Ulf W. Gedde · Mikael S.

Hedenqvist
Minna Hakkarainen · Fritjof Nilsson
Oisik Das

Applied Polymer
Science
Applied Polymer Science
Ulf W. Gedde • Mikael S. Hedenqvist
Minna Hakkarainen • Fritjof Nilsson • Oisik Das

Applied Polymer Science

Ulf W. Gedde Mikael S. Hedenqvist
Fibre and Polymer Technology, Fibre and Polymer Technology,
KTH Royal Institute of Technology KTH Royal Institute of Technology
Stockholm, Sweden Stockholm, Sweden

Minna Hakkarainen Fritjof Nilsson

Fibre and Polymer Technology, Fibre and Polymer Technology,
KTH Royal Institute of Technology KTH Royal Institute of Technology
Stockholm, Sweden Stockholm, Sweden

Oisik Das
Structural and Fire Engineering Division
Department of Civil, Environmental
and Natural Resources Engineering
Luleå University of Technology
Luleå, Sweden

ISBN 978-3-030-68471-6 ISBN 978-3-030-68472-3 (eBook)

https://doi.org/10.1007/978-3-030-68472-3

# Springer Nature Switzerland AG 2021

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on
microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer
software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even
in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and
therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true
and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied,
with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains
neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

This book has a part of its roots in a textbook Polymer Physics by Ulf Gedde published in 1995. We
extend our gratitude to colleagues in our department who made contributions to this book: Maria
Conde Braña, Kristian Engberg, Anders Hult, Jan-Fredrik Jansson, Håkan Jonsson, Sari Laihonen,
Fredrik Sahlén, Marie Louise Skytt, G€oran Wiberg, Jens Viebke and Bj€ orn Terselius. Richard Jones
of the Cavendish Laboratory, University of Cambridge, UK, has read through all the chapters and
made some very constructive contributions. Mats Ifwarson, Studsvik Material AB, Sweden; Josef
Kubat, Chalmers University of Technology, Sweden; and Andrew Keller, University of Bristol, UK,
have also made contributions in their specialty fields.
Less than a year ago, Gedde and Hedenqvist completed a book entitled Fundamental Polymer
Science, which contained 11 chapters covering, as the title indicates, the fundamental aspects.
A second text of a more fundamental character entitled Essential Classical Thermodynamics was
completed a few months later by Gedde. Both these books have been published by Springer Nature.
Polymer Physics published in 1995 had chapters on thermal analysis, microscopy, spectroscopy and
scattering methods, which have been extensively updated and rewritten in this new book (Applied
Polymer Science). These fields have developed a great deal since 1995 and the more novel methods
have now been included. We felt before writing the book that chromatography and computer
simulation/modelling are extremely important fields and that these should also be included. This
completes the experimental method part of the book; please note that simulation is really an
experimental method although it uses a computer. To these five chapters, we have added three
chapters on mechanical properties, transport properties and polymer processing. Each of these
subjects forms an entire course at the KTH Royal Institute of Technology (Stockholm), both at the
master’s level and for the more advanced, postgraduate (Ph.D.) students. The beauty is to provide this
material together with the fundamental parts and ensemble of the advanced experimental methods to
the students. The motivation for the two remaining chapters is also strong: the environmental aspects
of polymers in the context of other materials used by mankind and finally, but not least, the solutions
to all the problems given as exercises at the end of each chapter. We thus proudly present three
strongly allied books, Fundamental Polymer Science, Essential Classical Thermodynamics and
Applied Polymer Science. They are linked by referencing in each chapter in each book to other
chapters in all the books. The style is essentially the same in all the books and we have indeed strived
to use a language of similar character in all the chapters. Some mathematical ‘language’ is necessary,
but we have kept it at a level suitable for students at technical universities. We have also prepared
texts that explain the mathematics in more detail, cf. a chapter in the Essential Classical Thermody-
namics book and as sections in the chapters in the other two books. The presence of solutions to the
large number of problems appearing in each chapter is important; we know that most students like to
solve problems, but it is good to provide solutions in case they lack the time to solve some of the
problems. This book has a wide range of possible readers: students in chemistry, physics, materials

v
vi Preface

science, biotechnology and civil engineering. It is suitable for master’s students, graduate (Ph.D.)
students, and engineers and scientists who have left academia for industry. Furthermore, the book is
rich in illustrations that will enable efficient knowledge transfer.
Except from one of the authors, our native language is not English, and although we have been
writing many scientific papers and several textbooks before this, we felt a strong need for the
assistance of two scientifically skilled Englishmen. Roger Brown (editor of Polymer Testing since
many years) and Anthony Bristow have corrected the written language and made it pleasant to read.
Anthony Bristow has been assisting us for 40 years and he is a dear friend to us with an outstanding
language ear and, just as important, a profound understanding of science in general and polymer
science in particular. Geddes’ younger son, Samuel Gedde, has proven very skilled in making much
of the graphics using Adobe Illustrator.
This book would not have been possible without the help of other scientists who provided
inspiration, direct input before and after reading the text, and micrographs: Richard H. Boyd
(a principal mentor of Gedde and Hedenqvist; whose impact is great although he has been dead for
several years); David Bassett (a gentleman with insightful knowledge of semicrystalline polymer
morphology, who also provided beautiful micrographs); Bernard Lotz (scientific inspiration and with
vast knowledge on polymer physics and crystallography, providing feedback on polymer microscopy
and some beautiful micrographs); Grant D. Smith (a master of relaxation processes and computer
simulation); Henning Kausch (inspiration and input concerning polymer mechanics and fracture);
Karin Odelius (who provided input concerning the chromatographic analysis of polymers);
Vigneshwaran Shanmugam (a young but talented researcher who aided Oisik Das in revising the
polymer processing chapter); Elaine Espuche and Matteo Minelli (for the constructive inputs to the
transport property chapter); Mats Johansson, Juha Heikanen, Yuxiao Cui and Tomas Larsson (for the
valuable comments on the spectroscopy section of Chap. 3); Istvan Furo and Tore Brinck (providing
deep knowledge about quantum chemistry); Henrik Hillborg (giving insight in recent applications of
statistical methods); Jakob Wohlert (reviewing the molecular mechanics sections); Mikael Hancke
and Yuriy Serdyuk (giving valuable comments on finite element modelling); and Tohr Nilsson for the
proofreading of simulation chapter.
We are indebted to David Packer, the publisher at Springer Nature. He has been extremely
supportive in the process of preparing the books and has provided many great ideas, such as the
suggestion that we should write this companion book on applied polymer science. David is an
extremely experienced and knowledgeable publisher, and it has been an honour to work together
with him! We are indebted to Springer Nature for their patience in waiting for the manuscript to arrive
and for performing an excellent job in transforming the manuscript to this pleasant form. We are also
grateful to our families for their support during the writing process.
This book was produced during very difficult conditions with the Covid-19 pandemic spread
across the entire world. The close collaboration between us (the authors), Springer Nature (David
Packer, Zachary Evenson, Barbara Amorese, Jeffrey Taub and Amrita Unnikrishnan) and Sandrine
Pricilla Anthonisamy and her team at Straive made possible the production of this book. We are so
grateful to you for the careful work transforming our text and graphics into the final book.

Stockholm, Sweden Ulf W. Gedde

Mikael S. Hedenqvist
Minna Hakkarainen
Fritjof Nilsson
Oisik Das
Contents

1 Thermal Analysis of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Thermo-analytical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.1 Differential Thermal Analysis and Differential Scanning Calorimetry . . . . 6
1.2.2 Thermogravimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2.3 Dilatometry and Thermomechanical Analysis . . . . . . . . . . . . . . . . . . . . . 16
1.2.4 Dynamic Mechanical Thermal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.2.5 Thermal Optical Analysis and In Situ Structural Assessment Under
Controlled Thermal History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.2.6 Dielectric Thermal Analysis (DETA) . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.3 Thermal Behaviour of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.3.1 Semicrystalline Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.3.2 Amorphous Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.3.3 Liquid Crystalline Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
1.3.4 Polymer Degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
1.3.5 Further Applications of Thermal Analysis in Polymer Science and
Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
1.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
1.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2 Microscopy of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.2 Optical Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.2.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.2.2 Polarized Microscopy and Related Techniques . . . . . . . . . . . . . . . . . . . 67
2.3 Electron Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.4 Atomic Force Microscopy and Related Techniques . . . . . . . . . . . . . . . . . . . . . . 73
2.5 Novel Techniques in Polymer Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.6 Preparation of Specimens for Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
2.6.1 Preparation of Samples for Optical Microscopy . . . . . . . . . . . . . . . . . . . 79
2.6.2 Preparation of Samples for Scanning Electron Microscopy . . . . . . . . . . . 79
2.6.3 Preparation of Samples for Transmission Electron Microscopy . . . . . . . . 79
2.6.4 Preparation of Samples for Atomic Force Microscopy . . . . . . . . . . . . . . 84
2.6.5 Artificial Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

vii
viii Contents

2.7 Applications of Microscopy in Polymer Science and Engineering . . . . . . . . . . . 85

2.7.1 Semicrystalline Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
2.7.2 Liquid Crystalline (LC) Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.7.3 Polymer Blends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.7.4 Composites Including Nanocomposites . . . . . . . . . . . . . . . . . . . . . . . . . 95
2.7.5 Native Polymers and Polymeric Biomaterials . . . . . . . . . . . . . . . . . . . . 99
2.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
2.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3 Spectroscopy and Scattering of Radiation by Polymers . . . . . . . . . . . . . . . . . . . . . 105
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.2 Vibrational Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.2.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.2.2 IR and Raman Spectrophotometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
3.2.3 Quantitative Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
3.2.4 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.2.5 Applications of Vibrational Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . 116
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 123
3.3.1 Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.3.2 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
3.3.3 Polymer Applications of NMR Spectroscopy . . . . . . . . . . . . . . . . . . . . . . 130
3.4 Other Spectroscopic Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.4.1 X-ray Photoelectron Spectroscopy (XPS) . . . . . . . . . . . . . . . . . . . . . . . . 139
3.4.2 Electron Spin Resonance Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.4.3 UV-VIS Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
3.5 Scattering Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
3.5.1 Light Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
3.5.2 Wide-Angle X-ray Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
3.5.3 Small-Angle X-ray Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
3.5.4 Electron Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
3.5.5 Neutron Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
3.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
4 Chromatographic Analysis of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
4.2 General Concepts in Chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.3 Size Exclusion Chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
4.3.1 Molar Mass and Molar Mass Distribution . . . . . . . . . . . . . . . . . . . . . . . 176
4.3.2 Principles of SEC Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
4.3.3 Separation in the SEC Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
4.3.4 Detection of the Eluting Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
4.3.5 Relative Narrow Standard Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4.3.6 Universal Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
4.3.7 Sample Preparation for SEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
4.3.8 Molar Mass Averages from SEC Chromatograms . . . . . . . . . . . . . . . . . 185
4.3.9 Development of New Coupled SEC Systems . . . . . . . . . . . . . . . . . . . . . 186
Contents ix

4.4 High-Performance Liquid Chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

4.4.1 Principles of an HPLC System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
4.4.2 Separation Modes in HPLC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
4.5 Gas Chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
4.5.1 Principles of a GC System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
4.5.2 Integrating GC with Mass Spectrometry (MS) . . . . . . . . . . . . . . . . . . . . 192
4.6 Qualitative and Quantitative Analysis by HPLC and GC . . . . . . . . . . . . . . . . . . 193
4.7 Sample Preparation Before GC or HPLC Analysis . . . . . . . . . . . . . . . . . . . . . . 194
4.8 Application of GC and HPLC for the Analysis of Polymers . . . . . . . . . . . . . . . 197
4.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
4.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
5 Simulation and Modelling of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
5.2 Quantum Chemistry (QC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
5.2.1 QC: Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
5.2.2 Computational Quantum Mechanics: Formalism . . . . . . . . . . . . . . . . . . 207
5.2.3 Molecular Orbital (MO) Methods, One-Electron Technique . . . . . . . . . . 210
5.2.4 Molecular Orbital (MO) Methods, Many-Electron Technique . . . . . . . . . 213
5.2.5 Semi-empirical MO Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
5.2.6 Ab Initio Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
5.2.7 Density Functional Theory (DFT) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.3 Molecular Dynamics (MD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
5.3.1 MD: Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
5.3.2 Force-Field Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
5.3.3 Coarse-Graining . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
5.3.4 The Basic MD Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
5.3.5 Ensembles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
5.3.6 The Phase Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
5.3.7 A Practical MD Example: Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
5.4 Monte Carlo Methods (MC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
5.4.1 MC: Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
5.4.2 Importance Sampling: Metropolis Hastings and Biased Sampling . . . . . . 236
5.4.3 Macromolecular Starting Configurations with MC . . . . . . . . . . . . . . . . . 239
5.4.4 Macromolecular Energy Minimization with MC . . . . . . . . . . . . . . . . . . 241
5.5 Mesoscale Modelling, Including Dissipative Particle Dynamics . . . . . . . . . . . . . 242
5.6 Statistical Methods, Including Group-Contribution Methods . . . . . . . . . . . . . . . 243
5.7 The Finite Element Method (FEM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
5.7.1 FEM: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
5.7.2 Applied FEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
5.7.3 FEM: Mathematical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
5.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
5.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6 Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
6.2 Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
6.2.1 Normal stress and Shear Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
x Contents

