Section 4.2.2. The main changes were
the enlargement of the vessel in order to process quantities up to 4 kg and
the possibility to use overpressure during injection.
In order to apply pressure on the vessel, proper sealing is crucial.
In combination with the high temperature of the melt however, sealing
becomes quite difficult. Therefore, it was decided to only heat the bottom
part of the vessel and to keep the upper part as cold as possible.
The mould design was not only based on the geometrical specifications
of the leaf spring, the injection strategy was also taken into account, with
the most specific feature being a distribution channel along the length of
the leaf spring, Figure A.3.
A.4 Production
In order to enlarge the time window for fibre impregnation with the reac-
tive mixture, the oligomer melt (2 kg) was kept at 170◦ C as compared to
190◦ C for the production of the A-4 sized plates. Lowering the temper-
ature of the oligomer-catalyst mixture, decreases the reaction speed and
thus the viscosity of the entering liquid. In order to complete polymeri-
sation in a reasonable time-frame and to optimise the matrix properties,
the mould temperature was still kept at 190◦ C. Stirring time for the cat-
alyst/oligomer mixture was 45 s. The resin mixture was injected into
A.4 Production 173
mechanical
stirrer
T control
N2 inlet
manometer
heating Pt100 shell

mantle
Figure A.2: Leaf spring injection set-up. Only the bottom part
of the vessel was heated and overpressure is realised by N2 . Cata-
lyst and resin are mixed by a mechanical stirrer and the temper-
ature is controlled with a PT100
braided
preform
outlet
distribution
channel
outlet
inlet
Figure A.3: Leaf spring mould set-up. A distribution channel

along the length of the leaf spring facilitates mould filling
Table A.1: Processing parameters for leaf spring production

with thermoset and thermoplastic matrix
Epoxy1 Epoxy2 pCBT1 pCBT2
mould rotation 0◦ 0◦ 0◦ 180◦
distribution channel empty empty empty R
overflow time (s) 75 90 60 120
filling time (s) 600 900 300 300
pCBT3 pCBT4 pCBT5 pCBT6
mould rotation 180◦ 180◦ 90◦ 90◦
distribution channel R R R/S R/S
overflow time (s) 100 N.A. 90 N.A.
filling time (s) 300 375 420 270
R: roving, S: 5 cm of sealing at the ends of the channel
the mould with an overpressure of maximum 2.5 bar and assisted by an

underpressure of 0.8 bar.
Table A.1 shows the processing parameters for the production of both
thermoset and thermoplastic leaf springs. From the production of these
prototypes, it became clear that the process is not yet very reproducible.
Both the time to reach the first overflow and the time to reach air-free
overflow (filling time) changes although other parameters are kept con-
stant. There was no major difference in the irreproducibility between the
very low viscous oligomers and the epoxy resin. Even though there are a
number of factors that have to be taken into account, such as the pressure
profile, the fibre preform plays a major role.
The preform was braided around a mandrel with the inner section of
the mandrel being 60 mm wide. This leads to a preform width of around
70 mm, which is substantially larger than the 66.5 mm of the actual leaf
spring and hence of the mould cavity. The needed compression of the
braided preform to fit into the mould cavity causes an increase in fibre
volume fraction and hence a decrease in transverse permeability of 22%,
calculated with the formulas of Gebart [183]. The amount of preform
compression changed from one production run to another. Sometimes
fibres from the outer braid were trapped inside the injection line and/or
distribution channel, therefore altering the transverse permeability and
thus causing process irreproducibility.
Another effect of the need to compress the fibre preform in the middle
section, is the impossibility to closely fit the preform to the mould cavity
in the widening region of the leaf spring. This leads to gaps exceeding
3 mm between the fibre preform and the mould cavity. In combination
with a non-perfect fit of the preform at the outer sides of the leaf spring,
short cut flows are almost unavoidable. These short cut flows are respon-
A.5 Conclusion 175
sible for the sometimes very sudden overflow. It is therefore crucial to

produce a preform, which fits perfectly inside the mould cavity without
major handling. These edge effects are however well known in the liquid
moulding industry and not inherent to the in-situ polymerisation process.
Figure A.4: Thermoset (dark) and thermoplastic (white) leaf

springs
The time for this resin-catalyst combination to complete polymerisa-

tion and crystallisation is only 30 minutes, which, in an automated pro-
cess, can be the total cycle time, since the mould does not need to cool
down before demoulding. For this prototype set-up however, a minimum
cycle time of three hours was needed due to thermal cycling of the mould,
preform preparation excluded. The prototype is shown in Figure A.4 and
weighed 2.9 kg after proper trimming.
Although the time to first overflow and the total filling time for the
different production runs differ significantly, the final weight and hence
matrix content did not. The impregnation quality was however not inves-
tigated.
A.5 Conclusion
The production of this thermoplastic leaf spring prototype illustrates the
possibility to scale-up the liquid moulding process for thermoplastic resins.
Although optimisation of the braided preform is necessary to improve
process reproducibility, the encountered problems were inherent to the
RTM process and not related to the resin system.
Appendix B
Principles of X-ray
scattering
B.1 Introduction
X-rays are a high-energy electromagnetic radiation and have a range of
wavelengths between 0.1 and 100 Å. In the electromagnetic spectrum they
are bound by ultraviolet light at the long wavelength side and by gamma
radiation at the short wavelength side.
The production of X-rays is associated with the loss of energy of elec-
trons. Conventional X-ray sources generate X-rays by accelerating elec-
trons by an electric field after which they are slowed down by multiple
collisions with a metal target material. During these collisions both white
radiation and characteristic radiation, which depends on the target ma-
terial, occur. For X-ray scattering studies, a monochromatic beam is pre-
ferred in order to avoid an unwanted background. Suitable filters must
hence be chosen to single out the desired characteristic line of the target
material.
B.2 X-ray Scattering

The interaction of X-rays and matter leads to both absorption and scat-
tering of the X-rays. Scattering is defined as the deflection of an X-ray
photon from its original path by collision with an electron or an atom
[155]. If this collision is not accompanied by a loss of energy, the scatter-
ing is called elastic scattering. The elastic scattered radiation retains the
wavelength of the incident beam, the scattering is hence called coherent.
It is this scattering that can give rise to systematic interference effects.
When, however, the wavelength is changed after the collision (taking place
177
178 Principles of X-ray scattering
q nl = 2dsinq
q
d
Figure B.1: Diffraction of X-rays by a crystal, illustrating

Bragg’s law
with loss of energy), incoherent scattering, which gives rise to a continuous

background, arises.
X-ray scattering and diffraction are often used interchangeably, how-
ever, diffraction is preferred when the specimen under consideration has
a regularity in its structure as for example in crystals. The diffraction
of X-rays by crystals was discovered by Max von Laue in 1912 who de-
scribed the phenomenon in terms of diffraction from a three-dimensional
grating. In the same year, W.L. Bragg noticed the similarity of diffraction
to ordinary reflection and deduced a simple equation treating diffraction
as reflection from lattice-planes.
Bragg considered the crystal to be constructed of different equidistant
planes. X- rays impinge on the crystal planes under an angle θ with respect
to the surface. The condition of positive interference is fulfilled when the
incoming X-rays leave the sample under the same angle θ, as shown in
Figure B.1, and when the path difference between X-rays diffracted from
two different planes is equal to a multiple of the X-ray wavelength. This
results in Bragg’s law:
nλ = 2dsinθ (B.1)
in which λ is the X-ray wavelength, d the interplanar distance and n a
positive integer number representing the order of the reflection.
There are two major techniques of X-ray scattering that are often used
in polymer science, namely Wide-Angle X-ray Diffraction (WAXD) and
Small-Angle X-ray Scattering (SAXS) both corresponding to a different
level of structural organisation in polymers. Whereas WAXD is concerned
with scattering phenomena occurring at angles above 3◦ and is hence re-
lated to atomic distances as they occur in the unit cell of polymer crystals
or in the ‘packing’ of a liquid-like or glassy material, SAXS deals with
scattering at angles below 3◦ providing information about structures at a
nanometer level such as alternating crystalline-amorphous layer stacking.
Going outside the scope of X-ray wavelengths, it is also possible to in-
vestigate the spherulitic superstructure of polymers by Small-Angle Light
Scattering (SALS) which also involves scattering by interaction with the
electron clouds.
B.3 Wide-angle X-ray Diffraction 179
Figure B.2: Typical WAXD patterns of a sample consisting of

(a) large crystals, (b) small crystals and (c) a sample in the amor-
phous state, adapted from [102]
B.3 Wide-angle X-ray Diffraction

Wide-angle X-ray diffraction is used to investigate structures of 1 nm or
smaller. The size of the regular domains determines the width of the
diffraction peaks as illustrated in Figure B.2. Even when there is no
apparent order present in the sample, like for example in a polymer melt,
the diffraction pattern does not completely disappear. For amorphous
materials there is still local order, in the sense that there is a reasonably
well-defined average number of nearest neighbours and scattering from
these gives rise to the amorphous halo in Figure B.2(c) [102].
For most non-polymeric materials, either the amorphous scattering
or the crystalline diffraction is present. Since polymers are, however,
semi-crystalline, the crystalline diffraction peaks are superimposed on the
amorphous halo. By comparing the intensity scattered from the crystals
(diffraction peaks) to the total scattered intensity, the degree of crys-
tallinity can be determined. This requires, however, knowledge about the
shape of the amorphous halo or well-separated diffraction peaks between
which the amorphous halo can be interpolated.
B.4 Small-angle X-ray Scattering

Structures with dimensions between 1 and 1000 nm can be investigated
using small-angle X-ray scattering. For semi-crystalline polymers, this
range coincides with the stacking of crystalline and amorphous layers.
In order to process the experimentally obtained intensity profiles, linear
correlation functions or interface distribution functions can be calculated.
B.4.1 Correlation function analysis [180, 184]

The scattering behaviour of a semi-crystalline polymer can be related to
the electron density distribution %(r) measured along a trajectory normal
to the inner surfaces of a stack. This trajectory will pass through amor-
phous regions with density %a and crystallites with a density %c in the
core. The average density within a stack lies between these two limits
according to the local crystallinity and will be called h%i. The correlation
function, K(x), Equation B.2, is the average of the product of the electron
density differences at two points a distance x apart and is a measure of the
probability of finding the density the same at these two points, showing
a peak at the most probable repeat distance, namely the long period Lp
of a two-phase structure.
K(x) = h[%(x0 ) − h%i] [%(x0 + x) − h%i]i (B.2)
Linear correlation functions are calculated by a cosine transformation

of the background corrected intensity profile I(s), Equation B.3. Since
data are not recorded over an infinite s-range, extrapolation of I(s) both
to low and high angles is necessary to calculate the correlation function.
Z∞
K(x) = I(s)s2 cos(2πxs)ds (B.3)
0
The periodic structure of the semi-crystalline polymer determines the

shape of the correlation function. The effect of stepwise perturbations
of the ideal two-phase structure on the correlation function is shown in
Figure B.3. Throughout all perturbations, a central linear section in the
auto-correlation region of K(x) remains unchanged. By extrapolating this
linear section, it is possible to derive the parameters of the ‘corresponding
ideal two-phase system’ [180].
The value of the correlation function at x = 0 is often referred to as
the total scattering power or invariant and is defined in Equation B.4.
Z∞
Q = K(0) = I(s)s2 ds (B.4)
0
According to Strobl and Schneider [184], the invariant of the correspond-

ing ideal two-phase system, Qid , equals the intercept of the linear regres-
sion in the auto-correlation triangle (LRAT).
The invariant of the (ideal) two-phase system can be used to normalise
the correlation function resulting in CF (x) and γ(x), Equation B.5, B.6
B.4 Small-angle X-ray Scattering 181
Figure B.3: Electron density distribution %(x) and related cor-

relation function K(x) for lamellar systems of different regularity.
(a) Periodic two-phase system. (b) Effect of long-spacing varia-
tions. (c) Effect of additional thickness fluctuations. (d) Effect of
introduction of diffuse phase boundaries. Adapted from [184]
Figure B.4: Normalised correlation function CF (x) indicating

the determination of the structural parameters
and Figure B.4.