6.2.2 The Stress Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273

6.2.3 The Mohr Stress Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
6.3 Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
6.3.1 Introduction: Uniaxially Loaded Specimens . . . . . . . . . . . . . . . . . . . . . 278
6.3.2 The Strain Tensor and Other Strain Concepts . . . . . . . . . . . . . . . . . . . . 282
6.3.3 The Mohr Strain Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
6.4 Assessment of the Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
6.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
6.4.2 Tensile Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
6.4.3 Dynamic Mechanical Analysis, Dilatometry and Thermal
Mechanical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
6.4.4 Fracture Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
6.4.5 Surface Mechanics Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
6.5 Definition of Mechanical Parameters from the Tensile Test . . . . . . . . . . . . . . . . 290
6.5.1 The Stress–Strain Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290
6.5.2 The Stiffness and the Tensile Modulus . . . . . . . . . . . . . . . . . . . . . . . . . 291
6.6 The Three-Dimensional View for Modulus and Strain . . . . . . . . . . . . . . . . . . . 292
6.7 Energy During Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
6.7.1 Energy Stored During Elastic Deformation . . . . . . . . . . . . . . . . . . . . . . 296
6.7.2 Energy Transformed During Viscous Flow and Viscoelastic
Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
6.8 Oriented Polymers and Multiaxial Stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
6.9 Linear Viscoelasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
6.9.1 The Spring and the Dashpot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
6.9.2 The Maxwell (iso-Stress) Element . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
6.9.3 The Voigt–Kelvin (iso-Strain) Element . . . . . . . . . . . . . . . . . . . . . . . . . 306
6.9.4 The Burgers Element and Other Linear Viscoelastic Models . . . . . . . . . 308
6.10 Correlations Between Stress Relaxation and Dynamic Mechanical Data . . . . . . . 311
6.10.1 From Stress Relaxation Modulus to Dynamic Mechanical Modulus . . . 311
6.10.2 From Dynamic Mechanical Modulus to Dynamic Mechanical
Compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
6.11 Boltzmann Superposition Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
6.12 The Influence of Strain Rate on the Viscoelastic Properties . . . . . . . . . . . . . . . . 316
6.13 The Influence of Temperature on the Viscoelastic Properties:
Time–Temperature Shifting and Superposition . . . . . . . . . . . . . . . . . . . . . . . . . 316
6.14 Non-linear Viscoelasticity as Illustrated with Creep Behaviour . . . . . . . . . . . . . 318
6.15 Short-Term Mechanical Properties of Selected Polymers . . . . . . . . . . . . . . . . . 322
6.16 Rubber Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
6.17 Some Examples of Parameters Affecting the Mechanical Properties . . . . . . . . . 327
6.18 Yielding of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
6.18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
6.18.2 Theories of Yielding in Amorphous Polymers . . . . . . . . . . . . . . . . . . . 334
6.18.3 Yielding in Semicrystalline Polymers . . . . . . . . . . . . . . . . . . . . . . . . . 335
6.19 Fracture of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
6.19.1 Introduction and the Brittle–Ductile Transition . . . . . . . . . . . . . . . . . . 339
6.19.2 Fracture Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
6.19.3 Ductile Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
6.19.4 Creep Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
6.19.5 Impact Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
6.19.6 Dynamic Fatigue Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
Contents xi

6.20 Cellulose Fibre Systems and Related Materials . . . . . . . . . . . . . . . . . . . . . . . . . 363

6.21 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
6.22 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
7 Transport Properties of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
7.2 Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
7.2.1 Basics of the Random Walk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
7.2.2 The Velocity Autocorrelation Function . . . . . . . . . . . . . . . . . . . . . . . . . 381
7.2.3 Diffusion; Temperature and Solute Size . . . . . . . . . . . . . . . . . . . . . . . . 383
7.3 Solubility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386
7.4 Fick’s Laws of Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
7.5 Methods of Solution of the Diffusion Equation . . . . . . . . . . . . . . . . . . . . . . . . . 395
7.6 Transport Properties of Elastomers and Melts . . . . . . . . . . . . . . . . . . . . . . . . . . 400
7.6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
7.6.2 Models of Diffusion in Elastomers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
7.7 Transport Properties of Semicrystalline Polymers . . . . . . . . . . . . . . . . . . . . . . . 412
7.7.1 Geometrical Effects of Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412
7.7.2 Molecular Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
7.8 Concentration-Dependent Diffusivity and Swelling in Flexible Polymers . . . . . . 418
7.9 Transport Properties of Glassy Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
7.9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
7.9.2 Dual Solubility Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
7.9.3 The Dual Mobility Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422
7.9.4 Anomalous and Case II Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
7.10 Barriers and Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
7.10.1 Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
7.10.2 Membrane Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
7.11 Techniques for Measuring Permeability, Diffusivity and Solubility . . . . . . . . . . 433
7.12 Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
7.12.1 The Heat Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
7.12.2 Surface Boundary Heat Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 441
7.13 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
7.14 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448
8 Processing of Polymeric Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
8.2 Polymer Processing: A Complex Applied Polymer Science Discipline . . . . . . . . 458
8.3 Compounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
8.4 Injection Moulding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461
8.5 Extrusion and Associated Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
8.6 Thermoforming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471
8.7 Producing Hollow Objects: Rotational Moulding and Blow Moulding . . . . . . . . 472
8.8 Compression Moulding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 474
8.9 Calendering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
8.10 Additive Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
8.10.1 Steps in the Development of Digital Model . . . . . . . . . . . . . . . . . . . . . 478
8.10.2 3D Printing Methods for Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
xii Contents

8.11 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483

8.12 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
9 Plastics and Sustainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
9.2 Contribution of Plastics to Sustainable Society . . . . . . . . . . . . . . . . . . . . . . . . . . 489
9.3 Bio-based Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
9.3.1 Thermoplastic Materials Derived from Bio-based Monomers . . . . . . . . . . 493
9.3.2 Biopolymer-Based or Biopolymer-Derived Thermoplastics . . . . . . . . . . . 494
9.3.3 Bio-based Thermosets and Coatings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494
9.3.4 Production of Polymers from Waste–Greenhouse Gases . . . . . . . . . . . . . . 495
9.3.5 Challenges and Opportunities for Bio-based Materials . . . . . . . . . . . . . . . 495
9.4 End-of-Life Management: From Waste to Resource . . . . . . . . . . . . . . . . . . . . . . 496
9.4.1 Polymer Loop: Mechanical Recycling . . . . . . . . . . . . . . . . . . . . . . . . . . . 497
9.4.2 Monomer and Molecule Loops: Chemical Recycling . . . . . . . . . . . . . . . . 499
9.4.3 Additive Manufacturing as Enabling Tool for Plastic Recycling . . . . . . . . 499
9.4.4 Dynamic Covalent Chemistry as Enabling Tool for Plastic Recycling . . . . 500
9.5 Degradation of Plastics in Natural Environments . . . . . . . . . . . . . . . . . . . . . . . . 500
9.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
9.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
10 Solutions to Problems Given in Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
10.1 Chapter 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
10.2 Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
10.3 Chapter 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
10.4 Chapter 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
10.5 Chapter 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
10.6 Chapter 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519
10.7 Chapter 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
10.8 Chapter 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
10.9 Chapter 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
Chapter 1
Thermal Analysis of Polymers

1.1 Introduction

According to the definition originally proposed in 1969 by the Nomenclature Committee of the
International Confederation for Thermal Analysis (ICTA; renamed in 1992 as the International
Confederation for Thermal Analysis and Calorimetry, ICTAC) and later reaffirmed in 1998
(Hemminger 1998; Hemminger and Sarge 1998), thermal analysis includes the analytical methods
by which a physical property of a substance is measured as a function of temperature while the
substance is subjected to a controlled temperature program. Thermal analysis involves a physical
measurement and not, strictly speaking, an analysis. Figure 1.1 presents a summary of the different
thermal analytical methods available. This chapter includes differential scanning calorimetry (DSC)
and differential thermal analysis (DTA), thermogravimetry (TG), thermomechanical analysis (TMA,
DMTA), thermal optical analysis (TOA), thermal analysis focussing on structure (atomic force
microscopy (AFM) and wide-angle X-ray scattering (WAXS)) and dielectric thermal analysis
(DETA). Section 1.2 describes the different methods. The use of thermal analysis on polymers
requires special attention, as is discussed in Sect. 1.3. Examples from the melting and crystallization
of flexible-chain polymers, the glass transition of amorphous polymers, phase transitions in liquid
crystalline polymers and chemical reactions including degradation are presented to illustrate the
complex behaviour of polymers.
Thermo-analytical methods are powerful. The range of phenomena that can be studied by thermal
analysis is impressive. Only small amounts (typically milligrams) of sample are required for the
analysis. Calorimetric methods record exo- and endothermic processes, i.e. first-order phase
transitions such as melting and crystallization and also most chemical reactions, such as polymeriza-
tion, cross-linking, depolymerization and degradation. DSC, TMA and DMTA also reveal second-
order phase transitions, including the glass transition. Figures 1.2 and 1.3 show the characteristic
features of first- and second-order phase transitions. Further information about the strict definitions of
transitions according to Ehrenfest are presented by Gedde (2020). First-order transitions are
characterized by distinct changes (jumps) in enthalpy (H ), entropy (S) and volume (V), whereas
the chemical potential (μ) is continuous with an abrupt change in its temperature derivative (i.e. in the
molar entropy). Second-order phase transitions show continuity in the primary parameters (H, S and
V ), but the temperature derivatives show a distinct change associated with the transition., i.e. the
volume expansion coefficient (α ¼ (1/V )(∂V/∂T)p) and heat capacity (Cp ¼ (∂H/∂T )p show jumps
(Δα and ΔCp). The second temperature derivative of the chemical potential changes sign at the
transition temperature for the second-order phase transition.

# Springer Nature Switzerland AG 2021 1

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_1
2 1 Thermal Analysis of Polymers

Fig. 1.1 Thermo-analytical methods

Fig. 1.2 Characteristics of first-order phase transitions showing volume (V ), entropy (S), enthalpy (H ) and chemical
potential (μ) as functions of temperature (T ). (From Gedde (2020) with permission from Springer Nature)

Fig. 1.3 Characteristics of second-order phase transitions showing volume (V ), entropy (S), enthalpy (H ) and chemical
potential (μ) as functions of temperature (T ). (From Gedde (2020) with permission from Springer Nature)

The behaviour presented in Figs. 1.2 and 1.3 is idealized; in reality, the thermal behaviour of
polymers is more complicated. The lines displayed in Figs. 1.2 and 1.3 are equilibrium lines, i.e. at
each temperature, the system is at equilibrium, which means a low free energy state. The envisioned
path is thus characteristic of a reversible process. Characteristic of a reversible path is that the entropy
production (ΔSi) is zero. The entropy production is the sum of the entropy changes of the given
system (ΔSsyst) undergoing the phase transition and of the surroundings (ΔSsurr), the latter change
being a direct consequence of the changes within the system. A spontaneous irreversible process
shows according to Clausius (1854) a positive ΔSi. Consider a simple case: liquid water at 253 K is
spontaneously converted into ice, which is an irreversible process with a positive entropy production.
The entropy change of the system is calculated by considering a fully reversible path: the reversible
isobaric heating of water from 253 K to 273 K; the phase transformation at the equilibrium melting
temperature (273 K) – note that chemical potentials of water and ice are equal at 273 K and 1 atm
pressure; and finally the reversible isobaric cooling of ice from 273 K to 253 K:
1.1 Introduction 3

ð
273 ð
253
Cp,1 dT Δ H cryst Cp,s dT
Δ Ssyst ¼ þ þ ð1:1Þ
T 273 T
253 273

where Cp,l and Cp,s are the heat capacities of water and ice, respectively. Note that both ΔHcryst and
ΔSsyst are negative. The crystallization process at 253 K is irreversible, although the heat produced
(being equal to ΔHmelt (253 K)) is transferred in a reversible fashion to the surroundings:
ΔH melt ð253 KÞ
ΔSsurr ¼ ð1:2Þ
253

where ΔHmelt (253 K) is calculated from the enthalpy of melting at 273 K by applying Kirchhoff’s
law (cf. Gedde 2020). Note that both ΔHmelt (253 K) and ΔSsurr are positive. The entropy production
(ΔSi) is:
ΔSi ¼ ΔSsyst þ ΔSsurr ¼
ð
273 ð
253
Cp,l dT ΔH cryst Cp,s dT ΔH melt ð253 KÞ ð1:3Þ
þ þ þ >0
T 273 T 253
253 273

A comprehensive description of the second law of thermodynamics and entropy production is

presented by Gedde (2020). Glassy polymers and semicrystalline polymers both show marked
deviations from the equilibrium state. Figure 1.4 illustrates the behaviour of polymer glasses. An

Fig. 1.4 Volume (V ) or enthalpy (H ) of a fully amorphous polymer as functions of temperature (T ) at constant
pressure. Line a: cooling the liquid, which undergoes a kinetic glass transition. Line b: heating the glass at the same rate
as it was vitrified; the path (V or H ¼ f (T ) follows essentially the same track as during cooling. Line c: heating the
vitrified glass according to path A at a faster rate than during the previous cooling. Note the superheating. Line d:
heating the vitrified glass according to path A at a slower rate than during the previous cooling. Note the decrease in
excess volume and in excess enthalpy occurring close to the fictive temperature. The equilibrium line is shown as a
broad line, and the underlying thermodynamic glass transition is indicated
4 1 Thermal Analysis of Polymers

equilibrated liquid which is cooled at a constant rate shows a decrease in both V and H following the
equilibrium line. At a certain temperature, depending on the cooling rate, both V and H start to show a
marked deviation from the equilibrium line, and instead they follow what is referred to as the glassy
line. This change of behaviour is a kinetic phase transition. The formation of a glass from a liquid is
denoted vitrification. Most glasses show an excess volume and an excess enthalpy with respect to the
equilibrium low free energy state. When the glass is heated at the same rate as it was cooled during
vitrification, it mimics the path in V–T and H–T that it followed during cooling. The temperature at
which the glass line intersects the equilibrium line is denoted the fictive temperature (Fig. 1.4). If the
glassy structure is heated at a faster rate, the first part of the process follows the same path as during
the heating at a slower rate (path c resembles path b), but the crossover to the equilibrium line is
delayed due to superheating. The fictive temperature is nevertheless the same as in case b. If the
glassy structure is heated more slowly, the first part is the same as during heating at the faster rate
(path d resembles path b), but at temperatures just below the kinetic glass transition temperature, the
glassy structure undergoes a change often referred to as volume recovery (a decrease in volume) and
enthalpy recovery (a decrease in enthalpy). Even here, the fictive temperature is the same as in cases b
and c. The invariant in the three heating scan experiments (denoted b–d in Fig. 1.4) is the fictive
temperature. Hence, the fictive temperature is a suitable measure of the structure of the glassy
material. A low fictive temperature indicates a structure with only little excess volume and little
excess enthalpy. However, even though all recorded glass transitions are of the kinetic type, it was
early realized that a thermodynamic second-order glass transition must exist (Fig. 1.4; cf. Gedde and
Hedenqvist 2019a).
Folded-chain polymer crystals deviate markedly from the equilibrium, minimum free energy
structure which, for a homopolymer with a regular polymer chain structure, is an extended chain
crystal. Figure 1.5a shows that the thickness of virgin crystals (Lc*) decreases with increasing degree
of supercooling (ΔT ¼ Tm0 – Tc. where Tm0 is the equilibrium melting temperature and Tc is the
crystallization temperature):
C1
L∗
c ¼ þ δLc ð1:4Þ
ΔT

where C1 and δLc are constants. This means that crystals formed at low temperatures (high ΔT ) are
thin, whereas crystals formed at high temperatures (i.e. low ΔT ) are thick. During prolonged
crystallization at a given temperature, it is possible that crystals grow in thickness, i.e. undergo
crystal thickening, particularly at high temperatures. Further details about crystal thickening and the
relationship between Lc* and ΔT are presented by Gedde and Hedenqvist (2019b). The
Gibbs–Thomson equation relates melting temperature (Tm) and crystal thickness (Lc) according to:
 
2σ f T 0 1
T m ¼ T 0m  0 m  ð1:5Þ
Δhm ρc Lc

where σ f is the fold surface energy, Δhm0 is the enthalpy of melting of 100% crystalline polymer and
ρc is the crystal density. Figure 1.5b shows Gibbs–Thomson data for linear polyethylene. Further
details about the Gibbs–Thomson equation including its derivation and history are presented by
Gedde and Hedenqvist (2019b).
Figure 1.5c shows different paths depending on the heating rate: fast heating means that the time
available for structural changes driven by lowering of the free energy (i.e. crystal thickening and
perfection) is so short that essentially no change occurs prior to melting. This is referred to as zero-
entropy production path melting, a term coined by Wunderlich (1964). Heating more slowly enables
structural rearrangements of the crystals to take place prior to the final melting. This reorganization
path yields crystals which are thicker and have a higher melting temperature than the original crystals
(red path in Fig. 1.5c). Obviously the melting according to the zero-entropy production path provides
the most relevant information about the original crystal structure.
1.1 Introduction 5

Fig. 1.5 (a) Virgin crystal thickness (Lc*) of linear polyethylene as a function of the reciprocal degree of supercooling
(ΔT ). Drawn after data collected by Ramos et al. (2018). From Gedde and Hedenqvist (2019b) with permission from
Springer Nature. (b) Melting temperature as a function of reciprocal crystal thickness (Lc) for linear polyethylene.
Drawn after data collected by Wunderlich (1980). From Gedde and Hedenqvist (2019b) with permission from Springer
Nature. (c) Chemical potential (μ; lines are in blue) as a function of temperature for a crystal of thickness Lc, for a crystal
of infinite thickness (denoted 1) and for the liquid phase (in green). With rapid heating, the sample follows path zepp,
i.e. zero-entropy production path (marked in black). During slower heating (path reorg), sufficient time is given to
induce changes in the structure (crystal thickening) prior to the final melting (marked in red)

DMTA and DETA are used to study secondary transitions (sub-glass processes). DSC and TMA
sensitively determine thermodynamic quantities, e.g. heat capacity, enthalpy of melting and thermal
expansion coefficient. Modulated DSC is a relatively new technique, which gives a better separation
of overlapping endo- and exothermic processes. TG is a valuable tool for the determination of the
content of volatile species and fillers in polymeric materials and also for studies of polymer
degradation. TG can also be used to assess transport properties such as diffusivity, solubility and
permeability. Morphological information providing insight into the nature of the phase transitions in
both crystalline and liquid crystalline polymers can be obtained by TOA and by combination with
modern atomic force microscopy (AFM) and X-ray scattering techniques (WAXS).
A more recent trend is to combine instruments ‘on line’ in order to gain more information about
structure and processes. Combined TG and GC/MS provide direct information about the gases
evolved from a polymeric material exposed to a degrading atmosphere. Fast scanning calorimetry,
an emerging discipline since 2000, provides new opportunities to study polymer phase transitions
over a much wider temperature range than before.
6 1 Thermal Analysis of Polymers

1.2 Thermo-analytical Methods

It is not the purpose of this chapter to give full details of thermo-analytical methods. In general, it is
always good to start a comprehensive study by using DSC, which is versatile, simple and time-
efficient and provides a wealth of data that can help the interpretation of information given by the
more specialized techniques.

1.2.1 Differential Thermal Analysis and Differential Scanning Calorimetry

Calorimetry is the discipline by which heat is measured, the amount of heat (in joules) that is required
to increase the temperature by, e.g. 1 K in a certain amount of substance. The first calorimeters
(a term coined by Lavoisier) were designed by the Scottish physician and chemist Joseph Black in
1760 (Ramsey 1918) and in an improved version by the French scientist Antoine-Laurent Lavoisier
(1789). These instruments used samples with substantial mass. Accurate temperature-measuring
components which are of great importance for quality calorimetric measurements – thermocouples,
resistance thermometers and optical pyrometers – were developed in the late nineteenth century.
Thermometry is the simplest and oldest method of thermal analysis, where a sample is heated at a
constant heat flow rate. Any first-order phase transition is recorded as an invariance in temperature
during the constant rate of supply of heat to the system. Henry Le Chatelier (1887) used such
instruments in the late nineteenth century on chemical systems to study changes during the heating
of clay. The first DTA method was conceived 2 years later by the English metallurgist William
Chandler Roberts-Austen. Stone (1951) introduced the modern DTA equipment using an apparatus
which permitted the flow of gas through the sample during the temperature scan. The undesirable
feature of the classical DTA (Fig. 1.6b) in sensor-sample interactions was overcome by Boersma
(1955), and this type of instrument has since been referred to as ‘Boersma DTA’ (Fig. 1.6c). The
temperature sensors are placed outside the sample and the reference.
A differential scanning calorimeter is basically an instrument which accurately measures the
calorimetric properties of a sample. A DSC based on the DTA principle is referred to as a heat-flux
differential scanning calorimeter (Fig. 1.6c). Such instruments measure the difference in temperature
(ΔT) between sample and reference and convert ΔT to absorbed or evolved heat via a mathematical
procedure. The conversion factor is temperature-dependent. There are two main types of heat-flux
DSC instruments. The most common type uses a metal disc on which the sample and the reference are
positioned and the temperatures of the two are accurately measured. The second type consists of two
‘heavy’ cylindrical containers for sample and reference. The cylinders are in good thermal contact
with a heated furnace, and the temperatures are measured in the cylinders. Characteristic of the heat
flux DSC instruments is that the sample and reference are heated by a single heat source. Temperature
sensors are positioned close to the sample and the reference. Watson et al. (1964) introduced the first
power-compensation differential scanning calorimeter, the main elements of which are shown in
Fig. 1.6a. Power-compensation DSC relies on the ‘null-balance’ principle (Fig. 1.7), where the
temperature of the sample holder is kept the same as that of the reference holder by continuous and
automatic adjustment of the heater power. The sample and reference holders are individually heated,
and a signal proportional to the difference between the heat power input to the sample and to the
reference, dH/dt, is recorded.
Several new developments in the instrumentation have been made during the last decades. The
temperature scan is controlled, and data are collected and analyzed by computers in the modern
instruments. Instruments allowing the simultaneous measurement of differential temperature (ΔT)
and sample mass, i.e. combined heat-flux DSC and TG as well as combined heat-flux DSC and TOA,
1.2 Thermo-analytical Methods 7

Fig. 1.6 Schematic

representation of (a)
power-compensation DSC,
(b) ‘classical’ DTA and (c)
‘Boersma’ DTA, i.e. heat-
flux DSC. S and R denote
respectively sample and
reference

Fig. 1.7 Principle of power-compensation DSC

are now commercially available. Pressurized DSC instruments have been in use since the late 1960s.
Davidson and Wunderlich (1969) constructed a DTA cell, which could be used at pressures up to
500 MPa, and Blankenhorn and H€ohne (1991) constructed a high-pressure DSC operating between
20 and 300  C at a maximum pressure of 550 MPa. DSC instruments equipped with a UV cell were
developed during the 1980s.
A new class of highly sensitive (100 nW) isothermal or slowly scanning calorimeters to be used at
temperatures lower than 120  C was developed by Calvet and Prat (1963), Benzinger and Kitzinger
(1963) and Wads€o and co-workers (cf. Suurkuusk and Wads€ o 1982) starting in the 1960s. The first
8 1 Thermal Analysis of Polymers

Fig. 1.8 Schematic DSC

traces showing three
transition types

instrument in the series was a batch reaction calorimeter. These calorimeters consist of a calorimetric
vessel surrounded by thermopiles, often Peltier elements, through which the heat is conducted to or
from the surrounding heat sink.
Figure 1.8 shows the DSC thermogram of an undercooled, crystallizable polymer and illustrates
the principle of the power-compensation DSC. At low temperatures, there is no change in the heat
flow balance between the reference and the sample, and the DSC trace remains flat with no jumps.
When the glass transition is reached, an increase in the (endothermic) heat flow to the sample is
required in order to maintain the two at the same temperature. The change in level of the scanning
curve is thus proportional to the change in heat capacity (ΔCp) accompanying the glass transition. The
polymer crystallizes at a higher temperature, and exothermal energy is evolved, so that the heat flow
to the sample in this temperature region is less than the heat flow to the reference. The integrated
difference between the two, i.e. the area under the exothermal peak, is thus equal to the crystallization
enthalpy. At an even higher temperature, melting, which is an endothermic process, occurs. The heat
flow to the sample is then higher than that to the reference, and the peak points upwards. The area
under the endothermal peak is thus proportional to the melting enthalpy. The same information is also
obtained from the heat-flux DSC. At temperatures below the glass transition temperature, the sample
and the reference have only a small difference in temperature (ΔT ¼ Tr – Ts, where Tr is the reference
temperature and Ts is the sample temperature). The glass transition leads to an increase in the heat
capacity of the sample and a small increase in ΔT. Crystallization is exothermic and ΔT becomes
negative. The temperature difference between scanning baseline and the recorded value reflects the
heat evolved in the process. Finally, during melting (an endothermic process), ΔT is positive. It should
be noted that most heat-flux DSC software plots the differential heat flow rate as calculated from Tr –
Ts (according to Eq. (1.3)).
The following equation gives the evolved heat flow (i.e. dH/dt) in the heat-flux DSC:
dH dT dðT s  T r Þ
RT ¼ ðT s  T r Þ þ RT ðCs  Cr Þ r þ RT Cs ð1:6Þ
dt dt dt

where the subscripts ‘s’ and ‘r’ refer respectively to the sample and the reference, Ts and Tr are the
temperatures, Cs and Cr are the heat capacities and RT is the thermal resistance between the block and
the sample cell. The enthalpy change (ΔH) involved in an endothermic (or exothermic) process can be
derived from Eq. (1.6) under the assumption that Cs  Cr:
1
dH ¼ ΔTdt þ Cs dðΔT Þ ð1:7Þ
RT

which after integration yields:

1.2 Thermo-analytical Methods 9

ð
1
1 S
ΔH ¼ ΔTdt ¼ ð1:8Þ
RT RT
0

where ΔT ¼ Ts – Tr and S is the area under the ΔT-peak. The calibration factor (1/RT) is dependent on
the temperature, and this means that the heat-flux DSC has to be calibrated by several standards at
different temperatures.
An equation analogous to Eq. (1.6) can be obtained for the power-compensation DSC:

dH dq dT d2 q
¼  þ ðCs  Cr Þ r  RT Cs 2 ð1:9Þ
dt dt dt dt

where dq/dt is the differential heat flow. This equation can be integrated to yield the ΔH of a first-
order phase transition:
q ¼ ΔH ð1:10Þ

The calibration factor to transform peak surface area to enthalpy is independent of the thermal
resistance (1/RT) and is simply an electrical conversion factor without any temperature dependence.
Modern heat-flux and power-compensation DSC instruments both provide useful data. The power-
compensation instruments have some advantages, however, which are helpful in some applications.
The small mass of the furnace and the short heat conduction path in the power-compensation DSC
instruments allow a very fast system response to a sample process. A power-compensation DSC is
more powerful in resolving overlapping calorimetric processes than a heat-flux DSC, and the
possibility of using modulated DSC (see end of this section) increases the possibility of resolving
overlapping phase transitions in modern heat-flux DSC equipment. The maximum scanning rate is
also higher in the power-compensation systems than in the heat-flux instruments. This facilitates the
study of phase transitions under isothermal conditions after a temperature jump, although the
temperature range covered by the heat-flux DSC instruments is wider, from 80 K to 1800 K, than
that of the power-compensation DSC instruments (100–1000 K). Hemminger (1994) and H€ ohne
(1994) made detailed comparisons between heat-flux and power-compensation DSC instruments.
The accuracy in the determination of transition temperatures and enthalpies by DSC is dependent
on several factors. The instrument should be temperature- and energy-calibrated using extremely pure
samples, often metals, for the calibration. The calibration procedure is shown in Fig. 1.9. The
transition temperature (Ttr) is obtained by finding the intersection of the baseline and the leading
edge of the melting trace. Polymers have a low thermal conductivity, and the samples should be flat
and in good thermal contact with the sample pan. A standardized sample geometry and mass should
be used if samples in a series are to be compared. A heat-conductive, thermally inert liquid medium
may be used to improve the thermal contact between sample and sample pan. Oriented polymer
samples (e.g. fibres) shrink extensively during melting, and this may give rise to a noisy baseline and
even artificial endo- and exothermal peaks (Wiberg and Gedde 1998). A heat-conductive liquid in
contact with the sample and sample pan reduces the noise and eliminates artificial peaks. The sample
can also be clamped to prevent shrinkage. The melting point of an oriented specimen kept at constant
length is higher than that of a corresponding specimen that is allowed to shrink freely (Wiberg and
Gedde 1998). An inert purge gas should always be used in studies of physical phase transitions.
The correct determination of the peak temperature of a first-order transition requires a correction
for the thermal lag (temperature difference) between sample and thermometer, which is proportional
to the thermal resistance (RT) and the heat flow rate. If it is assumed that the dominant thermal
resistance is between the pan holder (or plate) and the sample pan, the leading edge of the melting of a
highly pure substance such as indium can be used to correct for the thermal lag (Fig. 1.9). Parallel
10 1 Thermal Analysis of Polymers

Fig. 1.9 DSC

thermograms showing the
melting of indium and
illustrating the calibration
procedure (upper curve)
and two first order
transitions (e.g. melting
and crystallization) in a
polymer (lower curve). The
correction for the thermal
lag of the transition peak
temperatures of a polymer
sample is shown

processes should be inhibited in order to obtain correct information about a phase transition. The
melting of polymer crystals is accompanied by crystal perfection and crystal thickening. Further
details about this particular case are presented in Sect. 1.3.1.
It is difficult to resolve two overlapping phase transitions. Slow heating of small samples is the best
way of resolving overlapping peaks. Modulated temperature DSC also makes it possible to resolve
overlapping phase transitions.
The accuracy in the determination of transition enthalpies depends on the level of the enthalpy
change; on the temperature region of the transition; on the temperature dependence of the heat
capacities before, during and after the transition; and on the linearity and control of the empty pan
measurement line. Rapid heating increases the sensitivity.
Heat capacity (Cp) can be determined with an accuracy of 2–5% by DSC. The heat capacity can be
determined by heating a sample of known mass (m) at a known heating rate (β) and measuring the
heat flow difference (HFD) between the sample and an empty sample pan (Fig. 1.10):
HFD
Cp ¼ ð1:11Þ
mβ

Heat capacity data with greater accuracy (2%) is obtained by the sapphire method (Fig. 1.10), in
which the thermograms are taken at the same heating rate for the sample and a sapphire sample, both
of known mass (m and ms). Corrections are made for empty sample pans, and the heat flow differences
(HFD and HFDs) are recorded. The heat capacity of the sample is calculated according to:
HFD  ms
Cp ¼ C ð1:12Þ
HFDs  m p,s

where Cp,s is the heat capacity of sapphire, which has been determined with great accuracy.
In modulated temperature DSC or alternating DSC, the temperature is periodically modulated
(f(t)) with respect to a steady temperature change (β) according to:
T ¼ T 0 þ βt þ f ðtÞ ð1:13Þ

where T is the actual temperature at time t and T0 is the starting temperature of the scan. The periodic
modulation can be either sinusoidal or triangular. Let us consider the first case:
f ðtÞ ¼ AT sin ωt ð1:14Þ

where AT is half the amplitude of the change in temperature in a full cycle. The heating rate (q) is
obtained by insertion of Eq. (1.14) in Eq. (1.13) and differentiating with respect to time:
1.2 Thermo-analytical Methods 11

Fig. 1.10 Assessment of heat capacity (Cp) by DSC: top, direct heating of sample and subtraction with the empty pan
curve; bottom, the sapphire method

q ¼ β þ AT ω cos ωt ð1:15Þ

The maximum heating rate is β + ATω, and the minimum heating rate is β  ATω. The important
parameters are the steady heating rate (β), AT and the period of the modulation (ω). The instrument
gives the alternating heat flow (HF) and the alternating heating rate: average values (hHFi and β),
amplitudes (AHF and Aq) and the phase angle (ϕ) between the heating rate and the heat flow. From
these, the different heat capacity values can be obtained:
hHFi
Cp,av ¼ average heat capacity
β
AHF
C∗
p ¼ complex heat capacity ð1:16Þ
Aq
C0p ¼ C∗
p cos ϕ reversing heat capacity
C00p ¼ C∗
p sin ϕ non‐reversing heat capacity

Figure 1.11 shows a thermogram recorded for quenched, almost fully amorphous poly(ethylene
terephthalate) (PET). The glass transition appears as an endothermic step in both the total and the
reversing heat flow thermograms. The non-reversing heat flow shows little or no change associated
with the glass transition. PET is a polymer that has a potential to crystallize on heating to
temperatures above the glass transition, where the segmental mobility is such that crystallization
occurs. This is referred to as cold crystallization. The exothermal peak in the total heat flow curve is
due to a non-reversing process (Fig. 1.11). The final melting includes both reversing and non-
12 1 Thermal Analysis of Polymers

Fig. 1.11 Modulated DSC

thermogram of rapidly
cooled PET drawn after
data obtained with TA
Instruments (1996)

reversing components to the total heat flow. The integral of the non-reversing process with respect to
temperature is an assessment of the entropy production (cf. Sect. 1.3.1). Separating the reversing and
non-reversing heat flows makes it possible to resolve overlapping thermal processes. In conventional
DSC, processes that overlap can be resolved by applying the modulated temperature DSC technique
provided that the processes have different rates. The reversing heat flow measures instantaneous
values, and they do not include relaxation effects. The non-reversing heat flow represents the
irreversible effects. Modulated DSC can also be used to assess thermal conductivity for materials
in the conductivity range from 0.1 to 1.5 W (K m)1 (Marcus and Blaine 1994).
A more recent important development is fast scanning calorimetry, a discipline that has developed
strongly since 2000. Recommended reading is Schick and Mathot (2016). The long-chain nature of
polymers is the underlying reason for the complexity of the phase structures of polymers, e.g. the
crystalline and amorphous phases in semicrystalline polymers, and the phases in polymer blends. In
none of these cases is an equilibrium reached in the true thermodynamic sense, since a crystal of finite
thickness has a higher Gibbs free energy than a crystal of infinite thickness. An immiscible polymer
blend consists of two distinct phases and an interface, and the latter typically does not have an
equilibrium configuration. The melting and crystallization of polymers are both complex processes
because the different paths that these processes can take depend strongly on the initial conditions and
on the rate of temperature change involved. A typical polymer crystal is far from its lowest free
energy state, which for most homopolymers is an infinitely thick crystal (cf. Gedde and Hedenqvist
2019b, c). When a typical polymer crystal is heated, the structure gradually changes with time before
the crystal finally melts. This means that the recorded melting point, which reflects the structure of the
polymer crystal just prior to melting, depends on both the initial structure and the heating rate
(Fig. 1.12). A slower heating triggers a greater change in structure, i.e. more extensive crystal
thickening and hence a higher recorded melting point than after rapid heating. Figure 1.12 also
shows that polymers are capable of crystallizing over a very wide temperature range, i.e. from the
glass transition temperature (Tg) to just below the equilibrium melting point (Tm0) (cf. Gedde and
Hedenqvist 2019c). Some industrial moulding processes involve very high cooling rates, and it is
important to be able to determine the temperature at which the polymer is crystallizing under such
extreme conditions. Conventional DSC instruments lack this ability, their maximum cooling rate
being typically 300 K min1 (Schawe and Pagatscher 2016). These instruments have a signal time
constant of approximately 1 s. High-performance conventional DSC instruments such as the
PerkinElmer DSC 8500, which became available as the new millennium started (Pijpers et al.
2002), use small sample masses (<1 mg), and the maximum cooling rate is impressively high,
700 K min1.
1.2 Thermo-analytical Methods 13

Fig. 1.12 Illustration of the effect of heating rate on the recorded melting temperature; upper row and left-hand sketch
in the lower row. The lower right-hand sketch presents a comparison between the available crystallization temperature
range using classical DSC (C-DSC; red cuboid) and fast scanning DSC (green cuboid)

The first demonstrations of fast scanning calorimeters date back more than 50 years through the
work of Hager (1964), which allowed heating rates up to 500 K s1. Further developments by Allen
and co-workers (Allen et al. 1994; Efremov et al. 2004a, b) provided a foundation for a substantial
expansion of heating rates up to 100 MK s1 (Toda et al. 2016). Fast scanning calorimeters analyze
samples with ultra-low mass and use very tiny calorimeters with a very low heat capacity (Cp,cal)
(Fig. 1.13). The calorimeter is cooled by the surrounding gas, and the maximum cooling rate is
controlled by the temperature difference between the sample/calorimeter and the gas (ΔT ) and the
thermal resistance between the gas and the sample/calorimeter (Rcal-gas) according to:
ΔT
Ploss to gas ¼ ð1:17Þ
Rcal‐gas

where the energy transfer to the gas per unit time is equal to the electrical power transferred to the
calorimeter:
Pelectr ¼ Ploss to gas ð1:18Þ

The maximum scanning rate (β) is given by:

ΔT
Pelectr ¼ ¼ Cp,cal  β ð1:19Þ
Rcal‐gas
14 1 Thermal Analysis of Polymers

Fig. 1.13 Sketch of a fast

scanning calorimeter.
Drawn after Shawe and
Pogatscher (2016)

If Cp,cal is low, the surrounding gas is cold and Rcal-gas is low, the scanning rate (β) can be adjusted
to very high values. The calorimeter is a chip-based sensor with built-in heater and a thermometer that
must be in excellent thermal contact with the calorimeter. Everything is very low mass and compact.
The sample must fulfil certain requirements: (a) good thermal contact with both the gas and the
calorimeter, (b) thin sample geometry to minimize the internal temperature gradients and (c) the need
to minimize the sample mass means that much surface and little bulk are assessed, and in order to
properly assess the bulk properties, larger samples are preferred, which is thus a compromise. This
technology gives instruments with a signal time constant less than 1 ms. The first commercial
instrument of this type was the Flash DSC, Mettler Toledo which permits heating from 0.1 K s1
to 40,000 K s1. The cooling rate range is narrower: 0.1 K s1 to 10,000 K s1. Some non-commercial
instruments use sensors with a lower heat capacity allowing faster maximum cooling rates. The Flash
DSC is a power-compensated DSC (cf. Fig. 1.7).