K(x)
CF (x) = (B.5)
Qid
K(x)
γ(x) = (B.6)
Q
There are two routes to determine the local volume fractions of the
phases present in a semi-crystalline polymer (φ and (1−φ)). Firstly, φ can
be calculated from the value of the flat minimum of CF (x) which equals
−φ/(1 − φ). In this context, φ necessarily corresponds to the minority
fraction, irrespective of being crystalline or amorphous. In order to use
this value, the minimum has to be flat. This only occurs if the distance
distributions of the two characteristic lengths do not overlap. If so, the
length, associated with the minority fraction, can be determined by taking
the solution of LRAT at y = the value of the flat minimum.
Most often however, the minimum is not flat and only the second
method, based on the quadratic expression (q.e.), is viable to determine
φ, Equation B.7.
2
LRATy=o = A = φ(1 − φ) = φ(1 − φ)Lp (B.7)
OS
with A the solution of the LRAT at the value of y = 0, Figure B.4 and Os
the specific surface area of the phase boundary. By replacing O2S by Lp ,
the assumption is made that the lamellar crystals have infinite lateral
ρc
ρs
ρa
Figure B.5: Schematic representation of a heterogeneous system
consisting of partially crystalline regions dispersed in an amor-
phous matrix
dimensions. After solving the quadratic expression, the characteristic

lengths can be then determined by l1 = φLp and l2 = (1−φ)Lp = Lp −l1 1 .
The correlation function analysis is a helpful tool to determine the
structural parameters of a semi-crystalline polymer. According to Babi-
net’s principle however, it is impossible to decide wether φ or (1 − φ)
and hence l1 or l2 corresponds to the crystalline or the amorphous phase.
Therefore, independent information about the crystalline structure is needed
to make the correct assignment. Moreover, near φ = 0.5, evaluation of
scattering curves is inherently difficult and insensitive because the scatter-
ing behaviour changes very slowly. No evaluation procedure can remove
this basic difficulty and SAXS measurements then need to be comple-
mented by other data.
Physical meaning of the invariant

For homogeneous systems, meaning that the irradiated volume is densely
filled with stacks, the invariant of an ideal two-phase systems is correlated
to the electron density difference according to Equation B.8 [184]:
Qid = C 0 ϕ(1 − ϕ)(%c − %a )2 (B.8)
with ϕ and (1 − ϕ) the overall volume fraction of the crystalline and

amorphous phase which for a homogeneous system equals the local volume
fractions φ and (1 − φ), (%c − %a ) the temperature dependent electron
density difference between the crystalline and amorphous phase, C 0 a
constant accounting for the relative scale of the measured intensities.
Heterogeneous systems, however, are often encountered e.g. during
the investigation of crystallisation kinetics where, at intermediate stages,
a sample is only partially filled with stacks, illustrated in Figure B.5. In
this case, the local degree of crystallinity is larger than the overall degree
of crystallinity by a factor αs , which corresponds to the fraction of semi-
crystalline stacks.
1 or l = φL and l = (1 − φ)L = L − l as φ is mostly considered to be the
2 p 1 p p 2
crystalline fraction and l1 < l2
The scattering pattern of such a structure can be roughly described

as a superposition of two curves, a scattering curve Is that is related
to the internal structure of the stack and a second curve Isa , reflecting
the supermorphology formed by the dispersion of semi-crystalline stacks
inside an amorphous matrix. Even though the invariant only depends on
the mean-square fluctuation of the electron density, it may be affected
by instrument resolution [180]. The first component of the scattering
curve, Is will always be observed, whereas the second part can become
unobservable if the dimensions of the structural elements are large, hence
scattering at very low angles.
If the complete scattering curve is recorded, Equation B.8 remains
valid. In the other extreme case, namely if Isa is not recorded at all, the
correlation function is proportional to αs and the invariant is given by
Equation B.9 [184]:
Qid = Cαs φ(1 − φ)(%c − %a )2 (B.9)
The same equation was derived by Goderis et al. [180], starting from
Equation B.8 and assuming not only large distances scattering at small
angles but also that the lower contrast factor, (%s − %a ), leads to an inten-
sity that is low comparable to that due to the partially crystalline sub-
regions. Whereas Strobl and Schneider concluded that the interpretation
of the invariant becomes difficult if not impossible when heterogeneities
cannot be excluded (e.g. at the intermediate stages of crystallisation),
Goderis et al. were able to use Equation B.9 during the melting of linear
polyethylenes.
B.4.2 Interface distribution function [163, 185]

Although used less often, the characteristic lengths of the ideal two-
phase system can also be calculated by the interface distribution function,
IDF (x). This function is the second derivative of the normalised correla-
tion function, γ(x) and represents the probability distribution of finding
two interfaces at a distance x.
The interface distribution function can also be calculated by a cosine
transformation of the interference function, IF F (s), which can be deter-
mined from the measured intensity profile, Equation B.10.
IF F (s) = (I(s) − B)s4 − P (B.10)
where the constants B and P can be determined by linear regression of

a Porod plot (I(s)s4 versus s4 ). The slope of this plot yields B, whereas
the intercept equals P . An additional constraint can be used to more
accurately determine both constants, Equation B.11. The interface dis-
Figure B.6: Shape of interface distribution function when length

distributions do not overlap
tribution function is then calculated using Equation B.12.

Z∞
IF F (s)ds = 0 (B.11)
0
Z∞
IDF (x) = − IF F (s) cos(2πxs)ds (B.12)
0
Figure B.6 represents a typical IDF(x) function when the distributions

of the characteristic lengths do not overlap and the most probable lengths
can be determined by taking the maximum of each peak. Note that the
correlation function approach yields ‘number average’, rather than ‘most
probable’ lengths.
With the interface distribution function, it is however not possible to
assign these lengths to either the amorphous or crystalline phase. More-
over, since the length distributions usually overlap, only one maximum is
visible, which is assigned to the length with the most narrow distribution,
often considered to be the crystalline phase.
Appendix C
Thermoplastic
Polyurethanes
C.1 Introduction
The main disadvantage when processing thermoplastic composites is their
high melt viscosity. One route to overcome this issue is in-situ polymerisa-
tion where prepolymers impregnate the fibrous reinforcement after which
they are polymerised. This route was extensively described in the main
part of this thesis. At the start of this research however, another polymer
system, namely a thermoplastic polyurethane, was investigated because
of its capability to undergo a reversible depolymerisation upon melting,
hence lowering its viscosity drastically. This appendix will describe some
of the characteristics of this polymer system.
C.2 Thermoplastic polyuretane (TPU)

The general term polyurethane has become known for a multitude of
polymers prepared according to the diisocyanate-polyaddition principle.
The characteristic structural element of almost all of these polymers is the
urethane group formed in the course of the polyaddition, Equation C.1.
The characteristic functional group originates from the two basic building
blocks of polyurethanes, isocyanates and polyols.
(C.1)
Thermoplastic polyurethanes are polyurethane elastomers with a ther-

moplastic character. Generally, they can be considered a linear segmented
187
188 Thermoplastic Polyurethanes
block copolymer. The linearity is obtained from using bifunctional raw

materials and avoiding secondary reactions. The rigid block consists of
repeating groups of diisocyanate and a short-chain diol (or chain exten-
der). The flexible block consists of repeating groups of diisocyanate and
a long-chain diol (or polyol).
A very important side-reaction is the reaction of isocyanates with wa-
ter (Equation C.2), which is essential for the formation of polyurethane
foam. In this reaction, the primary addition product is carbamic acid.
Since carbamic acid is not stable, it splits off carbon dioxide, forming the
corresponding amine. The carbon dioxide, which is formed, takes over the
role of a blowing agent, which leads to the formation of the macromolec-
ular skeleton.
(C.2)
The thermoplastic polyurethane used in this study is commercially

available from The Dow Chemical Company under the tradename of
IsoplastTM . The clear grade of IsoplastTM , with almost 90% of light trans-
mission and minimal yellowness has a level of clarity comparable to poly-
carbonate and amorphous nylon and is therefore ideal for container- or
bowl-type applications where fluid levels need to be inspected or working
parts need to be viewed. Processing of IsoplastTM is enhanced by the low
melt viscosities attained by the reversible depolymerisation.
In 1999, a U.S.patent was granted to The Dow Chemical Company for
the production of fibre reinforced composites prepared from depolymeris-
able and repolymerisable polymer by a pultrusion process [25]. In this
patent, the IsoplastTM resins were mentioned as suitable resins for this
process. Since then, a spin-off company was founded, Fulcrum Compos-
ites Inc., which further developed the composite applications of IsoplastTM
under the tradename FulcrumTM . These developments however focus on
one production process, namely pultrusion. In this short study, however,
the feasibility of using this particular TPU in resin transfer moulding was
investigated.
C.3 Reversible (de)-polymerisation

The thermodynamics of a system can be described by the Gibbs free en-
ergy. A reaction is only possible if the Gibbs free energy of the system
decreases (∆G < 0) during this reaction. Equation C.3 shows the rela-
tionship between the Gibbs free energy change, enthalpy change, ∆H,
and entropy change, ∆S, with T the absolute temperature:
∆Gp = ∆Hp − T · ∆Sp (C.3)

C.4 Characterisation 189
For polymerisation, the entropy change is almost always negative.

Therefore, almost all polymerisation reactions are exothermal. If the
temperature rises, the entropy term becomes more important and at a
certain temperature, (T · ∆Sp ) will equal ∆Hp . At higher temperatures,
the Gibbs free energy change will be positive. As a result, polymerisation
is impossible, and moreover, depolymerisation will occur. The tempera-
ture at which ∆Gp reaches zero is called the ceiling temperature, Tc .
∆Gp = 0 = ∆Hp − Tc · ∆Sp (C.4)

∆Hp
Tc = (C.5)
∆Sp
For most polymer systems (but not all), the degradation temperature
is lower than the ceiling temperature. Therefore, most systems do not
exhibit a thermodynamically driven depolymerisation. In general, ther-
moplastic polyurethanes possess a ceiling temperature below their degra-
dation temperature and this ceiling temperature can be easily altered via
a change in the type of constituents, Table 1. Experiments show that the
depolymerisation reaction of TPU’s is completely reversible and equilib-
rium is attained almost instantaneously [186].
Table C.1: Ceiling temperatures for different types of poly-

urethanes [186]
Compound Tc (◦ C)
Alkyl–NH–COO–Alkyl 250
Aryl–NH–COO–Alkyl 200
Alkyl–NH–COO–Aryl 180
Aryl–NH–COO-Aryl 120
C.4 Characterisation
C.4.1 Molecular Structure
The molecular structure of the TPU was investigated with both NMR
spectroscopy (Nuclear Magnetic Resonance spectroscopy) and FT-IR spec-
troscopy (Fourier Transform Infrared spectroscospy)1 . The NMR spec-
troscopy analysis involved the dilution of 40 mg of sample in 0.7 ml of
dimethylsulfoxide (d6). In order to make the dilution, the sample was
heated up to 70◦ C. The FT-IR analysis involved the preparation of a KBr
dispersion from the grinded specimen.
1 measurements were performed at the Chemistry department of the Science Faculty
of the K.U.Leuven
The NMR spectra show that the TPU results from the reaction of MDI
(methylene diphenyl isocyanate), Equation C.6; with two chain extenders,
1,4-butanediol (60 mol%), Equation C.7 and ethylene glycol (40 mol%),
Equation C.8.
(C.6)
(C.7)
(C.8)
This structure was confirmed by FT-IR analysis. Both diols are con-
sidered chain extenders because of their low molecular weight, therefore
being part of the hard segment. From these analysis, it can be concluded
that this TPU does not contain a soft segment. By comparing this chem-
ical structure with those listed in Table C.1, the ceiling temperature will
be around 180-200◦ C, which is below the temperature range in which the
polymer starts to flow (flow temperature).
Since we are dealing with a reactive system, molecular weight rebuild-
up after cooling is important. To examine this rebuild after processing,
small tensile bars l = 90 mm were produced with a miniature injection
moulding machine2 . The temperature of the melt reservoir was kept at a
constant temperature of 240◦ C, whereas the mould was at room temper-
ature.
Molecular weight and molecular weight distribution were determined
by GPC (Gel Permeation Chromotography)3 . The TPU was dissolved in
tetrahydrofurane (THF) and both UV and refractive index detectors were
used to determine Mw . Table C.2 lists the average molecular weight as
well as the polydispersity index.
Although it is claimed that the depolymerisation is completely re-
versible, this is not observed. First of all, it is known that due to shearing,
the molecular weight decreases in injection moulding. In our case how-
ever, this is unlikely to be the primary cause since shearing takes place
before the repolymerisation. The high cooling rate in injection mould-
ing can however influence the repolymerisation time. Since the TPU is
cooled down almost immediately, insufficient time is available to complete
the repolymerisation.
2 available at the Chemical Engineering department of the Engineering Faculty of
the K.U.Leuven
3 measurements were performed at the Chemistry department of the Science Faculty
of the K.U.Leuven
C.4 Characterisation 191
Table C.2: Molecular weight of TPU

as received injection moulded
Mn (kg/mole) 75.6 40.3
Mw (kg/mole) 189.0 74.5
Mw /Mn (-) 2.5 1.9
Figure C.1: DSC thermogram of TPU
C.4.2 Thermal Analysis

DSC (Differential Scanning Calorimetry) measurements were performed
at a heating rate of 5◦ C/min, with a temperature range from 50 to 350◦ C
and a sample weight of around 5 mg. TGA (Thermogravimetrical Anal-
ysis) measurements were also performed at a heating rate of 5◦ C/min,
with a temperature range from room temperature to 450◦ C and a sample
weight of around 5 mg. Isothermal measurements were also performed at
200 and 250◦ C.
Figure C.1 shows the result of the DSC measurements. The first small
peak shows the evaporation of the absorbed water. Since the TPU is
amorphous, there is no melting peak visible. The peak that starts at
around 280◦ C is the foaming peak. At high temperatures, the material
foams due to volatising diols. On top of that, the TPU reacts with the
surrounding moisture, which cannot be prevented by drying. The same
onset of foaming was seen in the TGA ramp measurement. The isothermal
TGA experiments showed that at 200◦ C there was no weight loss over a
period of 60 minutes and the sample only showed minor foaming. At
250◦ C however, the weight loss was constant with time and going up to
almost 20% after 60 minutes. This sample showed significant foaming.