1.2.2 Thermogravimetry

The fundamental components of thermogravimetry (TG) have existed for thousands of years.
Mastabas or tombs in ancient Egypt (2500 BC) have wall carvings and paintings displaying both
the balance and the fire. The two were however first linked together in the fourteenth century in
studies of gold refining. The first thermogravimetric experiments were conducted by Hannay (1877),
Ramsay (1877) and Ångstr€om (1895). Honda (1915) pioneered the modern TG analytical technique
in 1915, and Cahn and Schultz (1963) designed the first automatic TG instrument. At about the same
time, a high-precision instrument was introduced by Mettler (Wiedemann 1964).
TG is carried out in a thermobalance, which is an instrument with which it is possible to make
continuous measurements of sample mass as a function of temperature or time with a regulated
constant temperature. A typical TG instrument includes a recording balance, a furnace, a furnace
temperature controller and a computer (Fig. 1.14). Modern thermobalance instruments are equipped
with a DTA, which allows simultaneous calorimetric recording and simplifies the temperature
calibration of the instrument. Figure 1.15 shows a schematic TG trace for a filled polymer. The
furnaces can be used at temperatures up to 2400  C or more in a great variety of atmospheres,
including corrosive gases. The sensitivities of the modern recording balances are in the μg range,
although some instruments (e.g. those based on quartz crystals which are coated with the sample)
permit reading down to pg. A comprehensive review of the early developments of instruments has
been presented by Wunderlich (1990). One important aspect of the assessment of mass is the presence
1.2 Thermo-analytical Methods 15

Fig. 1.14 Schematic

presentation of a TG
apparatus

Fig. 1.15 Sample mass as

a function of temperature.
Schematic curve for a
polymer filled with CaCO3

of a buoyancy force ( fb) due to the displacement of a gas volume equal to the sample volume (Vs),
which reduces the measured sample mass (mb) according to:
f b ¼ V s  ρ  g ) mb g ¼ V s  ρ  g ) mb ¼ V s  ρ ð1:20Þ

where ρ is the density of the gas, calculated for an ideal gas according to the familiar formula:
pV ¼ nRT ð1:21Þ

where p is the gas pressure, V is the gas volume (i.e. ¼ Vs), n is the number of moles of gas and T is the
absolute temperature. The density of the gas is calculated using the relationship n ¼ m/M, where m is
the mass of gas and M is the molar mass:
m pM m pM
pV ¼  RT ) ¼ )ρ¼ ð1:22Þ
M RT V RT

The buoyancy mass is thus given by:

pM
mb ¼ V s  ρ ¼ V s  ð1:23Þ
RT

Two parameters are gradually changed during a TG run: Vs and T; since temperature is increasing
while Vs is decreasing, both these effects lead to a gradual decrease in mb with increasing temperature.
A correction for m ¼ f(T) can be made by applying Eq. (1.23).
It should be noted that the sample size and form affect the shape of the TG curve. A large sample
may develop temperature gradients within the sample, may suffer a temperature deviation from the
16 1 Thermal Analysis of Polymers

set temperature due to endo- or exothermic reactions and may experience a delay in mass loss due to
diffusion obstacles. Finely ground samples are preferred in quantitative analysis for these reasons.
TG is often combined with other techniques to analyze the evolved gas such as infrared
(IR) spectroscopy and gas chromatography (GC), the latter equipment is often being further com-
bined with either mass spectrometry (MS) or IR spectroscopy, to enable the volatile products to be
identified. More recent reviews on modern TG instrument developments can be found in Brown and
Gallagher (2007), Gabbott (2008) and Loganathan et al. (2017).

1.2.3 Dilatometry and Thermomechanical Analysis

Dilatometers (from Latin: dilatation (extension) and metrum (measurement)) are in principal of two
types: volume dilatometers and linear dilatometers. The classical way of measuring sample volume
as a function of temperature is with a mercury dilatometer (Fig. 1.16a), where the sample is immersed
in a liquid mercury and the level of the liquid in a capillary is measured by direct reading, an optical
method or using an electrical device. Mercury can be used at temperatures between its melting point
and boiling point, i.e. from 19  C to 356  C. Other immersion liquids than mercury have been used,
e.g. water, alcohols and silicone fluids. Bekkedahl (1974) reported an accuracy of 0.3 m3 (kg)1 for a
mercury dilatometer. The cylinder-piston dilatometer shown in Fig. 1.16b is suitable for viscous
liquids such as polymer melts. A constant force is applied to the piston, and the motion of the piston
records the volume change. Pressure dilatometry can be carried out with this instrument. Problems
with leakage at the piston end and too high frictional forces at the seal have been solved in modern,
well-designed instruments. The bellows-type dilatometer shown in Fig. 1.16c is suitable for both
solids and liquids, and it is the recommended method for pressure dilatometry. The accuracy of this
instrument is of the order of 1 m3 (kg)1. The pressure dilatometer designed by Zoller et al. (1976) is
commercially available.

Fig. 1.16 Volume dilatometers: (a) mercury dilatometer; (b) cylinder-piston dilatometer; (c) bellows-type dilatometer.
Drawn after Zoller (1990)
1.2 Thermo-analytical Methods 17

Fig. 1.17 Schematic

presentation of a TMA
(top). Bottom left,
expansion probe for
measurement of the linear
thermal expansion
coefficient; bottom right,
penetration probe for the
measurement of stiffness

Linear dilatometers are commercially marketed by several companies. These instruments are
referred to as thermomechanical analyzers (TMA), and they measure not only the linear thermal
expansion coefficient but also the modulus as a function of temperature.
When thermal expansion or penetration (modulus) is being measured, the sample is placed on a
platform of a quartz sample tube (Fig. 1.17), since the thermal expansion coefficient of quartz is small
(about 0.6.106 K1) compared to that of the polymeric material. The quartz tube is connected to the
armature of a linear variable differential transformer (LVDT), and any change in the position of the
core of the LVDT, which floats friction-free within the transformer coil, results in a linear change in
the output voltage. The upper temperature limit for the currently available commercial instruments is
about 725  C. Gillen (1978) showed that the tensile compliance given by TMA using the penetration
probe and suitable compressive loads is comparable with the data measured by conventional
techniques on considerably larger samples.
Volume changes occur naturally in most polymers samples when they are subjected to mechanical
loads and to a very great extent under larger strains due to crazing and crack formation cf. Chap 6.
The volume strain (ΔV/V0) in a uniaxial tensile testing experiments (stress along direction 1) can be
determined by simultaneous measurement of longitudinal (ε1) and transverse strains (ε2 and ε3) using
strain gauge extensometers according to:
ΔV ð1 þ ε1 Þ  ð1 þ ε2 Þ  ð1 þ ε3 Þ  1  1  1
¼  ε1 þ ε2 þ ε3 ð1:24Þ
V0 111

Nothe that this approximation is only valid at small strains. The accuracy of the measurements is
high for glassy polymers due to their high stiffness. Problems due to indentation by the transverse
extensometer during stress-strain experiments on softer polymers such as low-density polyethylene
18 1 Thermal Analysis of Polymers

were reported by Delin et al. (1995). Other types of strain sensors, such as inductive extensometers
according to a design of Bertilsson et al. (1994), can be used in such cases. In these cases, dilatometers
can be used in which the stressed specimen is surrounded by a liquid in a container with an attached
capillary. Such equipment has been designed by Smith (1959), Coumans and Heikens (1980) and
Delin et al. (1995), who modified design of Coumans and Heikens to give better mechanical and
thermal stability. The thermal fluctuation must be no greater than 0.01  C in order to obtain quality
data on the volume strain.
James (2017) provides a modern review of dilatometry, and this is recommended reading.

1.2.4 Dynamic Mechanical Thermal Analysis

Dynamic mechanical thermal analysis (DMTA), also referred to as dynamic mechanical analysis
(DMA), measures stress and strain in a periodically deformed sample at different frequencies and
temperatures. It provides comprehensive information about the stress-strain properties at different
temperatures and about the relaxation processes, specifically the glass transition and the sub-glass
processes. When watching a bouncing ball, you will notice that the ball is losing energy in each
bounce. This is a simple illustration that materials are never completely elastic and hence that part of
the mechanical energy is converted to heat. DMTA provides a means to measure the elastic (storage)
and viscous (loss) components of the modulus at different frequencies (cf. Chap. 6).
Figure 1.18 shows the stress response to a constrained strain function, a sinusoidal variation. An
elastic material that obeys Hooke’s law (stress / strain) shows a stress response which is in phase

Fig. 1.18 Stress response

to a sinusoidal strain for (a)
an elastic material and (b) a
viscous material
1.2 Thermo-analytical Methods 19

Fig. 1.19 (a) Schematic presentation of a torsion pendulum (left) and two apparatuses with forced oscillation:
controlled strain (centre) and controlled stress (right). The samples in the different sketched experimental setups are
marked with capital S. (b) Different loading configurations

with the strain function (Fig. 1.18a). A viscous material, on the other hand, follows Newton’s law
(stress / strain rate), and the stress is 90 out of phase from the sinusoidal strain (Fig. 1.18b).
Anelastic materials, i.e. most polymers, belong to this group while being moderately strained, and
they show a mixed response with a stress function at an angle between 0 and 90 from the strain
function; the modulus has two different components, an elastic (storage) component which is in phase
with the strain and a viscous (loss) component which is 90 out of phase with the strain. It should be
noted that the strains involved in the measurements need to be small (less than 0.5%) to avoid a
non-linear response. There are essentially two main types of instruments: the free oscillation
instruments such as the torsion pendulum and the forced oscillation instruments (Fig. 1.19).
The torsion pendulum sketched in Fig 1.19a (left) has preferentially a cylindrical sample, which is
rigidly held at one end, and supports an inertia rod at the other end. The inertia rod is set into
oscillation, and the polymer sample is subjected to a sinusoidal torsion, which, depending on the
relative size of the viscous component, gradually dampens down. The frequency range of operation is
between 0.01 and 50 Hz. The real (G’) and imaginary (G’’) parts of the shear modulus are given by:

2lMω2
G0 ¼ ð1:25Þ
πr 4

and

2l2 M 2 Λ
G00 ¼ ω  ð1:26Þ
πr 4 π
20 1 Thermal Analysis of Polymers

where l is the length of the sample rod, M is the moment of inertia of the inertia rod, ω is the angular
frequency, r is the radius of the sample rod and Λ is the logarithmic decrement, defined as:
 
1 An
Λ ¼  ln ð1:27Þ
k Anþk

where k is the number of swings and Ai is the amplitude of the ith swing. It is possible to vary the
angular frequency by changing the moment of inertia (M ) of the inertia rod.
The forced oscillation techniques shown in Fig 1.19a (centre and right sketches) can be used over a
wider frequency range than the torsion pendulum, ranging from 104 to 104 Hz. A motor can be used
to control amplitude and frequency of the strain, while both the strain and stress are measured as a
function of time using a strain-sensitive device (e.g. a linear variable differential transformer
(LVDT)) and a load cell. Commercial instruments are also available controlling in which the
sinusoidal stress is controlled, and the dynamic data are obtained by measurement of the strain as a
function of time. Uniaxial extension, bending, torsion and shear are used in different commercially
available instruments (Fig. 1.19b). The strain (ε) is dependent on time (t) in such measurements
according to:
ε ¼ ε0 sin ωt ð1:28Þ

where ω is the angular frequency and ε0 is the strain amplitude. The stress (σ) is given by:
σ ¼ σ 0 sin ðωt þ δÞ ð1:29Þ

where δ is the phase angle displacement of the stress function with respect to the strain function and
σ 0 is the stress amplitude. Note that δ varies between the limits 0 (elastic) and 90 (viscous). The
storage modulus (E´; the elastic component) is given by
 
σ0
E0 ¼  cos δ ð1:30Þ
ε0

and the loss modulus (E´´) by:

 
σ0
E00 ¼  sin δ ð1:31Þ
ε0

Other frequently used quantities are:

sin δ E00
tan δ ¼ ¼ ð1:32Þ
cos δ E0

and
 
E00
δ ¼ arctan ð1:33Þ
E0

The complex modulus (E*) is calculated according to:

E∗ ¼ E0 þ iE00 ð1:34Þ
1.2 Thermo-analytical Methods 21

Fig. 1.20 The logarithm of the storage shear modulus (G´) and the logarithm of the loss shear modules (G’’) as
functions of temperature for a linear aliphatic copolyester with 30% crystallinity. Isochronous data taken at 1 Hz.
Drawn after data of Boyd and Aylwin (1984)

Figure 1.20 shows a temperature scan at constant frequency (1 Hz) for an aliphatic copolyester
displaying a glass transition (denoted β) and a sub-glass process (denoted γ). Note that the G´´-peaks
for both processes coincide with the inflection points of the G´-steps.
Figure 1.21 shows dynamic mechanical data for styrene-butadiene rubber focussing on the glass
transition. The inflection point occurs at the same temperature as the maximum in G´´. It is important
to note that the maximum in tan δ occurs at a higher temperature, shifted in this case by 10  C from the
inflection point of G´. The recommendation is that the glass transition temperature be reported as the
inflection point of the G´ jump, which is the temperature associated with the maximum in G´´.
If temperature scans are run at different frequencies ( f ), the activation energy (Ea) of a sub-glass
process can be assessed using the Arrhenius law:
 
Ea E 1
f ¼ A  exp ) ln f ¼ ln A  a  ð1:35Þ
RT R T

where f and T denote the frequency and temperature of the maximum in G´´, A is a constant and R is
the gas constant. The activation energy is assessed in a plot referred to as an Arrhenius diagram
(Fig. 1.22) and is calculated from the slope according to:
Ea ¼ slope  R ð1:36Þ
22 1 Thermal Analysis of Polymers

Fig. 1.21 Complex modulus values (G´ and G´´) and the ratio of the two, tan δ ¼ G´´/G´ plotted as a function of
temperature for a styrene-butadiene rubber. Isochronous data taken at 1 Hz. Drawn after data of Wagner (2017)

Fig. 1.22 Sketch of isochronous (constant frequency) loss modulus data plotted as a function of temperature (upper
diagram). The peak values are then plotted in the Arrhenius diagram (lower diagram), which displays both a sub-glass
process (straight line) and a glass transition (curved line)

The glass transition follows another temperature law referred to as the Vogel–Fulcher–Tam-
mann–Hesse equation (Vogel 1921; Fulcher 1925a, b; Tammann and Hesse 1926) or the
Williams–Landel–Ferry equation (Williams et al. 1955):
 
B B
f ¼ A  exp  ) ln f ¼ ln A  ð1:37Þ
T  T0 T  T0
1.2 Thermo-analytical Methods 23

where A and B are constants. In the Arrhenius diagram, this equation appears as a curved line that
never goes beyond T0, i.e. at T ¼ T0, ln f ! – 1. Further information about the glass transition and
the sub-glass processes is found in Gedde and Hedenqvist (2019a) and in Chaps. 6 and 7. The
temperature shift as accompanied by a change in frequency for the glass transition is not controlled by
a single-valued activation energy. The sub-glass processes follow the Arrhenius law (Eq. (1.35)).
Time–temperature superposition is a method by which a series of isothermal dynamic mechanical
data (e.g. the logarithm of the storage modulus as a function of the logarithm of the frequency) can be
brought together in a so-called master curve at one selected temperature. The method involves both
vertical shifting (along the log storage modulus axis; minor) and horizontal shifting (along the log
frequency axis; major). The horizontal shift factor (log aT) is plotted versus 1/T, and for a sub-glass
process, it shows a straight line, i.e. follows the Arrhenius law, whereas for a glass transition process,
data shows curvature (i.e. follows the Vogel–Fulcher–Tammann equation). One extremely useful
‘property’ of master curves is that they span over a very wide frequency range, typically by ten orders
of magnitude. An in-depth presentation of the time–temperature superposition is found in Chap. 5.
Recommended further reading are Ferry (1980) and Boyd and Smith (2007a, b).
Dynamic modulus data (E* ¼ f(ω); J* ¼ f(ω)) can be converted into stress relaxation modulus data
(ER ¼ f (t)) and creep compliance data (J (t)) by applying the Boltzmann superposition principle
(Boltzmann 1847). The scheme of interconversions between these constitutive parameter functions
which also includes the relaxation time spectrum (H (τ)) is presented in detail in Chap. 6. An excellent
text on this topic is Boyd and Smith (2007a). The text by Menard and Menard (2017) is also
recommended.

1.2.5 Thermal Optical Analysis and In Situ Structural Assessment Under

Controlled Thermal History

Thermal optical analysis (TOA) is usually performed in a polarized light microscope equipped with a
hot-stage, which control the temperature of the sample. More information about polarized light
microscopy is given in Chap. 2.
The microscopic image of the typically 10 μm thick sample can be viewed directly in the
binoculars, or the intensity of the transmitted light can be recorded with a photodiode (Fig. 1.23).
A CCD camera connected to a PC enables effective recording of the structure and its development
with time/temperature (Fig. 1.24). The microscopy uses specially designed objective lenses with long
working distances to enable thermal insulation of the heat-sensitive objective from the hot specimen.
Combined TOA/DTA instruments are commercially available. Apparatuses for TOA of sheared
polymer melts were developed during the 1980s.
Atomic force microscopes are commonly equipped with hot-stages, and by preparation of special
samples, even a more detailed understanding of crystallization can be gained. Figure 1.25 shows
isothermal crystallization of a thin film of a star-branched poly(ε-caprolactone) and the assessment of
the growth rates along two crystallographic directions, i.e. along the a- and b-axes. Recommended
further reading about atomic force microscopy as an in situ tool to study crystallization and melting of
semicrystalline polymers are by Prud´homme (2016), Hobbs (Mullin and Hobbs 2011; Catto et al.
2011) and Vancso and Sch€onherr (2010).
X-ray diffraction (WAXS and SAXS) has developed strongly during the past decades. The use of
‘fast’ radiation methods, the synchrotron radiation, enables study of fast processes while controlling
temperature. Chapter 3 presents then fundamentals about the X-ray methods. Details about melting
and other reorganization processes arrives from data for crystallinity, long period (Lp i.e. crystal
thickness (Lc) + amorphous layer thickness (La)) and crystal phase including mesomorphic states. The
24 1 Thermal Analysis of Polymers

Fig. 1.23 Melting of a

crystalline polymer as
recorded by TOA

Fig. 1.24 Polarized photomicrographs showing growing superstructures in polymers based on poly(ε-caprolactone)
(PCL): (a) perfect spherulites in a star-branched (third generation dendrimer core) polymer with PCL arms with
51 repeating units crystallizing at 311 K. (b) Irregular spherulite in a star-branched (third generation dendrimer core)
polymer with PCL arms with 24 repeating units crystallizing at 315 K. (c) Axialite in a linear PCL with 39 repeating
units crystallizing at 325 K. From Nunez et al. (2004) with the permission of Marcel Dekker. The bottom graph shows
the spherulite radius (R) as a function of the crystallization time for a star-branched (third generation dendrimer core)
polymer with PCL arms with 51 repeating units crystallizing at 313.4 K. Drawn after data of Nunez et al. (2004). The
linear growth rate (Gr ¼ dR/dt) values for the individual spherulites show only a moderate variation
1.2 Thermo-analytical Methods 25

Fig. 1.25 Phase image of a thin film of a star-branched (third generation dendrimer core) polymer with PCL arms with
51 repeating units crystallizing at 326. Pictures (a) and (b) were taken at different crystallization times, 34 min and
51 min. Note the growing crystal (with flat orientation; denoted F2) from which the growth along the crystallographic a-
and b-axes can be determined from the slopes of the lines in graph c. From Nunez et al. (2008) with permission of
Taylor and Francis, London

following texts are recommended: (i) Kane et al. (1994), Mathot et al. (1996), Portale et al. (2013) and
Tence-Girault et al. (2019); these texts demonstrate the power of the technique to reveal mechanisms
and how this slots in with using other thermo-analytical techniques such as DSC, (ii) Graewert and
Svergun (2013) and Bolze et al. (2018).

1.2.6 Dielectric Thermal Analysis (DETA)

DETA, which is also referred to as dielectric spectroscopy, provides information about the segmental
mobility of a polymer. Chemical bonds between unlike atoms possess permanent electrical dipole
moments (cf. Gedde and Hedenqvist (2019d)). Many polymers with strong bond dipole moments
show no molecular dipole moments due to symmetry, i.e. the bond moments of a given central atom
(or group of atoms) counteract each other, and the vector sum of the dipole moments is zero. Other
polymers however have repeating unit structures such that the dipole moments can vectorially
accumulate into a repeating unit dipole moment in the possible conformational states. This group
of polymers can be studied by DETA. DMTA and DETA are thus two methods that detect segmental
mobility of polymers. DMTA can be used to study nonpolar and polar polymers, whereas DETA is
only able to detect segmental mobility in polar polymers. Polyethylene can, however, be investigated
by DETA by first oxidizing the polymer. Carbonyl groups are introduced, and the chains are thus
labelled with strong dipoles. DETA covers a much wider frequency range than DMTA. This is a great
advantage for DETA, because the relaxation processes in polymers are very broad and the analysis
becomes more precise when data covering a wider frequency range can be taken.
The polarizability (α) is the sum of dipole moments per unit volume. Figure 1.26 shows the
variation of α as a function of the frequency of the alternating electric field. The electric field induces
a distortion of the electronic clouds at a frequency of about 1015 Hz, which corresponds to the optical
ultraviolet range. The electronic polarizability (αe) is closely related to the refractive index (n) of
visible light according to the Lorentz–Lorenz equation (Lorenz 1870; Lorentz 1881):

n2  1 4πρN A
¼  αe ð1:38Þ
n2 þ 2 3M

where ρ is the density, NA is the Avogadro number and M is the molar mass.
26 1 Thermal Analysis of Polymers

Fig. 1.26 Schematic

representation of the
polarizability as a function
of the logarithm of the
frequency

Atomic polarization arises from small displacements of atoms under the influence of the electric
field at a frequency of approximately 1013 Hz (infrared range). The atomic polarizability cannot be
determined directly but is normally small compared to the electronic polarizability. Electronic and
atomic polarizations occur in all types of polymers, even polymers with no permanent dipole
moments. Polymers with permanent dipoles show no macroscopic polarization in the absence of an
external electric field. If an alternating electric field is applied and if the electric field frequency is
sufficiently low with reference to the jump frequency of segments of the polymer, the dipoles orient in
the field and the sample shows not only electronic and atomic polarization but also a dipolar
polarization. DETA, which operates in a frequency range from 104 to 108 Hz, is used to monitor
dipole reorientation induced by conformational changes. These are referred to as dielectric relaxation
processes.
Let us consider two parallel plates with opposite surface charges (σ) and a dielectric material
inserted between the plates. The electric field (E) causes a displacement field (D), which for a
configuration of charged parallel plates is related to the polarization (P) by the expression:
D ¼ κ0 E þ P ð1:39Þ

where κ0 is the permittivity of free space. The polarization (P) is the induced dipole moment per unit
volume and can expressed as:
P ¼ χκ0 E ð1:40Þ

where χ is called the susceptibility. Insertion of Eq. (1.40) in Eq. (1.39) yields:
D ¼ ð1 þ χ Þκ 0 E ¼ εκ 0 E ð1:41Þ

where ε ¼ 1 + χ is the dielectric constant. The capacitance (C) of a capacitor is defined as the amount
of charges that can be stored per unit voltage, and it can be shown that the dielectric constant (ε) of the
medium inserted between the plates is given by:
C
ε¼ ð1:42Þ
C0

where C0 is the capacitance of the capacitor in vacuum. Most dielectric work is concerned with a
periodic E field with an angular frequency ω, and the dielectric constant is conveniently represented
as a complex quantity:
1.2 Thermo-analytical Methods 27

ε∗ ¼ ε0  iε00 ð1:43Þ

where ε0 is referred to as the dielectric constant and ε’’ as the dielectric loss. It can be shown that the
complex displacement is given by:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
D∗ ¼ ε0 2 þ ε00 2 κ 0 E0 eiðωtδÞ ð1:44Þ

where δ is the phase difference between the applied E field and the displacement field. The phase
angle δ is given by:
ε00
tan δ ¼ ð1:45Þ
ε0

A simple assumption may be made that the polarization that originates from a reorientation of
dipoles is proportional to its displacement from equilibrium (χ o):

dχ
¼ kðχ  χ 0 Þ ð1:46Þ
du

where u is past time and k is a rate constant. The solution to Eq. (1.46) is:
 
χ ðt  uÞ ¼ χ 1 þ χ 0 1  eðtuÞ=τ ð1:47Þ

where χ 1 is the instantaneous susceptibility (due to electronic and atomic polarization) and τ is the
relaxation time. The angular frequency (ω) dependence of the complex dielectric constant can be
calculated using Eq. (1.47) and assuming that the Boltzmann superposition principle is valid:
ð εr  εu Þ
ε∗ ¼ εu þ ð1:48Þ
ð1 þ iωτÞ

where εu is the unrelaxed dielectric constant (the dielectric constant obtained at such a high frequency
that no dipole relaxation takes place) and εr is the relaxed dielectric constant, which is the value that
the dielectric constant takes at such a low frequency that all dipoles move with the E field. The
difference between the two limiting dielectric constants (εr – εu) is called the relaxation strength. The
real and imaginary parts of the complex dielectric constant are obtained from Eq. (1.47) as follows:
ðεr  εu Þð1  iωτÞ ðε  εu Þ  iωτðεr  εu Þ
ε0  iε00 ¼ εu þ ¼ εu þ r ð1:49Þ
ð1 þ iωτÞð1  iωτÞ ð 1 þ ω2 τ 2 Þ

ð εr  εu Þ
ε0 ¼ εu þ ð1:50Þ
ð1 þ ω2 τ2 Þ

and
ðεr  εu Þωτ
ε00 ¼ ð1:51Þ
ð1 þ ω 2 τ 2 Þ

The dielectric constant and dielectric loss obtained according to Eqs. (1.50) and (1.51) are plotted
as a function of log ωτ in Fig. 1.27. A maximum in ε00 and an inflection point in ε0 appear at the
angular frequency ω ¼ τ1. It should be noted that the dielectric constant takes the relaxed value
28 1 Thermal Analysis of Polymers

Fig. 1.27 Dielectric

constant (ε0 ) and dielectric
loss (ε00 ) as a function of
angular frequency (ω)
assuming a single
relaxation time (τ), β ¼ 1,
and symmetric broadening
according to the Cole–Cole
equation (Eq. (1.52)) with a
symmetric broadening
factor β ¼ 0.2. In both
cases, the following
limiting values for the
dielectric constant have
been used: εu ¼ 3 and
εr ¼ 5

(εr ¼ 5) at low frequencies (ω < 0.01τ1) and the unrelaxed value (εu ¼ 3) at frequencies greater than
100 τ1.
It is very uncommon that a single relaxation time model can describe relaxation processes in
polymers, and Eq. (1.48) has been modified empirically. The Cole–Cole equation (Cole and Cole
1941) includes a symmetric broadening factor β, which takes values between 0 and 1:
ð εr  εu Þ
ε∗ ¼ εu þ   ð1:52Þ
1 þ ðiωτ0 Þβ

where τ0 is the central relaxation time.