From these measurements, it is clear that foaming and its prevention is
a major issue when processing a TPU. Especially when high temperatures
need to be kept for a substantial amount of time, foaming is very likely to
occur. Even with extensive drying of the polymer, the fibre reinforcement
and the processing environment need to be moisture free.
C.4.3 Viscosity
In order to use a TPU in an RTM-like process, the viscosity during pro-
cessing needs to be below 1 Pa·s. A rather large viscosity drop in the melt
is however expected since, above the ceiling temperature, a reversible de-
polymerisation takes place.
The viscosity was measured with a Dynamic Stress Rheometer at
1 rad/s4 . The sample was kept under an inert N2 -atmosphere. The tem-
perature ranged from 220 to 280◦ C at a heating rate of 5◦ C/min. Isother-
mal viscosity measurements were performed on a Capillary Rheometer, a
30 × 1 mm2 die was used and temperature ranged from 220 to 260◦ C.
Figure C.2 shows the results of the dynamic measurements. Although
the rheometer was kept under an inert N2 -atmosphere, foaming occurred
at the edges. Due to this foaming, the repolymerisation and thus the
viscosity rise was not complete during cooling, but still a clear trend is
seen. The viscosity drops more than two decades when the temperature
is raised only 60◦ C.
In order to avoid free edges where foaming can occur, isothermal mea-
surements were performed with a capillary rheometer, Figure C.3. No
corrections (Bagley plots, Weissenberg-Rabinowitz) were made to the raw
data.
Although there is a substantial drop in viscosity above the flow tem-
perature, related to the reversible depolymerisation, the lowest viscosity
measured is around 10 Pa·s, which is well above the upper limit for use
in RTM.
C.5 Conclusion
The feasibility to use the thermoplastic polyurethane, IsoplastTM , in an
RTM-like process, was investigated. Its chemical structure, behaviour un-
der elevated temperatures and viscosity were determined. From this short
study, it is clear that the resin as such is not very suitable for use in RTM.
Even though there is a substantial viscosity drop due to the reversible de-
polymerisation, the lowest measured viscosity is 10 Pa·s, which is an order
of magnitude too high. Further increasing the temperature to decrease
4 measurements were performed at the Chemical Engineering department of the
Engineering Faculty of the K.U.Leuven

C.5 Conclusion 193
Figure C.2: Dynamic viscosity measurement showing the tem-

perature dependence of the complex viscosity
Figure C.3: Viscosity versus temperature for different shear

rates resulting from capillary viscosity measurements
the viscosity even more is not an option since foaming due to volatising
diols cannot be avoided anymore at temperatures above 280◦ C.
On top of the viscosity barrier, processing should take place in an inert
atmosphere, free from all moisture since moisture also induces foaming.
Due to the above mentioned constraints, it was decided to further focuss
the research on the CBTr resin.
Nederlandstalige
samenvatting
Polymeermatrixcomposieten zijn een interessante materiaalklasse omdat

zij goede mechanische eigenschappen combineren met een lage densiteit.
Onafhankelijk van de versterking zijn er twee grote klassen van matrices,
namelijk thermoplasten en thermoharders. Omwille van de hoge smeltvis-
cositeit van thermoplasten is het moeilijk om de vezelversterking te im-
pregneren. Vandaar dat continu vezelversterkte thermoplasten slechts een
klein deel van de huidige composietmarkt innemen ondanks hun verhoogde
impactweerstand en recyclagemogelijkheden.
De hoge smeltviscositeit heeft bovendien tot gevolg dat de produc-
tiemethodes voor thermoplastische composieten sterk verschillen van ther-
mohardende composieten. Bij gemiddelde drukken zijn er twee routes om
continu vezelversterkte thermoplasten te produceren: (1) het verlagen van
de vloei-afstand van het hars en (2) het verlagen van de viscositeit tijdens
de impregnatie, Figuur 1.
In dit werk zal de viscositeit verlaagd worden door in-situ polymerisatie
van cyclische oligomeren, CBTr hars. De composieten kunnen dan gepro-
duceerd worden met een productiemethode typisch voor thermoharders,
namelijk RTM1 .
Situering van het onderzoek

Het voornaamste doel van in-situ polymerisatie van thermoplasten is de
rechtstreekse impregnatie van de vezelversterking zonder bijkomende tus-
senstappen. Er moet echter aan een aantal voorwaarden voldaan worden
om de versterking rechtstreeks te impregneren door een vloeibare matrix
[14, 15]:
• De viscositeit van de matrix tijdens de impregnatie is zeer laag,

<1 Pa·s.
1 Resin Transfer Moulding
195
196 Nederlandstalige samenvatting
Figuur 1: Productie-routes voor thermoplastische composieten

met een continue versterking
• Zodra de versterking volledig gempregneerd is, moet de matrix vast

worden in een voldoende korte tijd.
• De fysische eigenschappen van de matrix volstaan om de mecha-

nische stabiliteit van het composiet te waarborgen.
Indien het productieproces plaatsvindt in een gesloten mal zijn er bij-

komende vereisten:
• De vorming van bijproducten tijdens de polymerisatiereactie moet

vermeden worden [17].
• De reactie-exotherm is best zo klein mogelijk om variaties in eigen-

schappen door de dikte te vermijden [44].
Het is duidelijk dat slechts een bepaald aantal polymeersystemen voldoen

aan deze voorwaarden. Het meeste onderzoek tot hiertoe heeft zich toege-
spitst op de anionische ring-openende polymerisatie van polyamides. Door
de nakende commercialisatie van cyclische oligoesters, zoals CBTr hars,
is er echter een hernieuwde aandacht ontstaan voor in-situ polymerisatie
van thermoplasten.
Nederlandstalige samenvatting 197
Figuur 2: Aspecten van het procesraam voor CBTr hars, (links)

impregnatie- en polymerisatietijd voor CBTr hars aangepast uit
[44], (rechts) viscositeitsevolutie voor CBTr hars boven en onder
het polymeer smeltpunt, aangepast uit [89]
Het procesraam voor de in-situ polymerisate van thermoplasten is zeer

belangrijk aangezien er gebruik gemaakt wordt van een reactief polymeer-
systeem. Een eerste aspect van dit procesraam is het tijdsraam voor im-
pregnatie. De impregnatietijd wordt gedefinieerd als de tijd waarop het
reactief mengsel een viscositeit van 1 Pa·s bereikt. De grootte van dit raam
hangt sterk af van de reactiekinetica van het gebruikte polymeersysteem
en wordt liefst niet te nauw gekozen.
Anderzijds moet natuurlijk rekening gehouden worden met de totale
productietijd die liefst zo kort mogelijk gehouden wordt. De totale pro-
ductietijd is niet alleen afhankelijk van de polymerisatietijd, maar ook
van de kristallisatietijd, het uiteindelijke composiet moet immers vast
zijn. Een compromis in reactiesnelheid is dus noodzakelijk om enerzijds
de impregnatietijd te vergroten en anderzijds de totale productietijd te
beperken.
De linkerzijde van Figuur 2 toont, voor het gebruikte CBTr hars, de
afhankelijkheid van de impregnatie- en polymerisatietijd van de temper-
atuur. Eén van de belangrijke voordelen van CBTr hars, ten opzichte
van de polyamides, is de mogelijkheid tot isotherm produceren, rechterz-
ijde Figuur 2. Hierbij kunnen polymerisatie en kristallisatie plaatsvinden
bij dezelfde temperatuur zonder een onaanvaardbare verhoging van de
polymerisatie-tijd.
Tot hiertoe heeft onderzoek op thermoplastisch-RTM (TP-RTM) zich
voornamelijk toegespitst op de karakterisatie van de onversterkte poly-
meersystemen met betrekking tot het procesraam en op de mechanische
eigenschappen van de composieten en dan voornamelijk voor de poly-
amides. Hoewel de mogelijkheid voor de productie van composieten met
CBTr hars vaak vermeld wordt, is het onderzoek naar de relatie tussen
de productie en de uiteindelijke eigenschappen erg beperkt gebleven.
Een eerste stap in het huidig onderzoek was het opzetten van een
productie-eenheid voor TP-RTM op laboschaal. Uitgaande van CBTr
hars wordt vervolgens glasvezelversterkt PBT (poly(butyleen tereftalaat))
geproduceerd. Aangezien de matrix gepolymeriseerd wordt tijdens het
productieproces, wordt er aandacht besteed aan de invloed van het pro-
ductieproces op de matrixeigenschappen van de composieten. Hiernaast
worden natuurlijk ook de mechanische eigenschappen van de composieten
bepaald.
Vervolgens wordt de invloed van het isotherme karakter van het pro-
ductieproces op de morfologie en eigenschappen van pCBT bestudeerd.
De laatste fase van dit onderzoek bestaat uit een meer fundamentele studie
van het kristallisatieproces en de invloed van de simultane polymerisatie
en kristallisatie.
Productie en eigenschappen van glasvezelver-

sterkt pCBT geproduceerd met TP-RTM
Productie
Het tijdsraam voor impregnatie tijdens de productie van glasvezelversterkt
pCBT met TP-RTM is relatief nauw. De mengtijd van katalysator en hars
evenals de harstemperatuur hebben een significante invloed op de vultijd.
Om het benodigde en beschikbare tijdsraam voor impregnatie op elkaar af
te stemmen, is het noodzakelijk om voor ieder individueel product een
gepaste injectiestrategie te kiezen. Vulsimulaties die de veranderende vis-
cositeit en de capillaire effecten in rekening brengen zijn daarvoor noodza-
kelijk. Deze simulaties zijn echter niet het onderwerp van deze studie.
Matrixeigenschappen
De eigenschappen van de matrix worden beı̈nvloed door het in-situ poly-
merisatie proces. De conversiegraad en het moleculair gewicht zijn lager
wanneer glasvezels aanwezig zijn tijdens de polymerisatie. De conver-
siegraad wordt echter onderschat door de sizing2 aanwezig op de glasve-
zels, Figuur 3. De GPC-resultaten tonen eveneens de nevenreactie tussen
de epoxide functionele groepen, aanwezig in de sizing, en pCBT. Hierdoor
wordt pCBT gedeeltelijk vernet.
Ook de kristallisatiegraad wordt onderzocht. Vergeleken met PBT,
snel afgekoeld vanuit de smelt, is de kristallisatiegraad van pCBT signifi-
cant hoger. Dit kan verklaard worden door de relatief lage onderkoeling bij
de procestemperatuur. Het effect van de glasvezels op de kristallisatiegraad
is dan weer sterk afhankelijk van de hoeveelheid vezels (en sizing) evenals
2 deklaagje aanwezig op de glasvezels om deze te beschermen tijdens de verwerking
en om het grensvlak tussen vezel en matrix te verbeteren

Figuur 3: Typisch GPC-chromatogram voor pCBT, glasvezel-

versterkt pCBT en de glasvezelsizing
van het contact tussen vezels en matrix tijdens de polymerisatie aange-

zien beide de hoeveelheid vernet pCBT beı̈nvloeden. Dit vernet pCBT
kan niet kristalliseren en verlaagt dus de kristallisatiegraad.
Een overzicht van de matrixeigenschappen in de composieten kan gevon-
den worden in Paragraaf 5.2.3, p 81 van de Engelse tekst.
Composieteigenschappen
De mechanische eigenschappen van de composieten zijn bepaald en ver-
geleken met de theoretische voorspellingen. Zoals verwacht hebben de
matrixeigenschappen geen effect op de vezelgedomineerde composietei-
genschappen zoals de longitudinale stijfheid en sterkte. Deze komen dan
ook goed overeen met de theoretische waarden. De transversale eigen-
schappen van unidirectionele composieten worden echter bepaald door de
matrix- en grensvlakeigenschappen.
Figuur 4 toont de invloed van de matrix- en grensvlakeigenschappen
op de composieteigenschappen. De invloed van de matrixeigenschappen
op de transversale modulus is beperkt. Zodra de conversiegraad en het
moleculair gewicht hoger zijn dan een kritische waarde, is hun effect op
de matrixmodulus immers zeer klein.
De grensvlakeigenschappen hebben daarentegen een niet te verwaar-
lozen effect op de transversale modulus. De grote daling in modulus wan-
neer de sizing verwijderd is, doet vermoeden dat in dat geval, de meeste
glasvezels niet hechten aan de matrix. Hierdoor dragen ze niet bij tot de
transversale modulus. De gemeten modulus is dus lager dan de matrix-
modulus omdat de matrix verzwakt is door ‘holtes’.
De matrixsterkte aan de andere kant wordt sterker beı̈nvloed door een
Figuur 4: Verband tussen matrix- en grensvlakeigenschappen

en mechanische eigenschappen van unidirectionele composieten
(links) transversale buigmodulus en (rechts) ‘werkelijke’ transver-
sale buigsterkte
lagere conversie en een lager moleculair gewicht. De transversale sterkte

daalt hierdoor meer dan 20%. Een verzwakking van het grensvlak zorgt
bovendien voor een verdere daling van de transversale sterkte.
De transversale sterkte van de unidirectionele composieten blijkt boven-
dien lager te zijn dan de matrixsterkte. Voor thermhardende composieten
is dit een vaak voorkomend fenomeen omwille van de spanningsconcen-
traties rond de vezels. Plastische vervorming van de thermoplastische ma-
trix zou echter deze spanningsconcentraties moeten kunnen compenseren.
Interlaminaire breuktaaiheidstesten en valimpacttesten werden uit-
gevoerd op pCBT versterkt met een glasvezel weefsel. De resultaten
werden vergeleken met epoxy- en PBT-composieten. Deze PBT com-
posieten werden snel afgekoeld vanuit de smelt. Ondanks het thermoplas-
tische karakter van pCBT, is zowel de interlaminaire breuktaaiheid als
de impactweerstand van de pCBT composieten beter vergelijkbaar met
deze van de thermohardende epoxy-composieten. Deze resultaten doen
vermoeden dat de pCBT matrix zich niet als een taaie thermoplastische
matrix gedraagt.
Een overzicht van de mechanische eigenschappen van de composieten

kan gevonden worden in Paragraaf 5.3.5, p 95 en in Paragraaf 5.4.3, p 102
van de Engelse tekst.
De huidige resultaten tonen duidelijk aan dat de gevolgde produc-

tieroute een effect heeft op de matrixeigenschappen en logischerwijze dus
ook op de composieteigenschappen die niet vezel-gedomineerd zijn. De
volgende fase van het onderzoek heeft zich dan ook toegespitst op de in-
vloed van het isotherm productieproces op de morfologie en mechanische
eigenschappen van de matrix.
Tabel 1: Mechanische eigenschappen van pCBT, geproduceerd

via verschillende routes
E (GPa) σ ∗a (MPa) ε∗a (MPa) taaiheidb (kJ/m2 )

pCBT190 3.2 ± 0.1 54 ± 5 1.7 ± 0.2 1.5 ± 0.2
pCBT230-190 3.1 ± 0.2 73 ± 14 2.3 ± 0.6 n.a.
pCBT240-S n.a. n.a. n.a. 2.5 ± 0.6
pCBT240-Q 2.4 ± 0.3 75 ± 6 3.5 ± 0.2 4.3 ± 0.8
PBT-Q 2.2 ± 0.1 67 ± 9 3.3 ± 0.1 8.2 ± 2.8
a Spanning en rek bij breuk voor pCBT190 en pCBT230-190, vloeispanning en -rek
voor pCBT240-Q en PBT-Q
b Charpy kerfslagwaarde
Eigenschappen en morfologie van pCBT

De mechanische composieteigenschappen zoals de transversale sterkte en
impactweerstand doen vermoeden dat de isotherm geproduceerde ma-
trix bros is. Om deze reden worden de eigenschappen van de onver-
sterkte pCBT matrix onderzocht. In een eerste stap wordt echter de
kristallisatiekinetica van CBTr hars bekeken.
Kristallisatiekinetica
De isotherme kristallisatiekinetica van CBTr hars werd bestudeerd met
behulp van DSC-metingen op verschillende temperaturen. Vooraleer de
kristallisatie kan starten, moet het moleculair gewicht van pCBT vol-
doende zijn toegenomen. Indien de polymerisatie- en kristallisatiestap
volledig gescheiden zouden zijn, zou de kristallisatiekinetica van CBTr
hars enkel een verschuiving in de tijd vertonen ten opzichte van isotherme
kristallisatie vanuit de smelt. Dit is echter niet het geval voor tempe-
raturen onder 200◦ C, waar de kristallisatie dus start tijdens de verdere
opbouw van het moleculair gewicht.
Kristallisatie tijdens de polymerisatie kan leiden tot een verschil in
kristalstructuur en morfologie [104] en dus een grote invloed hebben op
de finale polymeereigenschappen.
Mechanische eigenschappen
Tabel 1 toont de resultaten van 3-puntsbuig- en Charpy-testen. De bros-
heid van isotherm geproduceerd pCBT wordt duidelijk bevestigd door de
lage breukrek en Charpy kerfslagwaarde. De eigenschappen van pCBT
kunnen echter verbeterd worden door enerzijds polymerisatie en kristalli-
satie te scheiden (polymerisatie op hoge temperatuur) en anderzijds door
de afkoelsnelheid vanuit de smelt te vergroten.
De resultaten tonen duidelijk aan dat de brosheid van pCBT niet gere-
lateerd is aan het startmateriaal, CBTr hars, maar aan de gevolgde pro-
ductieroute. Herverwerken van de isotherm geproduceerde monsters leidt
dan ook tot een significante toename in taaiheid. Voor de productie van
composieten met RTM is een bijkomende verwerkingsstap echter geen
economisch-verantwoorde oplossing.
De vraag blijft echter wat de onderliggende oorzaak van de brosheid
is, met andere woorden, welke eigenschappen worden zo beı̈nvloed door
de isotherme productieroute dat ze leiden tot brosheid.
Fysische eigenschappen
Een eerste voorwaarde voor ductiliteit is een voldoende hoge conversie van
oligomeren naar polymeer, resulterend in een moleculair gewicht groter
dan het kritisch moleculair gewicht voor verstrengelingen, 50 kg/mol voor
PBT. Deze voorwaarde wordt vervuld door alle monsters.
Een hoge kristallisatiegraad kan ook leiden tot verbrossing van een
polymeer. De trage kristallisatie, gelijktijdig of gescheiden van de poly-
merisatie, leidt tot een kristallisatiegraad van meer dan 40%. Dit is consis-
tent met de hogere modulus en kan de brosheid (deels) verklaren. De DSC-
metingen tonen echter aan dat er nog verschillen zijn tussen gelijktijdige
en gescheiden polymerisatie en kristallisatie. Vandaar dat de kristalstruc-
tuur verder onderzocht werd.
Kristalstructuur en morfologie
Zoals verwacht vertoont zowel pCBT als PBT een α-kristalstructuur aan-
gezien de β-polymorf enkel voorkomt bij spanningsgeı̈nduceerde kristal-
lisatie. Er zijn echter wel duidelijke verschillen merkbaar op de hogere
ordeningsniveau’s.
Wanneer pCBT (of PBT) snel afgekoeld wordt vanuit de smelt, zijn de
gevormde lamellen dun en ongeordend ten opzichte van isotherm gepro-
duceerd pCBT, Figuur 5. Bovendien verschilt ook de sferulitische super-
structuur. Snel afkoelen leidt tot abnormale sferulieten, in tegenstelling
tot de normale sferulieten gevormd tijdens tragere kristallisatie.
Interessanter is echter het verschil tussen gelijktijdige en gescheiden
polymerisatie en kristallisatie. Beide kristallisatieroutes leiden tot de
vorming van normale sferulieten en hebben een vergelijkbare lameldikte.
De TEM-beelden, Figuur 6, tonen echter een verschil in overgang tussen
de kristallijne en amorfe fase. Wanneer pCBT isotherm geproduceerd
wordt, is er een duidelijk afgetekende scheiding tussen de kristallijne en
amorfe fase. Wanneer polymerisatie en kristallisatie daarentegen geschei-
den worden, is het contrast tussen kristallijn en amorf minder scherp en
is er als het ware een overgangslaag.
Het duidelijk contrast tussen de kristallijne en amorfe fase bij isotherm
Figuur 5: TEM beelden van de lamelstructuur van (links) snel

afgekoeld PBT en (rechts) isotherm geproduceerd pCBT
Figuur 6: TEM beelden van de lamelstructuur van (links)

isotherm geproduceerd pCBT en (rechts) pCBT traag afgekoeld
vanuit de smelt
geproduceerd pCBT wijst op de afwezigheid van verbindingsmolecules3 .

Deze verbindingsmolecules zijn essentieel voor de spanningsoverdracht en
een vermindering van deze molecules zorgt voor een verbrossing.
Door de gelijktijdigheid van polymerisatie en kristallisatie zijn er slechts
korte amorfe ketensegmenten aan het groeifront van de kristallen waar-
door de kans daalt dat deze segmenten zullen deel uitmaken van meerdere
lamellen en/of sferulieten [79]. Bovendien kan de transesterificatie-kataly-
sator een negatief effect hebben op het aantal verbindingsmolecules. Aan-
gezien de katalysator niet in de kristalstructuur kan ingepast worden, is
er een verhoogde katalysatorconcentratie aan het kristalgroeifront. De
verhoogde concentratie in combinatie met de mechanische spanningen
aanwezig omwille van de densiteitsverschillen tussen kristallijn en amorf,
kunnen leiden tot een lokale vermeerdering van de transesterificaties wat
op zijn beurt leidt tot een drastische reductie van het aantal verbind-
ingsmolecules.
De brosheid wordt dus niet enkel verklaard door de hoge kristallisatie-
3 E:tie-molecules
graad, maar ook door de afwezigheid van voldoende verbindingsmolecules

om een goede spanningsoverdracht te realiseren tussen de kristallijne en
amorfe fase.
Aanpassing van de matrix

Hoewel niet-isotherme productie de ductiliteit van pCBT substantieel kan
verbeteren, is dit geen optie voor de productie van composieten met RTM.
Impregnatie boven de smelttemperatuur zou immers het tijdsraam te
nauw maken en het doorlopen van een thermische cyclus is bovendien
tijd- en energierovend. Een andere mogelijkheid om de ductiliteit van de
matrix te verbeteren is een chemische aanpassing van de structuur zodat
de kristalliniteit beı̈nvloed wordt.
Vertakkingen in de polymeerketen verlagen de kristalliniteit. Vandaar
dat in een eerste poging een multi-funcioneel epxoy en een hypervertakt
polymeer worden toegevoegd in de hoop de kristalstructuur positief te
beı̈nvloeden. Hoewel de kristallisatiegraad verlaagd wordt door toevoeg-
ing van deze materialen, wordt de brosheid niet vermeden. De piste van
het aanpassen van de kristalstructuur door toevoeging van een kleine hoe-
veelheid ‘vreemd’ materiaal is echter interessant voor verder onderzoek.
Hierbij moet wel rekening gehouden worden met de invloed op de vis-
cositeit.
X-stralen metingen tijdens de kristallisatie

X-stralen metingen worden tijdens de isotherme polymerisatie en kristalli-
satie uitgevoerd en vergeleken met isotherme kristallisatie vanuit de smelt.
Om deze metingen mogelijk te maken, wordt gebruik gemaakt van syn-
chrotron straling omwille van de hoge intensiteit. De metingen werden uit-
gevoerd aan de Nederlands-Belgische bundellijn, DUBBLE BM26, aan de
European Synchrotron Radiation Facility, ESRF in Grenoble, Frankrijk.
WAXD en SAXS patronen4 werden gelijktijdig opgemeten, Figuur 7. Aan
de hand van deze patronen kan de kristallisatiekinetica evenals de ka-
rakteristieke lengtes van de semi-kristallijne twee-fasen structuur bepaald
worden.
Kristallisatiekinetica
De WAXD-gebaseerde kristallisatiekinetica bevestigt de eerdere DSC-re-
sultaten, namelijk dat het isotherme productieproces van CBTr hars re-
sulteert in gelijktijdige polymerisatie en kristallisatie. Deze gelijktijdig-
heid heeft een hogere kristalisatiegraad tot gevolg. Tijdens de koelstap
neemt de kristalliniteit verder toe, maar deze toename wordt teniet gedaan
4 Wide Angle X-ray Diffraction, Small Angle X-ray Scattering
Figuur 7: Gecorrigeerde WAXD (links) en SAXS (rechts) patro-

nen tijdens de isotherme polymerisatie en kristallisatie van CBTr
hars op 190◦ C
tijdens het heropwarmen, nog voor het smeltpunt is bereikt. Dit voortij-
dig smelten wijst op de vorming van kleinere, minder perfecte kristallen
tijdens het afkoelen.
Indien gepolymeriseerd wordt boven het evenwichtssmeltpunt van PBT5
op 250◦ C waarna de kristallisatie plaatsvindt op 190◦ C, wordt dezelfde
kinetiek gevolgd en dezelfde structuur bekomen dan wanneer pCBT vanuit
de smelt kristalliseert.
De glasvezels, aanwezig tijdens de productie van composieten, bevor-
deren de heterogene kiemvorming. Hierdoor wordt, voor de monsters
gekristalliseerd vanuit de smelt, de inductietijd voor kristallisatie verlaagd
en de kristallisatiesnelheid verhoogd.
Voor het isotherm productieproces, resulteert de toevoeging van glas-
vezels echter in een toename van de inductietijd. Deze toename is het
gevolg van een daling van werkelijke hoeveelheid katalysator omwille van
de glasvezelsizing. Dit heeft een daling van de polymerisatiesnelheid tot
gevolg. Na de (verlaatte) start van de kristallisatie, is wel een duidelijke
stijging van de kristallisatiesnelheid merkbaar.
Karakteristieke lengtes en lokale kristalliniteit

In de literatuur bestaat er nog steeds geen consensus over de toewijzing
van de twee karakteristieke lengtes, die volgen uit de correlatiefunctie-
analyse, aan de kristallijne en amorfe fase. Aangezien op kamertempera-
tuur de 001-diffractiepiek zichtbaar is in de WAXD-metingen, kan hieruit
de lameldikte bepaald worden en zo vergeleken worden met de karakte-
ristieke SAXS-lengtes. Aan de hand van deze metingen wordt de langste
lengte toegewezen aan de kristallijne fase. Het gevolg hiervan is echter
dat de lokale kristalliniteit significant groter is dan de globale WAXD-
gebaseerde kristalliniteit.
5 245◦ C [158]
De lameldikte blijkt amper afhankelijk van de thermische geschiedenis

van pCBT, hoewel, onverwacht, de kleinste lameldikte overeenkomt met
de simultane polymerisatie en kristallisatie. De kleine verschillen kun-
nen gerelateerd worden aan een meer uitgesproken overgangslaag wanneer
polymerisatie en kristallisatie gescheiden worden. Deze overgangslaag was
ook duidelijk zichtbaar in de TEM-beelden. De laterale afmetingen van
de lamellen zijn dan weer groter wanneer polymerisatie en kristallisatie
gelijktijdig optreden.
De glasvezels beı̈nvloeden de karakteristieke lengtes niet significant.
Aangezien de afstanden tussen de vezels en tussen de bundels groot zijn
in vergelijking met de lameldiktes, is dit niet verwonderlijk.
Tijdens de isotherme kristallisatie van CBTr hars daalt de gemiddelde
lameldikte zowel bij gelijktijdige als bij gescheiden polymerisatie en kris-
tallisatie. Deze daling wordt voortgezet tijdens de afkoeling. De hiermee
overeenkomende morfologie die in de literatuur wordt voorgesteld bestaat
uit semi-kristallijne stapels (kristallijne lamellen afgewisseld door amorfe
zones) die echter niet ruimtevullend zijn. Tijdens secundaire kristallisatie
worden dan nieuwe, dunnere lamellen gevormd door stapelverdikking of
door het tussenvoegen van nieuwe stapels.
Deze morfologie wordt echter niet ondersteund door de huidige experi-
mentele resultaten aangezien dit zou leiden tot een volledig amorfe fractie
van meer dan 40%. Bovendien neemt de fractie semi-kristallijne stapels
amper toe tijdens het afkoelen.
Morfologie overeenstemmend met WAXD en SAXS resultaten
De twee-fasen structuur die door de correlatiefunctie-analyse veronder-

steld wordt, is niet correct op kamertemperatuur voor stijvere polymeer-
ketens zoals polyesters en polyamides. Deze polymeren hebben een mor-
fologie bestaande uit een kristallijne, een mobiele en een immobiele6 amorfe
fase.
Bovendien werd door Strobl een nieuw kristallisatiemodel naar voren
geschoven waarbij de kristallen niet rechtstreeks vanuit de smelt groeien.
Eerst wordt een mesomorfe, mobiele maar gedeeltelijk geordende fase ge-
vormd. Deze mesomorfe fase wordt dikker en de interne ordening stijgt
totdat ze uiteindelijk overgaat in de kristallijne fase. Zo ontstaat een
korrelige kristalstructuur.
De kristallijne korrels kunnen dan verder stabiliseren door samenvoeg-
ing. Anderen hebben dan weer gesuggereerd dat deze samenvoeging niet
mogelijk is en dat het materiaal tussen de kristallijne korrels verglaasd tot
de immobiele amorfe fractie. Rekening houdend met het bestaan van een
derde fase, worden verschillende morfologische modellen voorgesteld.
6 E: rigid
Figuur 8: Morfologisch model: semi-dense lagen bestaande uit

lamellen, op hun beurt opgebouwd uit kristallijne korrels, ingebed
in een dense amorfe fase, afgewisseld door de amorfe fase met
lagere densiteit
Omwille van de beperkte laterale afmetingen van de kristallen, wordt