The real and imaginary parts of the complex dielectric constant are in this case given by:
  
ðεr  εu Þ 1 þ ðωτ0 Þβ cos βπ
2
ε0 ¼ εu þ  βπ  ð1:53Þ
β
1 þ 2ðωτ0 Þ cos 2 þ ðωτ0 Þ2β

and
 
ðεr  εu Þðωτ0 Þβ sin βπ
ε00 ¼  βπ
2  ð1:54Þ
β
1 þ 2ðωτ0 Þ cos 2 þ ðωτ0 Þ2β

The effect of a lowering of β from 1 (single relaxation time) to 0.2 is shown in Figs. 1.27 and 1.28.
An appreciable change in the dielectric constant occurs over a frequency range, which is greater than
12 orders of magnitude. The dielectric loss curve shows only a weak maximum at ωτ0 ¼ 1.
The Argand diagram (Fig. 1.28) is a useful diagram which reveals several important parameters of
the relaxation process. This type of diagram is also called a Cole–Cole plot or a complex plane plot.
The single relaxation time (β ¼ 1) data form a half circle which intercepts with the x-axis at εu and εr.
The symmetrically broadened relaxation (β ¼ 0.2) also results in a circular arc which intercepts the
x-axis at εu and εr, but the radius of the circle is considerable larger than that representing a single
relaxation time. The equation for the circle is given by:
1.2 Thermo-analytical Methods 29

Fig. 1.28 Argand diagram

showing the circular arcs of
the data presented in
Fig. 1.27

      2   2
0 εr  εu 2 00 εr  εu βπ ð εr  εu Þ βπ
ε  þ ε þ cot ¼ csc ð1:55Þ
2 2 2 2 2
0 00
which represents a circle with its centre at the (ε , ε ) coordinates (1/2 · (εr  εu), 1/2 ·
(εr  εu) · cot (βπ/2)) and a radius of 1/2 · (εr  εu) csc (βπ/2).
A number of other empirical equations have been proposed. The Davidson–Cole equation
(Davidson and Cole 1950) includes an asymmetric broadening factor (γ):
ð εr  εu Þ
ε∗ ¼ εu þ ð1:56Þ
ð1 þ iωτ0 Þγ

The combination of the Cole–Cole equation (Eq. (1.52) and the Davidson–Cole (Eq. (1.56)) is
referred to as the Havriliak–Negami equation after the inventors (Havriliak and Negami 1967):
ð εr  εu Þ
ε∗ ¼ εu þ  γ ð1:57Þ
1 þ ðiωτ0 Þβ

This equation is very flexible, and it can be fitted to most dielectric data. Figure 1.29 shows
dielectric data presented in an Argand diagram, where two circular arcs indicating symmetric
broadening (γ ¼ 1) of two relaxation processes α and β are fitted to the experimental data. With the
positions of the centres of the circular arcs, it is evident that the symmetric broadening factor is larger
for the α process than for the β process.
The broadening of the dielectric relaxation is expressed by, e.g. the Cole–Cole equation shown
in the time domain (dielectric permittivity ε ¼ f (t)) by the famous empirical Kohlrausch–
Williams–Watts (KWW) equation (Kohlrausch 1854; Williams and Watts 1970):
    
t β
εðtÞ ¼ εu þ ðεr  εu Þ  1  exp  ð1:58Þ
τ

where τ is the central relaxation time and β (<1) is the empirical broadening factor. The work of
Kohlrausch preceded that of Williams and Watts by more than 100 years, and it was also used for
other purposes.
The temperature dependence of a relaxation process is conveniently represented in an Arrhenius
diagram (Fig. 1.30). Sub-glass processes display an Arrhenius temperature dependence:
1 ΔE
f max / ¼ A1  e RT ð1:59Þ
τ0
30 1 Thermal Analysis of Polymers

Fig. 1.29 Argand plot

showing two dielectric
relaxation processes (α and
β) for a hyperbranched
polyester at 260 K. The
lines are obtained by fitting
the experimental data
indicated by the points.
Drawn after data of
Malmstr€ om et al. (1994)

Fig. 1.30 Arrhenius

diagram showing WLF and
Arrhenius temperature
dependence. The lines are
obtained from isothermal
dielectric loss data as a
function of log frequency

where ΔE is the activation energy and A1 is a constant. The glass transition adapts to the
Williams–Landel–Ferry (WLF) or Vogel–Fulcher–Tammann–Hesse (VFTH) temperature depen-
dence (Vogel 1921, Fulcher 1925a, b, Tammann and Hesse 1926, Williams et al. 1955):
A3
1
f max / ¼ A2  e TT0 ð1:60Þ
τ0

where A2 and A3 are constants and T0 is a temperature.

The instruments used for the determination of dielectric constants are based on Wheatstone
bridges (10–107 Hz), Schering bridges (104–10 Hz), distributed circuits and time-domain
spectrometers. Boyd (1980) and Boyd and Smith (2007b) provide detailed information about dielec-
tric spectroscopy including the apparatus used. The classical method for determining the dielectric
constant and dielectric loss by measuring the capacitance in a parallel plate capacitor (area A and
separation d ) using a Wheatstone bridge or a Schering bridge device expresses the measured
impedance in terms of a resistance (R) and a capacitance (C) in a parallel arrangement. The dielectric
constant (ε´) is given by:
1.3 Thermal Behaviour of Polymers 31

Cd C
ε0 ¼ ¼ ð1:61Þ
κ 0 A C0

where κ 0 is the permittivity of free space and C0 is the empty cell capacitance, given by:
Aκ 0
C0 ¼ ð1:62Þ
d

The dielectric loss (ε´´) is given by:

1
ε00 ¼ ð1:63Þ
ωRC0

The time-domain spectrometer charges the capacitor, i.e. the sample between two parallel metal
plates, by applying a step voltage. The time-dependent relative permittivity (ε (t)) is obtained by
measuring the time-dependent capacitance (C(t)):
CðtÞ
εðtÞ ¼ ð1:64Þ
C0

from which ε´(ω) and ε´´(ω) can be determined. Recommended further reading about dielectric
spectroscopy are the classical book McCrum et al. (1967), Boyd (1980), Feldman et al. (1992),
Kremer and Schonhals (2003), Boyd and Smith (2007b), Pethrick (2013) and Malas (2016).

1.3 Thermal Behaviour of Polymers

1.3.1 Semicrystalline Polymers

+
The melting of polymer crystals exhibits many instructive features of non-equilibrium behaviour.
It has been known since the 1950s that the crystals of flexible-chain polymers such as polyethylene
(PE) are lamella-shaped with the chain axis almost parallel to the normal of the lamella. The thickness
of the crystal lamellae (Lc) is of the order of 10 nm, corresponding to approximately 100 main chain
atoms, which is considerably less than the total length of a typical polymer chain. This led to the
realization that the chains must be folded in the crystals. The transverse dimensions (W ) are two to
three orders of magnitude greater than the lamella thickness. The melting temperature (Tm) of thin
crystal lamellae is related to the lamella thickness (Lc) according to the Gibbs–Thomson equation:

2σ f T 0m 1
T m ¼ T 0m   ð1:65Þ
Δh0m ρc Lc

where Tm0 is the melting temperature of an infinitely thick crystal (the so-called equilibrium melting
temperature), σ f is the fold surface free energy, Δhm0 is the heat of fusion and ρc is the density of the
crystalline phase. The Gibbs–Thomson equation has been explored in more full detail by Gedde and
Hedenqvist (2019b). Factors, such as molar mass, degree of chain branching and cooling rate,
influence the crystal thickness and hence the melting temperature.
The lamella shape of the polymer crystal (Lc/W  0.01–0.001, where W is the width of the crystal)
is the reason why rearrangement occurs at temperatures well below the melting point. The
32 1 Thermal Analysis of Polymers

Fig. 1.31 Melting point of

solution-grown crystals of
linear polyethylene
(0.05 wt.% polymer in
toluene at 81  C) as a
function of heating rate.
Drawn after data of
Hellmuth and Wunderlich
(1965)

equilibrium shape of the crystals can be calculated on the basis of the surface free energies of the fold
surface (σ f) and the lateral surface (σ L) (Gedde and Hedenqvist 2019b):
Lc σ
¼ ð1:66Þ
W σL

In linear polyethylene, the surface free energies are σ ¼ 93 mJ m2 and σ L ¼ 14 mJ m2, yielding
an equilibrium Lc/W value of 6.6 which is three orders of magnitude greater than the experimental
value obtained for crystals in samples crystallized under normal pressure. The crystal thickening of
polymer crystals expected to occur on the basis of these thermodynamic arguments has been verified
by small-angle X-ray scattering. The dependence of the melting point on the rate of heating of single
crystals of linear polyethylene presented in Fig. 1.31 agrees with this view. Crystal thickening occurs
at low heating rates to a greater extent than at high heating rates, and this leads in turn to a greater final
crystal thickness and a higher melting point after slow heating. The melting point value obtained at
higher heating rates is thus more in agreement with that of the original crystals. The melting
temperature is constant at approximately 121  C with a heating rate of 20 K min1 or higher. This
indicates the ‘true’ melting temperature of the original crystal because this relatively fast heating does
not permit significant crystal thickening and perfection prior to the final crystal melting. Such a
melting is referred to as zero-entropy-production melting (Wunderlich 1964).
Extended chain crystals of polyethylene are produced by high-pressure crystallization at elevated
temperatures, typically at 500 MPa and 245  C (Wunderlich and Arakawa 1964; Geil et al. 1964).
These μm-thick crystals display a melting behaviour distinctly different from that of the thin folded-
chain single crystals grown from solution. The increase in melting point with increasing heating rate
shown in Fig. 1.32 is due to superheating. The melting of such a sample is slow because of the small
surface-to-volume ratio of these extraordinarily thick crystals. In this particular case, it is thus
possible to supply heat to the crystal faster than it can melt. It is possible to eliminate the influence
of superheating by extrapolation of the data for melting peak temperature vs. heating rate to zero
heating rate. Liberti and Wunderlich (1968) applied a facile method for the assessment of the true
melting point based on data taken at sufficiently high heating rates to generate superheating based on
the following equation:
 z
dT
Tp ¼ Tm þ A  ð1:67Þ
dt

where Tp is the recorded melting peak temperature at a heating rate of dT/dt and A and z are adjustable
parameters. The use of small z-values permits the data to be adjusted to a straight line (Fig. 1.33b).
1.3 Thermal Behaviour of Polymers 33

Fig. 1.32 Melting point of

extended chain crystals of
linear polyethylene grown
at 0.48 MPa and 227  C as
a function of heating rate.
Drawn after data of
Hellmuth and Wunderlich
(1965)

Fig. 1.33 (a) Melting peak temperature (Tp) of a sample of polyamide 6 as a function of heating rate (dT/dt). The ‘true’
melting point (at zero heating rate) is indicated in the graph. Drawn after data of Liberti and Wunderlich (1968). (b) The
data displayed in graph (a) plotted according to Eq. (1.67)) showing a rectilinear relationship at a low z-value