het model waarbij de derde fase zich binnen de lamellen bevindt naar
voren geschoven. Aangezien deze beperkte afmetingen geen invloed blij-
ken te hebben op het gebruik van de kwadratische vergelijking voor het
bepalen van de lokale kristalliniteit, wordt bovendien verondersteld dat de
densiteit van deze fractie gelijk is aan de kristallijne densiteit of in ieder
geval dat het verschil met de kristallijne densiteit beperkt is zodat de zij-
delingse oppervlakken geen invloed hebben op de bepaling van de lokale
kristalliniteit. De derde fase wordt de dens amorfe fase (DAF) genoemd.
De resulterende morfologie is vereenvoudigd voorgesteld in Figuur 8.
De lokale kristalliniteit bepaald door SAXS, komt eigenlijk overeen
met de lokale fractie dens materiaal en is gerelateerd aan de WAXD-
gebaseerde globale kristalliniteit door zowel de fractie semi-dense lagen
als door de fractie dens amorf in de lamellen.
Eens de lagen ruimtevullend geworden zijn, wordt de verdere stijging
in kristalliniteit toegewezen aan de structurele overgang tussen dens amorf
en kristallijn. Aangezien de dens amorfe fase afneemt tijdens koelen, kan
deze fase enkel als mobiel beschouwd worden. Omdat de fractie van de
dense, amorfe fase in de lamellen bij de start van de kristallisatie boven-
dien ongeveer één is, wordt het kristallisatiemodel van Strobl door de
experimentele resultaten ondersteund. Daarenboven wordt het mogelijk
geacht dat bij verder afkoelen, de dens amorfe fase verglaasd en dus wordt
omgezet in de immobiele amorfe fase, gevonden bij kamertemperatuur.
De globale fractie dens amorf is kleiner voor simultane polymerisatie en
kristallisatie vergeleken met de fractie ontstaan tijdens gescheiden poly-
merisatie en kristallisatie. Het vergrootte tijdsraam voor kristallisatie
evenals een afname in verstrengelingen en defecten in de mesomorfe fase
zijn hiervoor verantwoordelijk.
Besluit
Ondanks de gekende voordelen van thermoplastische composieten, beslaan
ze slechts een klein deel van de totale markt. De hoge smeltviscositeit
van de matrix bemoeilijkt immers de impregnatie van continue vezelver-
sterkingen. In-situ polymerisatie van laag viskeuze thermoplastische pre-
polymeren maakt het gebruik van eenvoudigere productiemethodes echter
mogelijk. Het gebruikte polymeersysteem in deze studie, CBTr hars, laat
bovendien isotherme productie toe.
Omwille van de isotherme productie treden polymerisatie en kristalli-
satie gelijktijdig op. Dit leidt niet enkel tot een hoge kristalliniteit, maar
bovendien tot een kleinere fractie dens amorf materiaal en een daling in de
verbindingsmolecules tussen de kristallen. Deze morfologie leidt tot een
brosse matrix, wat op zijn beurt leidt tot een lage breukrek en een slechte
matrixtaaiheid. Hierdoor gaat het belangrijkste voordeel van thermoplas-
tische composieten verloren.
Om het succes van thermoplastisch RTM met CBTr hars te verzeke-
ren, is het cruciaal om de matrixtaaiheid te verhogen en dit liefst zonder te
raken een het isotherm karakter van het productieproces. Het aanbrengen
van vertakkingen door bijvoorbeeld copolymerisatie is een interessante
optie, maar moet verder onderzocht worden.
Andere productiemethodes, zoals bijvoorbeeld pultrusie, laten niet-
isotherme productie op een economisch-verantwoorde manier toe. Deze
productiemethodes zijn echter meestal slechts complementair aan RTM.
Een tweede belangrijk aspect in dit werk is de invloed van de vezelsiz-
ing. De vezelsizing verlaagt het moleculair gewicht van de matrix. Boven-
dien zijn de epoxide functionele groepen verantwoordelijk voor het gedeel-
telijk vernetten van de matrix. Niettegenstaande deze effecten, waren de
composieteigenschappen bevredigend. Het verwijderen van de sizing aan
de andere kant resulteert in ondermaatse transversale eigenschappen.
Het is duidelijk dat er nood is aan een specifieke sizing die de ver-
werking van vezels tot textielen toelaat en die eveneens compatibel is met
de polymerisatie. Als daarenboven gebruik kan gemaakt worden van het
reactief karakter van het proces om vezel en matrix chemisch aan elkaar
te binden, betekent dit een extra voordeel ten opzichte van de traditionele
productieprocessen.
Het laatste deel van dit werk bestaat uit een fundamentele studie van
de kristalstructuur. Uit deze studie blijkt dat de lamellaire structuur
van PBT niet enkel bestaat uit kristallijne lamellen en een amorfe inter-
lamellaire fase maar ook een een mobiele, dense en amorfe fase. Deze fase
verdeelt de kristallijne fase in blokken met gelimiteerde laterale dimensies.
Algemene conclusie Deze studie heeft niet alleen de mogelijkheid aange-

toond om thermoplastische composieten te produceren met RTM, de re-
sulterende eigenschappen van deze composieten werden ook bepaald en in
verband gebracht met het productieproces. Bovendien toonde de studie

aan dat een waaier aan technieken en kennisdomeinen moet aangesproken
worden om dit ingenieursprobleem ten volle te begrijpen.
Om de toekomst van de productie van composieten met CBTr hars
te verzekeren moeten verschillende onderzoeksstrategieën dan ook gelijk-
tijdig gevolgd worden. Enerzijds is er nood aan een meer fundamentele
studie van het polymeersysteem en zijn kristallisatieroute in combinatie
met mogelijke matrixwijzigingen om de taaiheid te verbeteren. Deze
studie moet bovendien aangevuld worden met een studie van het grensvlak
tussen matrix en vezel om zo tot een optimale sizing te komen.
Anderzijds is er nood aan een meer ingenieursgerichte benadering om
verschillende productieroutes voor composieten te onderzoeken en om zo
hun eventuele meerwaarde te bepalen. Hierbij kunnen (eenvoudige) mo-
dellen voor de impregnatie, polymerisatie en kristallisatie een doorbraak
van dit polymeersysteem in de industrie realiseren.
REFERENCES 211
References
[1] A.R. Cooper, Determination of molecular weight, Chemical Analy-
sis, vol. 103, John Wiley & Sons, 1989.
[2] J.-A.E. Månson, M.D. Wakeman, and N. Bernet, Comprehensive
composite materials, vol. 2 Polymer Matrix Composites, ch. 2.16
Composite processing and manufacturing-An overview, pp. 577–607,
Elsevier Science Ltd., 2000.
[3] A. Miller and A.G. Gibson, Impregnation techniques for thermoplas-
tic matrix composites, Polymers and Polymer Composites 4 (1996),
no. 7, 459–481.
[4] N. Svensson, R. Shishoo, and M. Gilchrist, Manufacturing of ther-
moplastic composites from commingled yarns - A review, Journal of
Thermoplastic Composite Materials 11 (1998), no. 1, 22–56.
[5] J.D. Muzzy and J.S. Colton, Advanced composites manufacturing,
ch. 3. The processing science of thermoplastic composites, pp. 81–
111, John Wiley & Sons Inc., 1997.
[6] I. Verpoest, Comprehensive composite materials, vol. 2 Polymer
Matrix Composites, ch. 2.18 Composite preforming techniques,
pp. 523–669, Elsevier Science Ltd., 2000.
[7] M.D. Wakeman and C.D. Rudd, Comprehensive composite materi-
als, vol. 2 Polymer Matrix Composites, ch. 2.27 Compression mold-
ing of thermoplastic composites, pp. 915–963, Elsevier Science Ltd.,
2000.
[8] A.M. Sastry, Comprehensive composite materials, vol. 2 Polymer
Matrix Composites, ch. 2.17 Impregnation and consolidation phe-
nomena, pp. 609–622, Elsevier Science Ltd., 2000.
[9] N. Bernet, V. Michaud, P.-E. Bourban, and J.-A.E. Månson, An
impregnation model for the consolidation of thermoplastic compos-
ites made from comingled yarns, Journal of Composite Materials 33
(1999), no. 8, 751–772.
[10] P.E. Bourban, N. Bernet, J.E. Zanetto, and J.-A.E. Månson, Ma-
terial phenomena controlling rapid processing of thermoplastic com-
posites, Composites Part A: Applied Science and Manufacturing 32
(2001), no. 8, 1045–1057.
[11] B.T. Åstrom, Manufacturing of polymer composites, 1st ed., Chap-
man & Hall, London, 1997.
[12] J.P. Fanucci, S. Nolet, and S. McCarthy, Advanced composites man-
ufacturing, ch. 7. Pultrustion of composites, pp. 259–295, John Wi-
ley & Sons, 1997.
[13] A.G. Gibson, Comprehensive composite materials, vol. 2 Polymer
Matrix Composites, ch. 2.29 Continuous molding of thermoplastic
composites, pp. 979–998, Elsevier Science Ltd., 2000.
212 REFERENCES
[14] T.C. Connor, R. Eder, E. Schmid, and U. Wild, EMS polymerisa-

tion moulding (EPM): A novel solution to thermoplastic composite
manufacturing, 1998 (Paris) (M.A. Erath, ed.), Proceedings of the
19th International Sampe Europe Conference, pp. 385–395.
[15] P.O. Mairtin, P. McDonnell, M.T. Connor, R. Eder, and C.M.
Ó Bradaigh, Process investigation of a liquid PA-12/carbon fibre
moulding system, Composites Part A: Applied Science and Manu-
facturing 32 (2001), no. 7, 915–923.
[16] C.W. Macosko, RIM: Fundamentals of reaction injection molding,
Hanser Publishers, 1989.
[17] A. Luisier, P.E. Bourban, and J.-A.E. Månson, Time-temperature-
transformation diagram for reactive processing of polyamide 12,
Journal of Applied Polymer Science 81 (2001), no. 4, 963–972.
[18] A. Luisier, In-situ polymerisation of lactam 12 for liquid moulding of
thermoplastic composites, Ph.D. thesis, Laboratoire de Technologie
des Composites et Polymères, EPFL, 2001.
[19] M. Greaney and C.M. Ó Bradaigh, Development of a polyamide
copolymer resin transfer molding system for thermoplastic compos-
ites, Proceedings of the 7th International Conference on Flow Pro-
cesses in Composite Materials (Newark, Delaware, USA) (S.G. Ad-
vani, ed.), 2004, pp. 157–161.
[20] W. Yan, K. Han, L. Qin, and M. Yu, Study on long fiber-reinforced
thermoplastic composites prepared by in situ solid-state polyconden-
sation, Journal of Applied Polymer Science 91 (2004), 3959–3965.
[21] T.J. Corden, I.A. Jones, C.D. Rudd, P. Christian, S. Downes, and
K.E. McDougall, Physical and biocompatibility properties of poly-²-
caprolactone produced using in sity polymerisation: a novel manu-
facturing technique for long-fibre reinforced composites, Biomaterials
21 (2000), 713–724.
[22] P. Christian, I.A. Jones, C.D. Rudd, R.I. Campbell, and T.J. Cor-
den, Monomer tranfer moulding and rapid prototyping methods for
fibre reinforced thermoplastics for medical applications, Composites
Part A: Applied Science and Manufacturing 32 (2001), 969–976.
[23] G. Jiang, M.E. Evans, I.A. Jones, C.D. Rudd, C.A. Scotchford,
and G.S. Walker, Preparation of poly(ε-caprolactone)/continuous
bioglass fibre composite using monomer transer moulding for bone
implant, Biomaterials 26 (2005), 2281–2288.
[24] N. Pini, C. Zaniboni, S. Busato, and P. Ermanni, Perspectives for
reactive molding of PPA as matrix for high-performance composite
materials, Proceedings of the 7th International Conference on Flow
Processes in Composite Materials (Newark, Delaware, USA) (S.G.
Advani, ed.), 2004, pp. 107–112.
[25] C.M. Edwards and E.L. d’Hooghe, Fiber-reinforced composite and
REFERENCES 213
method of making same, 1999, Patent US5891560.

[26] K. Boschmans, In-situ polymerisatie van continu vezelversterkte
thermoplasten, Master’s thesis, Departement MTM, K.U.Leuven,
2002-2003.
[27] K. van Rijswijk, K. Koppes, H.E.N. Bersee, and A. Beukers,
Processing window for vacuum infusion of fiber-reinforced anionic
polyamide-6, Proceedings of the 7th International Conference on
Flow Processes in Composite Materials (Newark, Delaware, USA)
(S.G. Advani, ed.), 2004, pp. 71–76.
[28] K. van Rijswijk, H.E.N. Bersee, W.F. Jager, and S.J.N Picken, Opti-
misation of anionic polyamide-6 for vauum infusion of thermoplastic
composites: choice of activator and initiator, Composites Part A:
Applied Science and Manufacturing in press (2005).
[29] D.P.N. Vlasveld, K. van Rijswijk, P. van Rhijn, H.E.N. Bersee,
A. Beukers, and S.J.N. Picken, Process considerations for liquid
molding of composites based on anionic polyamide 6, Proceedings
of ICCM-14 (San Diego, CA, USA), 2003, p. CDrom paper 1569.
[30] E.C. Botelho, N. Scherbakoff, M.C. Rezende, A.M. Kawamoto, and
J. Sciamareli, Synthesis of polyamide 6/6 by interfacial polyconden-
sation with the simultaneous impregnation of carbon fibers, Macro-
molecules 34 (2001), 3367–3375.
[31] E.C. Botelho, L. Figiel, M.C. Rezende, and B. Lauke, Mechanical
behavior of carbon fiber reinforced polyamide composites, Compos-
ites Science and Technology 63 (2003), 1843–1855.
[32] E.C. Botelho, Reaction injection pultrusion: A potential technique
to process polyamide 6,6/carbon fiber composites, Journal of Ad-
vanced Materials 36 (2004), no. 2, 49–53.
[33] X. Ning and H. Ishida, RIM-pultrusion of nylon-6 and rubber-
toughened nylon-6 composites, Polymer Engineering and Science 31
(1991), no. 9, 632–637.
[34] X. Ning and H.S. Ishida, Dynamic mechanical analysis of RIM
nylon-6, Journal of Polymer Science, Part B: Polymer Physics 29
(1991), no. 12, 1479–1492.
[35] B.G. Cho and S.P. McCarthy, Performance of nylon-6 compos-
ites produced by the reaction injection pultrusion process, Proceed-
ings of the 24th International SAMPE Technical Conference, 1992,
pp. T645–T659.
[36] B.G. Cho, S.P. McCarthy, J.P. Fanucci, and S.C. Nolet, Fiber rein-
forced nylon-6 composites produced by the reaction injection pultru-
sion process, Polymer Composites 17 (1996), no. 5, 673–681.
[37] A. Luisier, P.E. Bourban, and J.-A.E. Månson, Reaction injection
pultrusion of PA12 composites: process and modelling, Composites
Part A: Applied Science and Manufacturing 34 (2003), 583–595.
214 REFERENCES
[38] K.C. Hranac, New thermoplastic technologies heat up pultrusion,

Composites Technology Jan-Febr (2001).
[39] John H. Glenn Research Centre NASA, High Tg polyimide, Website,
2001.
[40] John H. Glenn Research Centre NASA, Low cost manufacturing for
high temperature polymer matrix composites, Website, 2002.
[41] J.L. Lee, Advanced composites manufacturing, ch. 10. Liquid com-
posite moulding, pp. 393–456, John Wiley & Sons Inc., 1997.
[42] F.M. Sweeney, Reaction injection moulding machinery and pro-
cesses, 1st ed., Marcel Dekker Inc., 1987.
[43] A.J. Ryan, J.L. Stanford, and X.Q. Tao, The processing, structure
and properties of structural composites formed by reaction injection-
molding (SRIM), Plastics Rubber and Composites Processing and
Applications 23 (1995), no. 3, 151–159.
[44] R. Eder and S.J. Winckler, Processing of advanced thermoplastic
composites using cyclic thermoplastic polyester, Proceedings of the
22nd International Sampe Europe Conference (Paris, France) (G.R.
Griffiths and R.F.J. McCarthy, eds.), Sampe Europe, 2001, pp. 661–
672.
[45] D. Gittel, H.-C. Ludwig, and P. Elsner, Guβpolyamid mit
faserverstärkungen, Proceedings of the 15th Stuttgarter Kunststoff-
Kolloqium (Stuttgart), 1997.
[46] D. Gittel, B. Möginger, and P. Eyerer, Composite materials using
RTM with ε-caprolactam, Proceedings of ECCM-8 (Naples, Italy),
1998, pp. 573–576.
[47] P. Rosso, K. Friedrich, and A. Wollny, Evaluation of the adhesion
quality between differently treated carbon fibers and an in-situ poly-
merized polyamide 12 system, Journal of Macromolecular Science,
Physics B41 (2002), no. 4-6, 745–759.
[48] S. Pillay, H. Ning, U.K. Vaidya, and G.M. Janowski, Liquid molding
of carbon fabric reinforced nylon matrix composite laminates, Pro-
ceedings of the 7th International Conference on Flow Processes in
Composite Materials (Newark, Delaware, USA) (S.G. Advani, ed.),
2004, pp. 65–70.
[49] K. van Rijswijk, D.P.N. Vlasveld, H.E.N. Bersee, and S.J.N. Picken,
Vacuum injection of anionic polyamide 6, Proceedings of ICCST-4
(Durban, South Africa), 2003.
[50] L. Zingraff, P.-E. Bourban, V. Michaud, and J.-A.E. Månson, Liquid
composite moulding of anionically polymerised polyamide 12, Pro-
ceedings of the 14th International Conference on Composite Mate-
rials, 2003, p. CDrom paper 1849.
[51] V. Michaud, L. Zingraff, J. Verrey, P.-E. Bourban, and J.-A.E.
Månson, Resin transfer molding of anionically polymerized poly-
REFERENCES 215
amide 12, Proceedings of the 7th International Conference on Flow

Advani, ed.), 2004, pp. 85–93.
[52] J. Verrey, V. Michaud, and J.-A.E. Månson, Capillary effects in
liquid composite moulding with non-crimp fabrics, Proceedings of
ICCM-14, 2003, p. CDrom paper 1658.
[53] M.D. Wakeman, L. Zingraff, M. Kohler, P.-E. Bourban, and J.-A.E.
Månson, Stamp-forming of carbon fibre/PA12 composite preforms,
Proceedings of ECCM-10, Composites for the future (Brugge, Bel-
gium) (H. Sol and J. Degrieck, eds.), 2002.
[54] L. Zingraff, P.-E. Bourban, M.D. Wakeman, M. Kohler, and J.-A.E.
Månson, Reactive processing and forming of polyamide 12 thermo-
plastic composites, Proceedings of the 23rd International Sampe Eu-
rope Conference (Paris, France) (F.J. Barnes and R.J. McCarthy,
eds.), 2002, pp. 237–248.
[55] M.D. Wakeman, L. Zingraff, P. Blanchard, and J.-A.E. Månson,
Stamp-forming of reactive-thermoplastic carbon fiber/PA12 compos-
ite sheet, Proceedings of the 7th International Conference on Flow
Advani, ed.), 2004, pp. 77–84.
[56] J. Verrey, V. Michaud, and J.-A.E. Månson, Processing of complex
parts with thermoplastic RTM techniques, Proceedings of the 24th
International Sampe Europe Conference (Paris, France) (K. Drech-
sler, ed.), 2003, pp. 553–561.
[57] J. Verrey, M.D. Wakeman, F. Roduit, V. Michaud, and J.-A.E.
Månson, Manufacturing cost comparison of thermoplastic and ther-
moset RTM for an automotive floor-pan, Proceedings of the 25th
Sampe Europe Conference (Paris, France) (K. Drechsler, ed.),
Sampe Europe, 2004, pp. 207–212.
[58] P. Hodge, Some applications of reactions which interconvert mono-
mers, polymers and/or macrocycles, Reactive & Functional Poly-
mers 48 (2001), 15–23.
[59] R.R. Burch, S.R. Lustig, and M. Spinu, Synthesis of cyclic oli-
goesters and their rapid polymerization to high molecular weight,
Macromolecules 33 (2000), no. 14, 5053–5064.
[60] J.U. Otaigbe, Advances in cyclomer technology for thermoplastic
composites manufacture, Trends in Polymer Science 5 (1997), no. 1,
17–23.
[61] A.J. Hall and P. Hodge, Recent research on the synthesis and ap-
plications of cyclic oligomers, Reactive & Functional Polymers 41
(1999), no. 1-3, 133–139.
[62] D.J. Brunelle, New methods of polymer synthesis, vol. 2, ch. Macro-
cycles for the synthesis of high molecular weight polymers, pp. 197–
216 REFERENCES
235, Chapman & Hall, 1995.

[63] D.J. Brunelle, Cyclic polymers, 2nd ed., ch. 6. Cyclic oligomers of
polycarbonates and polyesters, pp. 185–228, Kluwer Academic Pub-
lishers, 2000.
[64] D.A. Steenkamer and J.L. Sullivan, On the recyclability of a cyclic
thermoplastic composite material, Composites Part B: Engineering
29 (1998), no. 6, 745–752.
[65] D.J. Brunelle, T.L. Evans, T.G. Shannon, E.P. Boden, K.R. Stew-
art, L.P. Fontana, and D.K. Bonauto, Cyclic oligomeric carbon-
ates - New route to super high-molecular-weight polycarbonate - An
overview, Polymer Preprints (American Chemical Society, Division
of Polymer Chemistry) 198 (1989), 331–POLY.
[66] E.P. Boden, D.J. Brunelle, and T.G. Shannon, Efficient prepara-
tion of cyclic oligomeric bisphenol-a carbonates - Mechanism for
selective formation of cyclic oligomers, Polymer Preprints (Ameri-
can Chemical Society, Division of Polymer Chemistry) 198 (1989),
332–POLY.
[67] T.L. Evans, C.B. Berman, D.Y. Choi, J.C. Carpenter, and D.W.
Williams, Fundamentals of cyclic carbonate oligomer ring-opening
polymerization, Polymer Preprints (American Chemical Society, Di-
vision of Polymer Chemistry) 198 (1989), 333–POLY.
[68] K.R. Stewart, Melt polymerization of Bpa cyclic polycarbonate oli-
gomers - Rheokinetics of polymerization relevant to reactive pro-
cessing, Polymer Preprints (American Chemical Society, Division of
Polymer Chemistry) 198 (1989), 334–POLY.
[69] D.J. Brunelle, Encyclopedia of materials: Science and technology,
ch. Macrocyclic oligomers of engineering thermoplastics, pp. 4712–
4720, Elsevier Science, 2001.
[70] G.C. East and A.M. Girshab, Cyclic oligomers in poly(1,4-butylene
terephthalate), Polymer 23 (1982), 323–324.
[71] D.J. Brunelle, J.E. Bradt, J. Serth-Guzzo, T. Takekoshi, T.L. Evans,
E.J. Pearce, and P.R. Wilson, Semicrystalline polymers via ring-
opening polymerization: Preparation and polymerization of alkylene
phthalate cyclic oligomers, Macromolecules 31 (1998), no. 15, 4782–
4790.
[72] S.C. Hamilton and J.A. Semlyen, Cyclic polyesters: 5. Cyclics
prepared by poly(decamethylene terephthalate) ring-chain reactions,
Polymer 38 (1997), no. 7, 1685–1691.
[73] J.J.L. Bryant and J.A. Semlyen, Cyclic polyesters: 7. Preparation
and characterization of cyclic oligomers from solution ring-chain
reactions of poly(butylene terephthalate), Polymer 38 (1997), no. 17,
4531–4537.
[74] J.J.L. Bryant and J.A. Semlyen, Cyclic polyesters: 6. Preparation
REFERENCES 217
and characterization of two series of cyclic oligomers from solu-

tion ring-chain reactions of poly(ethylene terephthalate), Polymer
38 (1997), no. 10, 2475–2482.
[75] S.C. Hamilton, J.A. Semlyen, and D.M. Haddleton, Cyclic polyes-
ters: Part 8. Preparation and characterisation of cyclic oligomers
in six aromatic ester and ether-ester systems, Polymer 39 (1998),
no. 14, 3241–3252.
[76] P.A. Hubbard, W.J. Brittain, W.L. Mattice, and D.J. Brunelle,
Ring-size distribution in the depolymerization of poly(butylene
terephthalate), Macromolecules 31 (1998), no. 5, 1518–1522.
[77] C.L. Ruddick, P. Hodge, Y. Zhuo, R.L. Beddoes, and M. Helliwell,
Cyclo-depolymerisation of polyundecanoate and related polyesters:
Characterisation of cyclic oligoundecanoates and related cyclic oli-
goesters, Journal of Materials Chemistry 9 (1999), 2399–2405.
[78] S.D. Kamau, P. Hodge, and M. Helliwell, Cyclo-depolymerization of
poly(propylene terephthalate): Some ring-opening polymerizations
of the cyclic oligomers produced, Polymers for Advanced Technolo-
gies 14 (2003), 492–501.
[79] S. Miller, Macrocyclic polymers from cyclic oligomers of
poly(butylene terephthalate), Ph.D. thesis, Polymer Science and En-
gineering, University of Massachusetts, 1998.
[80] J.H. Youk, A. Boulares, R.P. Kambour, and W.J. MacKnight, Poly-
merization of ethylene terephthalate cyclic oligomers with a cyclic
dibutyltin initiator, Macromolecules 33 (2000), no. 10, 3600–3605.
[81] J.H. Youk, R.P. Kambour, and W.J. MacKnight, Polymerization
of ethylene terephthalate cyclic oligomers with antimony trioxide,
Macromolecules 33 (2000), no. 10, 3594–3599.
[82] S.R. Bahr, Cyclic thermoplastics: new oppertunities in reactive com-
pounding, Proceedings of Performance Compounding 2002 (San An-
tonio, TX, USA), 2002.
[83] S.R. Bahr, Mining new opportunities with cyclic polyesters, Pro-
ceedings of Functional Fillers for Plastics (Atlanta, USA), 2003.
[84] A.R. Tripathy, W. Chen, S.N. Kukureka, and W.J. MacKnight,
Novel poly(butylene terephthlate)/poly(vinyl butyral) blends prepared
by in situ polymerization of cyclic poly(butylene terephthalate) oli-
gomers, Polymer 44 (2003), 1835–1842.
[85] A.R. Tripathy, W. MacKnight, and S.N. Kukureka, In-situ copoly-
merization of cyclic poly(butylene terephthalate) oligomers and ε-
caprolactone, Macromolecules 37 (2004), 6793–6800.
[86] A.R. Tripathy, E. Burgaz, S.N. Kukureka, and W.J. MacKnight,
Poly(butylene terephthalate) nanocomposites prepared by in-situ
polymerization, Macromolecules 36 (2003), 8593–8595.
[87] Cyclics Corporation, Catalyst systems for polymerisation of CBTr
218 REFERENCES
resin, Personnal communication, 2001.

[88] M. Rösch and R.H.J. Eder, Processing cyclic thermoplastic polyes-
ter, Proceedings of the 9th International Conference on Fiber Rein-
forced Composites (Newcastle, UK) (A.G. Gibson, ed.), vol. 1, 2002,
pp. 54–62.
[89] Kunststofftechnik Beiner K.G., Report on polymer processing and
selection of resin systems for JET-ex process, Tech. report, Amiterm
project, 2002.
[90] J.D. Muzzy, Comprehensive composite materials, vol. 2 Polymer
Matrix Composites, ch. 2.02 Thermoplastics-Properties, pp. 57–76,
Elsevier Science Ltd., 2000.
[91] D. Bank, P. Cate, and M. Shoemaker, pCBT: A new material for
high performance composites in automotive applications, Proceed-
ings of the 4th Annual SPE Automotive Composites Conference
(Troy, Michigan, USA), 2004.
[92] R. Al-Zubi, Y.-F. Wang, and P. Larson, Rotomoldability of
cyclic polybutyleneterephthalate, Proceedings of Antec 2003 Plastics
(Nashville, USA), vol. 1, 2003, pp. 1257–1261.
[93] S. Miller, J. Donovan, and W. MacKnight, Toughness enhance-
ment through conversion of cyclic polybutylene terephthalate to lin-
ear PBT, Proceedings of Antec 2000, 2000.
[94] J. Ciovacco and S.J. Winckler, Cyclic thermoplastic properties and
processing, Proceedings of the 45th International SAMPE Sympo-
sium (Long Beach, CA, USA) (S. Loud, V. Karbhara, D.O. Adams,
and A.B. Strong, eds.), vol. 45, SAMPE Publishing, 2000.
[95] M. Rösch, Processing of advanced thermoplastic composites using
a novel cyclic thermoplastic polyester one part system designed for
resin transfer molding, Proceedings of the 25th SAMPE Europe
Conference (Paris, France) (K. Drechsler, ed.), Sampe Europe, 2004,
pp. 305–310.
[96] S.J. Winckler, J. Wang, and O. Hanitzsch, Processing thermoplastic,
resin film infusion materials, based on cyclic butylene terephthalate,
Proceedings of the 24th International SAMPE Europe Conference
(K. Drechsler, ed.), Sampe Europe, 2003, pp. 661–668.
[97] S.M. Coll, A.M. Murtagh, and C.M. Ó Bradaigh, Resin film infusion
of cyclic PBT composites: A fundamental study, Proceedings of the
25th Sampe Europe Conference (Paris) (K. Drechsler, ed.), Sampe
Europe, 2004, pp. 311–317.
[98] S.M. Coll, A.M. Murtagh, and C.M. Ó Bradaigh, Resin film infusion
of cyclic PBT composites: Consolidations analysis, Proceedings of
the 7th International Conference on Flow Processes in Composite
Materials (Newark, Delaware, USA), 2004, pp. 101–106.
[99] M. Repsch, U. Huber, M. Maier, S. Rief, D. Kehrwald, and
REFERENCES 219
K. Steiner, Process simulation of LPM (liquid polymer molding) in

special consideration of fluid velocity and viscosity characteristics,
Proceedings of the 7th International Conference on Flow Processes
in Composite Materials (Newark, Delaware, USA) (S.G. Advani,
ed.), 2004, pp. 305–309.
[100] F. Weyrauch, H.C. Stadtfeld, and P. Mitschang, Simulation and
control of the LCM-process with future matrix systems, Proceedings
of the 7th International Conference of Flow Processes in Compos-
ite Materials (Newark, Delaware, USA) (S.G. Advani, ed.), 2004,
pp. 95–100.
[101] D. Hull, Matrix-dominated properties of polymer matrix composite
materials, Materials Science and Engineering A 194 (1994), 173–
183.
[102] P.C. Painter and M.M. Coleman, Fundamentals of polymer science,
2nd ed., Technomic, 1997.
[103] M. Yokouchi, Y. Sakakibara, Y. Chatani, H. Tadokoro, T. Tanaka,
and K. Yoda, Structures of two crystalline forms of poly(butylene
terephthalate) and reversible transition between them by mechanical
deformation, Macromolecules 9 (1976), no. 2, 266–273.
[104] B. Wunderlich, Crystallization during polymerization, Advanced
Polymer Science 5 (1968), 568–619.
[105] C.A. Daniels, Polymers: Structure and properties, Technomin Pub-
lishing Co., Inc., 1989.
[106] L.E. Nielsen, Mechanical properties of polymers and composites,
vol. 2, Marcel Dekker Inc., 1974.
[107] T.J. Bessell, D. Hull, and J.B. Shortall, Effect of polymerization
conditions and crystallinity on mechanical properties and fracture
of spherulitic nylon 6, Journal of Materials Science 10 (1975), no. 7,
1127–1136.
[108] J.J. Janimak and G.C. Stevens, Inter-relationships between tie-
molecule concentrations, molecular characteristics and mechanical
porperties in metallocene catalysed medium density polyethylenes,
Journal of Materials Science 36 (2001), 1879–1884.
[109] I.H. Hall, Structure of crystalline polymers, ch. The determination
of the structures of aromatic polyesters from their wide-angle X-ray
diffraction patterns, pp. 39–78, Elsevier Applied Schience Publish-
ers, London and New York, 1984.
[110] P.L. Carr, R. Jakeways, and I.M. Ward, Tensile drawing, morphol-
ogy, and mechanical properties of poly(butylene terephthalate), Jour-
nal of Polymer Science, Part B: Polymer Physics 35 (1997), 2465–
2481.
[111] R.S. Stein and A. Misra, Morphological studies on polybutylene
terephthalate, Journal of Polymer Science: Polymer Physics Edition
220 REFERENCES
18 (1980), 327–342.
[112] E.J. Roche, R.S. Stein, and E.L. Thomas, Electron microscopy
study of the structure of normal and abnormal poly(butylene tereph-
thalate), Journal of Polymer Science, Polymer Physics Edition 18
(1980), 1145–1158.
[113] C.S. Park, K.J. Lee, S.W. Kim, K.Y. Lee, and J.D. Nam, Crys-
tallinity morphology and dynamic mechanical characteristics of PBT
polymer and glass fiber-reinforced composites, Journal of Applied
Polymer Science 86 (2002), 478–488.
[114] M.L. Di Lorenzo and M.C. Righetti, Crystallization of poly(butylene
terephthalate), Polymer Engineering and Science 43 (2003), no. 12,
1889–1894.
[115] H.J. Ludwig and P. Eyerer, Influence of the processing conditions
on morphology and deformation behavior of poly(butylene tereph-
thalate) (PBT), Polymer Engineering and Science 28 (1988), no. 3,
143–146.
[116] B. Möginger, C. Lutz, A. Polsak, and U. Fritz, Einfluss der verar-
beitung auf mechanische eigenschaften und morphologie von PBT,
Kunststoffe 81 (1991), no. 3, 251–255.
[117] A. Misra and S.N. Garg, Morphology and properties of poly(butylene
terephthalate), Journal of Polymer Science, Polymer Letters Edition
20 (1982), 121–125.
[118] S.Y. Hobbs and C.F. Pratt, Multiple melting of poly(butylene tereph-
thalate), Polymer 16 (1975), 462–464.
[119] J.T. Yeh and J. Runt, Multiple melting in annealed poly(butylene
terephthalate), Journal of Polymer Science, Part B: Polymer Physics
27 (1989), 1543–1550.
[120] I.S. Al-Raheil and A.M.A. Qudah, Morphology and melting be-
haviour of poly(butylene terephthalate), Polymer International 37
(1995), 47–52.
[121] M.E. Nichols and R.E. Robertson, The multiple melting endotherms
from poly (butylene terephthalate), Journal of Polymer Science, Part
B: Polymer Physics 30 (1992), no. 7, 755–768.
[122] H.G. Kim and R.E. Robertson, Multiple melting endotherms in
isothermally melt-crystallized poly(butylene terephthalate), Journal
of Polymer Science, Part B: Polymer Physics 36 (1998), no. 10,
1757–1767.
[123] M. Yasuniwa, S. Tsubakihara, K. Ohoshita, and S. Tokudome, X-
ray studies on the double melting behavior of poly(butylene tereph-
thalate), Journal of Polymer Science, Part B: Polymer Physics 39
(2001), 2005–2015.
[124] M.L. Di Lorenzo and M.C. Righetti, Morphological analysis of
poly(butylene terephthalate) spherulites during fusion, Polymer Bul-
REFERENCES 221
letin 53 (2004), 53–62.

[125] W. Michaeli and M. Koschmieder, Comprehensive composite mate-
rials, vol. 2 Polymer Matrix Composites, ch. 2.25 Processing princi-
ples for thermoplastic polymers, pp. 853–872, Elsevier Science Ltd.,
2000.
[126] E.-P. Chang and E.L. Slagowski, Morphology and physical properties
of poly(butylene terephthalate), Journal of Applied Polymer Science
22 (1978), 769–779.
[127] J. Vendramini, C. Bas, G. Merle, P. Boissonnat, and N.D. Alberola,
Commingled poly(butylene terephthalate)/unidirectional glass fiber
composites: Influence of the process conditions on the microstruc-
ture of poly(butylene terephthalate), Polymer Composites 21 (2000),
no. 5, 724–733.
[128] C.S. Park, K.J. Lee, J.D. Nam, and S.W. Kim, Crystallization ki-
netics of glass fiber reinforced PBT composites, Journal of Applied
Polymer Science 78 (2000), no. 3, 576–585.
[129] D. Philips, Characterisation and development of 3D-knitted compos-
ites, Ph.D. thesis, Faculteit Toegepaste Wetenschappen, Departe-
ment Metaalkunde en Toegepaste Materiaalkunde, K.U.Leuven,
1999.
[130] K. Vallons, Invloed van de matrixtaaiheid op de impactweerstand van
thermoplastische PBT composieten, Master’s thesis, Departement
MTM, K.U.Leuven, 2004-2005.
[131] J. Baets, Karakterisatie van continu vezelversterkte composieten met
in situ gepolymeriseerd CBT, Master’s thesis, Departement MTM,
K.U.Leuven, 2003-2004.
[132] I. Verpoest, H. Parton, F. Desplentere, S. Voskamp, A. Holmberg,
and P.-A. Löfgren, Design and production of a fibre reinforced ther-
moplastic leaf spring by RTM like technique, Proceedings of the 7th
International Conference on Textile Composites (Yonezawa, Yama-
gata, Japan), 2004.
[133] A. Bergeret, M.P. Bozec, J.-C. Quantin, and A. Crespy, Study of
interphase in glass fiber-reinforced poly(butylene terephthalate) com-
posites, Polymer Composites 25 (2004), no. 1, 12–25.
[134] W. Hale, H. Keskkula, and D.R. Paul, Compatibilization of
PBT/ABS blends by methyl methacrylate-glycidyl methacrylate-
ethyl acrylate terpolymers, Polymer 40 (1999), 365–377.
[135] P. Martin, J. Devaux, R. Legras, M. van Gurp, and M. van Duin,
Competitive reactions during compatibilization of blends of poly-
butyleneterephthalate with epoxide-containing rubber, Polymer 42
(2001), no. 6, 2463–2478.
[136] B. Kulshreshtha, A.K. Ghosh, and A. Misra, Crystallization kinet-
ics and morphological behavior of reactively processed PBT/epoxy
222 REFERENCES
blends, Polymer 44 (2003), 4723–4734.

[137] R.A. Kudva, H. Keskkula, and D.R. Paul, Compatibilization of ny-
lon 6/ABS blends using glycidyl methacrylate/methyl methacrylate
copolymers, Polymer 39 (1997), no. 12, 2447–2460.
[138] I.C. McNeill and M. Bounekhel, Thermal degradation studies of
terephthalate polyesters: 1. Poly(alkylene terephthalates), Polymer
Degradation and Stability 34 (1991), 187–204.
[139] F. Samperi, C. Puglisi, R. Alicata, and G. Montaudo, Thermal
degradation of poly(butylene terephthalate) at the processing tem-
perature, Polymer Degradation and Stability 83 (2004), 11–17.
[140] P.K. Mallick, Fiber-reinforced composites, 2nd ed., Marcel Decker,
New York, 1993.
[141] E.L. d’Hooghe, B. Hoek, and C.M. Edwards, Applications for
fulcrum thermoplastic composite technology, Proceedings of the
EPTA 5th World Pultrusion Conference (Leusden, the Nether-
lands), EPTA, 2000.
[142] J.E. Sumerak, Though choices for pultruders-polyurethane for high
performance pultrusions, Proceedings of EPTA 7th World Pultru-
sion Conference. Composite profiles speed & performance (Amster-
dam, The Netherlands), 2004.
[143] D.W. Dwight, Comprehensive composite materials, vol. 1 Fiber Re-
inforcements and General Theory of Composites, ch. 1.08 Glass fiber
reinforcements, pp. 231–261, Elsevier Science Ltd., 2000.
[144] P.A. Smith, Comprehensive composite materials, vol. 2 Polymer Ma-
trix Composites, ch. 2.04 Carbon Fiber Reinforced Plastics - Prop-
erties, pp. 107–150, Elsevier Science Ltd., 2000.
[145] S. Mall, Composites engineering handbook, ch. 16. Laminated poly-
mer matrix composites, pp. 811–890, Marcel Dekker, Inc., 1997.
[146] S. Goa and J. Kim, Cooling rate influences in carbon fibre/PEEK
composites. Part I. Crystallinity and interface adhesion, Composites
Part A 31 (2000), 517–530.
[147] T. Kuboki, P.-Y.B. Jar, and T.W. Forest, Influence of interlami-
nar fracture toughness on impact resistance of glass fibre reinforced
polymers, Composites Science and Technology 63 (2003), 943–953.
[148] P.Y. Jar, X. Gros, K. Takahashi, K. Kawabata, J. Murai, and Y. Shi-
nagawa, Evaluation of delamination resistance of glass fibre rein-
forced polymers under impact loading, Journal of Advanced Materi-
als 32 (2000), no. 3, 35–45.
[149] S.Y. Hobbs and C.F. Pratt, The effect of skin-core morpohology
on the impact fracture of poly(butylene terephthalate), Journal of
Applied Polymer Science 19 (1975), 1701–1722.
[150] S.R. Bahr, Transverse tensile properties of unidirectinal pCBT glass
REFERENCES 223
fibre composites, 2004, Private communication.

[151] L. Li, H. Rui, Z. Ling, and H. Shiming, A new mechanism in the
formation of PET extended-chain crystal, Polymer 42 (2001), 2085–
2089.
[152] B. Pettersson, Hyperbranched polymers-Unique design tools for
multi-property control in resins and coatings, Tech. report,
www.perstorp.com (scienctific publications).
[153] L. Boogh, G. Jannerfeldt, B. Pettersson, H. Björnberg, and J.-A.E.
Månson, Dendritic-based additives for polymer matrix composites,
Proceedings of the 12th International Conference on Composite Ma-
terials (Paris), 1999.
[154] L. Boogh, B. Pettersson, and J.-A.E. Månson, Dendritic hyper-
branched polymers as tougheners for epoxy resins, Polymer 40
(1999), 2249–2261.
[155] F.J. Balta-Calleja and C.G. Vonk, X-ray scattering of synthetic poly-
mers, Polymer Science Library, vol. 8, Elsevier Science Publishers,
1989.
[156] G.H. Stout and L.H. Jensen, X-ray structure determination, 2nd
ed., John Wiley & Sons, 1989.
[157] European Synchrotron Radiation Facility, A light for industry,
brochure, 2002.
[158] M. Pyda, Athas databank, 2005, internet address:
http://web.utk.edu/˜athas/databank/Intro.Html.
[159] B.S. Hsiao, Z. Wang, F. Yeh, Y. Gao, and K.C. Sheth, Time-resolved
X-ray studies of structure development in poly(butylene terephtha-
late) during isothermal crystallisation, Polymer 40 (1999), 3515–
3523.
[160] B. Goderis, M. Peeters, V.B.F. Mathot, M.H.J. Koch, W. Bras, A.J.
Ryan, and H. Reynaers, Morphology of homogeneous copolymers
of ethylene and 1-octene. III. Structural changes during heating as
revealed by time-resolved SAXS and WAXD, Journal of Polymer
Science, Part B: Polymer Physics 38 (2000), 1975–1991.
[161] H.G. Haubruge, A.M. Jonas, and R. Legras, Morphological study
of melt-crystallized poly(ethylene terephthalate). A. Comparison of
transmission electron microscopy and small-angle X-ray scattering
of bulk samples, Macromolecules 37 (2004), 126–134.
[162] B. Lee, T.J. Shin, S.W. Lee, J. Yoon, J. Kim, H.S. Youn, and
M. Ree, Time-resolved X-ray scattering and calorimetric studies on
the crystallization behaviours of poly(ethylene terephthalate) (PET)
and its copolymers containing isophtahlate units, Polymer 44 (2003),
2509–2518.
[163] C. Santa Cruz, N. Stricbeck, and H.G. Zachmann, Novel aspects in
the structure of poly(ethylene terephthalate) as revealed by means of
224 REFERENCES
small-angle X-ray scattering, Macromolecules 24 (1991), 5980–5990.

[164] Z.-G. Wang, B.S. Hsiao, B.B. Sauer, and W.G. Kampert, The nature
of secondary crystallization in poly(ethylene terephthalate), Polymer
40 (1999), 4615–4627.
[165] Z.-G. Wang, B.S. Hsiao, B.X. Fu, L. Liu, F. Yeh, B.B. Sauer,
H. Chang, and J.M. Schultz, Correct determination of crystal
lamellar thickness in semicrystalline poly(ethylene terephthalate) by
small-angle X-ray scattering, Polymer 41 (2000), 1791–1797.
[166] Z. Xia, H.-J. Sue, Z. Wang, C.A. Avila-Orta, and B.S. Hsiao, Deter-
mination of crystalline lamellar thickness in poly(ethylene tereph-
thalate) using small-angle X-ray scattering and transmission elec-
tron microcopy, Journal of Macromolecular Science, Physics B40
(2001), no. 5, 625–638.
[167] W. Gabriëlse, M. Soliman, and K. Dijkstra, Microstructure and
phase behavior of block copoly(ether ester) thermoplastic elastomers,
Macromolecules 34 (2001), 1685–1693.
[168] P.P. Huo, P. Cebe, and M. Capel, Real-time X-ray scattering study
of thermal expansion of poly(butylene terephthalate), Journal of
Polymer Science, Part B: Polymer Physics 30 (1992), 1459–1468.
[169] Y.A. Fakhreddine and P. Zoller, The equation of state of solid and
molten poly(butylene terephthalate) to 300◦ C and 200 MPa, Journal
of Polymer Science, Part B: Polymer Physics 29 (1991), 1141–1146.
[170] G.R. Strobl, From the melt via mesomorphic and granular crys-
talline layers to lamellar crystallites: A major route followed in poly-
mer crystallization?, The European Physical Journal E 3 (2000),
165–183.
[171] C. Schick, A. Wurm, and A. Mohamed, Vitrification and devitrifi-
cation of the rigid amorphous fraction of semicrystalline polymers
revealed from frequency-dependent heat capacity, Colloid and Poly-
mer Science 279 (2001), 800–806.
[172] S.Z.D. Cheng, R. Pan, and B. Wunderlich, Thermal analysis of
poly(butylene terephthalate) for heat capacity, rigid-amorphous con-
tent, and transition behavior, Makromolekulare Chemie 189 (1988),
2443–2458.
[173] P.-D. Hong, W.-T. Chuang, W.-J. Yeh, and T.-L. Lin, Effect of rigid
amorphous phase on glass transition behaviour of poly(trimethylene
[174] J. Lin, S. Shenogin, and S. Nazarenko, Oxygen solubility and specific
volume of rigid amorphous fraction in semicrystalline poly(ethylene
[175] M.C. Righetti, M.L. Di Lorenzo, M. Angiuli, and E. Tombari, Struc-
tural reorganization in poly(butylene terephthalate) during fusion,
Macromolecules 37 (2004), 9027–9033.
REFERENCES 225
[176] C. Alvarez, I. Sics, A. Nogalez, Z. Denchev, S.S. Funari, and

T.A. Ezquerra, Structure-dynamics relationship in crystallizing
poly(ethylene terephthalate) as revealed by time-resolved X-ray and
dielectric methods, Polymer 45 (2004), 3953–3959.
[177] R. Rastogi, W.P. Vellinga, S. Rastogi, C. Schick, and H.E.H. Meijer,
The three-phase structure and mechanical properties of poly(ethylene
terephthalate), Journal of Polymer Science, Part B: Polymer Physics
42 (2004), 2092–2106.
[178] B. Goderis, P.G. Klein, S.P. Hill, and C.E. Koning, A compara-
tive DSC, X-ray and NMR study on the crystallinity of isomeric
aliphatic polyamides, (accepted for publication).
[179] M. Pyda, E. Nowak-Pyda, J. Mays, and B. Wunderlich, Heat ca-
pacity of poly(butylene terephthalate), Journal of Polymer Science,
Part B: Polymer Physics 42 (2004), 4401–4411.
[180] B. Goderis, H. Reynaers, M.H.J. Koch, and V.B.F. Mathot, Use
of SAXS and linear correlation functions for the determination of
the crystallinity and morphology of semi-crystalline polymers. Ap-
plication to linear polyethylene, Journal of Polymer Science, Part B:
Polymer Physics 37 (1999), 1715–1738.
[181] H. Parton and I. Verpoest, Thermoplastic liquid composite molding:
Production and characterizations of composites based on cyclic oli-
gomers, Proceedings of the 7th International Conference on Flow
Processes in Composite Materials (Newark, Delaware, USA), 2004,
pp. 57–64.
[182] F. Desplentere, H. Parton, I. Verpoest, S. Voskamp, A. Holmberg,
and P.-A. Löfgren, Design and production of a fibre reinforced ther-
moplastic leaf spring by RTM like technique, Proceedings of the
Sampe Europe Conference (Paris), pp. 209–213.
[183] B.R. Gebart, Permeability of unidirectional reinforcements for
RTM, Journal of Composite Materials 26 (1992), no. 8, 1100–1133.
[184] G.R. Strobl and M. Schneider, Direct evaluation of the electron den-
sity correlation function of partially crystalline polymers, Journal of
Polymer Science: Polymer Physics Edition 18 (1980), 1343–1359.
[185] B. Goderis, H. Reynaers, R. Scherrenberg, V.B.F. Mathot, and
M.H.J. Koch, Temperature reversible transitions in linear polyethy-
lene studied by TMDSC and time-resolved, temperature-modulated
WAXD/SAXS, Macromolecules 34 (2001), 1779–1787.
[186] W. Van Pelt, Processing of intractable polymers using reactive sol-
vent, Ph.D. thesis, Polymer Technology, T.U.Eindhoven, 2001.
Curriculum Vitae
Personal information
Full name: Hilde Parton
Home address: Tulpenlaan 1, 3052 Oud-Heverlee (Blanden), Belgium
E-mail: hildeparton@yahoo.com
Date of birth: 10/11/1978

Place of birth: Leuven
Nationality: Belgian
Status: Married
Education
2001-2006 PHD IN MATERIALS SCIENCE
Department of Metallurgy and Materials Engineering
Faculty of Engineering
1996 - 2001 MASTER IN CHEMICAL ENGINEERING

(burgerlijk ingenieur scheikunde)
Department of Chemical Engineering
Faculty of Applied Sciences
Graduated cum magna laude (83%)
Master Thesis
completed in the Socrates framework at the IKV, RWTH, Aachen
Promotor: Prof. Dr.-Ing. Dr.-Ing. E.h. W. Michaeli
Co-Promotor: Prof. dr. ir. P. Moldenaers
Analysis of process-relevant parameters using the direct strand-
deposition process for long glass-fibre reinforced polymers
227
228 Curriculum Vitae
Publications
International journals
1. Parton, H. and Verpoest, I., In situ polymerization of thermoplastic
composites based on cyclic oligomers, Polymer Composites. 26(1):
p. 60-65. (2005)
2. Parton, H.; Baets, J.; Goderis, B.; Lipnik, P.; Devaux, J. and Ver-
poest, I., Properties of poly(butylene terephthalate) polymerized from
cyclic oligomers and its composites, Polymer. 46: p. 9871-9880.
(2005)
3. Parton, H.; Goderis, B.; Basiura, M.; Baets, J. and Verpoest, I.,
Time-resolved X-ray measurements during the isothermal polymeri-
sation and crystallisation of cyclic butylene terephthalate oligomers,
in preparation
Conference proceedings
First author
1. Parton, H. and Verpoest, I. , Reactive processing of textile reinforced
thermoplastics, in Proceedings of 14th International Conference on
Composite Materials. San Diego. Paper 1201 (2003).
2. Parton, H.; Baets, J.; Lipnik, P.; Devaux, J. and Verpoest, I., Liquid
moulding of textile reinforced thermoplastics, in Proceedings of 11th
European Conference on Composite Materials. Rhodos, Greece.
CDrom paper A069 (2004).
3. Parton, H. and Verpoest, I., Thermoplastic Liquid Composite Mold-

ing: Production and Characterization of Composites based on Cyclic
Oligomers, in Proceedings of 7th International Conference on Flow
Processes in Composite Materials. Newark, Delaware, USA. p. 57-
64 (2004).
229
230 Publications
Co-authored
1. Brast, K.; Parton, H. and Michaeli, W. Characterizing the Direct
Strand-Deposition Process to Manufacture Long-Fiber Reinforced Ther-
moplastic Components. in Proceedings World Compounding Congress
WCC 2000, Neuss/Dsseldorf, 2000, Kapitel Compounding Technol-
ogy
2. Verpoest, I.; Parton, H.; Desplentere, F.; Voskamp, S.; Holmberg,
A. and Löfgren, P.-A., Design and production of a fibre reinforced
thermoplastic leaf spring by RTM like technique, in Proceedings of
7th International Conference on Textile Composites. Yonezawa, Ya-
magata, Japan (2004).
3. Desplentere, F. and Parton, H. and Verpoest, I. and Voskamp, S.
and Holmberg, A. and Lofgren, P.-A., Design and production of a
fibre reinforced thermoplastic leaf spring by RTM like technique, in
Proceedings of Sampe Europe. Paris. p. 209-213 (2005)
4. Baets, J.; Parton, H.; Devaux, J. and Verpoest, I., Characteriza-
tion of continuous reinforced thermoplastics with in-situ polymer-
ized CBT (cyclic butylene terephthalate), in Proceedings of Theplac
2005, Lecce, Italy, 2005

In Situ Polimerization TP For Continuous Composites

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

In Situ Polimerization TP For Continuous Composites

Uploaded by

Copyright:

Available Formats

Katholieke Universiteit Leuven

Characterisation of the in-situ

Promotor: Proefschrift ingediend tot

Characterisation of the in-situ

Jury: Proefschrift voorgedragen tot

All rights reserved. No part of the publication may be reproduced in any

that things are very similar wherever you go.

Abstract The advantages of continuously reinforced thermoplastic com-

Samenvatting De voordelen van continu vezelversterkte thermoplasten

φ, (1 − φ) Crystalline or amorphous local volume fraction −

ρ Mass density kg/m3

ϕ, (1 − ϕ) Crystalline or amorphous overall volume fraction −

% Electron density e− /cm3

B Constant background factor a.u.

Ca∗ Modified capillary number −

D Lateral dimensions of semi-crystalline stacks Å

Dhkl Crystallite size in hkl-direction Å

Eabs Absorbed energy J

Eim Impact energy J

GIc Corrected crack propagation energy J/m2

GI Crack propagation energy J/m2

Mm Monomer molecular weight g/mole

Mn Number average molecular weight kg/mole

Mw Weight average molecular weight kg/mole

Os Specific surface area of the phase boundary Å2

P Porod constant a.u.

Q Heat flow W/g

Q Volumetric flow rate m3 /s

R Ideal gas constant, 8.31 J/(mole · K)

s Modulus of scattering angle Å

tpol Polymerisation time min

tstir Stirring time s

Vf Fibre volume fraction %

Vm Matrix volume fraction %

vs Superficial velocity m/s

S-B please refer to page 53

semi Processed semi-isothermally

-S Cooled down slowly

ED- Entropy driven

2.6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . 47

4 Materials and Methods 51

5 Composites: Production and Properties 71

6 Properties and morphology of pCBT 105

7 Time-resolved X-ray measurements 125

8 Conclusions and Outlook 167

A Production of a leaf spring prototype 171

B Principles of X-ray scattering 177

C Thermoplastic Polyurethanes 187

Nederlandstalige Samenvatting 195

Curriculum Vitae 227

Polymer composites are an interesting material class since they com-

ties is a major advantage of thermoplastics. One of the drawbacks is the

2.1 Manufacturing of continuous fibre rein-

2.1.1 Impregnation of fibrous reinforcement

2.1.2 Processing routes

Film Hybrid Powder Solvent In-situ

Final composite part

Figure 2.1: Processing routes for thermoplastic composites with

Facilitating impregnation usually results in the production of preforms

2.1.3 Facilitating impregnation

Film stacking In the film stacking process, alternate layers of (thin)

Figure 2.2: Different types of hybrid yarns (a) commingled, (b)

Hybrid yarns and fabrics The second method of intimate mingling