Polyethylene crystals of intermediate thickness display approximately zero-entropy-production

melting, i.e. the melting point is almost independent of heating rate. A positive entropy production of
the crystal subsystem is due to a rearrangement of the crystals by lamella thickening and/or a
reduction of the defect content or to superheating. Equilibrium melting, where an equilibrium crystal
is converted into an equilibrium melt, involves no entropy production. The polyethylene crystals of
intermediate thickness are clearly not in equilibrium, so in this case the metastable crystals are
converted to a metastable melt without either rearrangement or superheating. Interestingly, as shown
in Fig. 1.34, linear polyethylene with a broad molar mass distribution contains crystals with a great
variety of thicknesses, displaying in a single sample reorganization, zero-entropy-production melting
and superheating with increasing crystal thickness, i.e. with increasing melting point. Modulated
DSC provides direct information about the entropy production. Figure 1.11 shows a good example of
positive entropy production in poly(ethylene terephthalate) by first cold crystallization at
temperatures just above Tg and at higher temperatures by rearrangement of crystals (crystal
thickening, perfection of crystals and recrystallization) in the melting temperature range.
Fast scanning calorimetry provides a wide heating rate range. Figure 1.35 shows melting peak
temperature data for an isotactic polypropylene sample recorded at heating rates between 10 and
34 1 Thermal Analysis of Polymers

Fig. 1.34 Difference in

melting point (ΔT ) as
recorded at 10  C min1
from that recorded at high
heating rates (region I:
crystal thickening) or zero
heating rate (region III:
superheating)
plotted vs. melting point as
recorded at 10  C min1.
Region II refers to zero-
entropy-production
melting. Drawn after data
by Gedde and Jansson
(1983a) on linear
polyethylene

Fig. 1.35 Melting peak

temperature for an isotactic
polypropylene sample
(crystallized during cooling
at 1  C s1) as a function of
the square root of the
heating rate (dT/dt). Drawn
after data of Schawe and
Pogatscher (2016)

10,000  C s1. Tp gradually decreases with increasing heating rate in heating rate region I, which
suggests that a reorganization (crystal thickening) of the crystalline phase occurs during heating. In
region II, superheating dominates, and the reorganization of the crystalline phase is suppressed. The
‘true’ melting temperature is obtained by linear extrapolation to zero heating rate of the data at
heating rates within region II (200–10,000  C s1) to zero heating rate. This is an elegant method that
considers both reorganization and superheating effects on the melting peak temperature and makes it
possible to assess of the transition temperature associated with zero-entropy-production melting.
In general, polymers melt over a wide temperature range, typically more than 30  C. This is due,
firstly, to their multicomponent nature: polymers always exhibit a distribution in molar mass and
occasionally also in monomer sequence (copolymers). The different molecular species crystallize at
different temperatures, and this leads to a significant variation in crystal thickness and, according to
the Gibbs–Thomson equation, in the melting temperature. A polymer sample which has crystallized
under isothermal conditions and has then been rapidly cooled to a lower temperature frequently
exhibits bimodal melting for these reasons (Fig 1.36a). The high-temperature peak is associated with
the high molar mass species, which have crystallized under isothermal conditions, and the
low-temperature peak (denoted a) is due to the low molar mass component which is able to crystallize
only during the subsequent cooling phase. By extraction in hot p-xylene, most of the low melting
species (of confirmed low molar mass) were removed, and this left the smaller peak denoted b.
Figure 1.36b shows a blend of linear and branched polyethylene, the latter being segregated by the
crystallization at high temperature and thus appearing as the low-melting peak.
1.3 Thermal Behaviour of Polymers 35

Fig. 1.36 (a) Bimodal melting of linear polyethylene (Mn ¼ 8.4 kg mol1 and Mw ¼ 90 kg mol1) crystallized at 398 K
for 24 h and thereafter rapidly cooled to 300 K (thermogram a). Thermogram b shows the melting trace of the samples
after extraction in p-xylene at 375 K and drying. Drawn after data of Gedde and Jansson (1983b). (b) Bimodal melting
of a HDPE/LDPE 50/50 blend. Drawn after data of Gedde et al. (1981b)

Figure 1.37a shows the effect of extending the crystallization time under isothermal conditions of a
blend of two narrow linear polyethylene fractions. After 468 s, both parts of the high and low molar
mass fractions have crystallized and later melted in peak 2. It seems, based on the single high-
temperature melting peak, that the two fractions have co-crystallized. However, on prolonged
isothermal crystallization (45,000 s), the high molar mass fraction showed crystal thickening yielding
melting peak 3, whereas the low molar mass fraction remains intact in its extended chain crystal
structure and it melted in peak 2. Separate crystallization of the two narrow fractions was thus
confirmed. A linear polyethylene which was first crystallized during constant rate cooling and then
heated to 401 K, kept at this temperature for 24 h and then cooled at constant rate, showed two
melting peaks (Fig. 1.37b): peak 1 refers to material which melted at 410 K and crystallized during
the subsequent cooling stage. The material which remained crystalline at 401 K underwent crystal
thickening and crystal perfection. This material melts within peak 2. The examples of bimodal
melting presented in Figs. 1.36 and 1.37 involve two crystallite populations with the same unit cell
structure, an orthorhombic crystal. The difference in melting temperature is due to differences in
crystal thickness, which in turn is due to differences in molar mass and degree of chain branching. It is
very unusual that different polymers co-crystallize. Figure 1.38a shows bimodal melting of a blend of
polyethylene and isotactic polypropylene: in this case, the two polymers have different unit cell
structures, and crystallization leads to a complete segregation. Many crystalline polymers show more
than one crystal structure. Figure 1.38b shows a bimodal phase transition accompanying the melting
of semicrystalline trans-polybutadiene, where peak 1 is due to a polymorphic transition and peak 2 is
due to crystal melting.
The melting of the once-folded orthorhombic crystals of n-C294H590 (peak 1) leads to recrystalli-
zation (peak 2) into extended chain crystals that melt in the high-temperature peak 3 (Fig. 1.39).
DSC is one of the major techniques for assessing the crystallinity of polymers. Other important
methods are wide-angle X-ray scattering and density measurements (cf. Gedde and Hedenqvist
2019b). The degree of crystallinity is a major factor affecting basically all the material properties
of semicrystalline polymers such as stiffness, gas permeability and density.
36 1 Thermal Analysis of Polymers

Fig. 1.37 (a) Melting traces of a linear polyethylene 50/50 blend (narrow fractions: 2.5 and 66 kg mol1, respectively)
crystallized at 392 K for 468 s (upper) and 45,000 s (lower) and thereafter rapidly cooled to 300 K. (Drawn after data of
Rego Lopez et al. (1988)); (b) bimodal melting of linear polyethylene after annealing at 401 K for 24 h followed by
rapid cooling to 300 K. (Drawn after data of Gedde et al. (1981a))

Fig. 1.38 Left: bimodal melting of a 50/50 blend of linear polyethylene and isotactic polypropylene. (Drawn after
Aumnate et al. (2019)). Right: Bimodal melting of trans-polybutadiene due to a polymorphic transition. (Drawn after
data of Moraglio et al. (1965))

The mass crystallinity (wc) obtained from the melting enthalpy (Δhm) is based on the area under
the DSC melting peak. The choice of baseline is crucial, particularly for polymers of low crystallinity,
e.g. poly(ethylene terephthalate) (PET). Another problem arises from the fact that the melting
enthalpy is temperature-dependent. What temperature should be selected? Gray (1970) and
1.3 Thermal Behaviour of Polymers 37

Fig. 1.39 Melting trace of

a strictly monodisperse low
molar mass polyethylene
after rapid cooling from the
melt. (Drawn after data of
Ungar (1990))

Richardson (1976) have proposed two rigorous methods: the total enthalpy method and the peak area
method. Excellent agreement has been obtained for crystallinity data for samples of linear, branched
and chlorinated polyethylene and PET between the DSC and total enthalpy method and the X-ray
diffraction (WAXS) and density methods. Referring to Fig. 1.40, the mass crystallinity (wc) at
temperature T1 is given by (cf. Gedde and Hedenqvist 2019b):
Δhm
wc ¼ ð1:68Þ
Δh0m ðT 1 Þ

where Δhm0(T1) is the melting enthalpy of 100% crystalline polymer at T1, which is given by:
Tð1
   
Δh0m ðT 1 Þ ¼ Δh0m T 0m þ Cpa  Cpc dT ð1:69Þ
T 0m

where Cpa and Cpc are the heat capacities of the amorphous and crystalline components. This equation
is based on the well-known Kirchhoff law (cf. Gedde 2020). Heat capacity data for different polymers
were collected by Wunderlich and Baur (1970). A more recent compilation of data can be found on
the web in ATHAS (Wunderlich 1995), which contains thermodynamic data for 200 polymers and
small molecules.
The assessment of Δhm is best carried out by linear extrapolation of the post-melting scanning
baseline to establish the intersection with the pre-melting scanning baseline according to Fig. 1.40.
The point of intersection usually coincides with the onset of the non-linear thermal curve (Fig. 1.40).
The greatest problem is to determine the crystallinity of polymers that degrade at low temperatures
within the crystal melting range. Poly(vinyl chloride) belongs to this group. The baseline definition is
always a problem due to early thermal degradation and the low overall crystallinity, which leads to
melting over a broad temperature range. In these cases, the Kirchhoff law correction of Δhm0 shown
in Eq. (1.69) is essential. Crystallinity values obtained by DSC according to this methodology agree
with crystallinity data obtained by WAXS and density measurements. In some cases, the volume
crystallinity (ϕc) is of greater interest than the mass crystallinity (wc). The volume crystallinity is
calculated according to:
wc
ϕc ¼ ð1:70Þ
wc þ ð1  wc Þ  ρρc
a

where ρa is the density of the amorphous phase.

38 1 Thermal Analysis of Polymers

Fig. 1.40 Melting thermogram showing the total enthalpy method used to assess mass crystallinity

Important information about the structure and properties of semicrystalline polymers is obtained
from DMTA and dielectric spectroscopy. Figure 1.41 shows a sketch of the semicrystalline structure.
Composite equations are useful to analyze dynamic mechanical data. If we assume, for simplicity,
that the two phases are anisotropic, two limiting types of behaviour can be envisioned, upper bound
(Eq. (1.71)) and lower bound (Eq. (1.72)):
EU ¼ ϕc  Ec þ ð1  ϕc Þ  Ea ð1:71Þ

1 ϕ 1  ϕc
¼ cþ ð1:72Þ
EL Ec Ea

where EU is the upper-bound tensile modulus, which is related to the tensile stress acting along the
spherulite radius (r) (i.e. the same strain in both phases), and EL is the lower-bound tensile modulus,
which is related to the tensile stress acting along both x and y (i.e. the same stress in both phases).
Figure 1.42 shows the upper and lower moduli as functions of volume crystallinity, assuming that
Ec ¼ 10 Ea.
Figure 1.42b shows that the relaxation strengths (Gu – Gr) of the two mechanical relaxation
processes (α and β) diminish with decreasing crystallinity. It can be concluded that both the glass
transition (α process) and the sub-glass process (β process) occur in the amorphous phase of the
semicrystalline polymer. Boyd (1979) used the shear modulus version of the Halpin–Tsai equation
(Eq. (1.73); exchange E with G) to model Illers torsion pendulum data for linear polyethylene
covering wide ranges both in crystallinity and temperature. The Halpin–Tsai equation contains
three adjustable parameters at any given temperature: Ga, Gc and ξ (the aspect ratio). The Halpin–Tsai
equation for tensile modulus is given by:
1.3 Thermal Behaviour of Polymers 39

Fig. 1.41 Sketch of

crystal-amorphous layer
stack useful in making a
simple view of the
composite structure of a
semicrystalline polymer.
An uniaxial stress acting
along r ensures that the
strain is the same in both
phases (upper bound). An
uniaxial stress acting either
along x or y ensures that the
stress is the same in both
phases (lower bound)

Fig. 1.42 (a) Upper-bound and lower-bound moduli as functions of volume crystallinity (ϕc) with constrained modulus
values for the crystalline component (1 au) and amorphous component (0.1. au). (b) The dependence of volume
crystallinity (ϕc) of the unrelaxed and relaxed shear moduli associated with the two mechanical relaxations α and β for
poly(ethylene terephthalate) (PET). (Drawn after data of Boyd (1985))

Ea ðEc þ ξðð1  ϕc ÞEa þ ϕc Ec ÞÞ 1 þ ξ  EU

E¼ ¼ ð1:73Þ
ð1  ϕc ÞEc þ ϕc Ea þ ξEa ξ þ E1 L

By varying the aspect ratio, it is possible to model both upper-bound (ξ ! 1) and lower-bound
(ξ ! 0) conditions and anything in between. A good fit was obtained using a single (universal) aspect
ratio value (ξ  1; close to the lower bound) for all linear polyethylenes in the temperature range from
200 to +100  C. Values for Ga and Gc were obtained at each temperature (Fig. 1.43). It is fascinating
to see that the relaxed rubber modulus is several hundred MPa. The rubber phase in semicrystalline
linear polyethylene is very rigid.
40 1 Thermal Analysis of Polymers

Fig. 1.43 Shear modulus of the crystalline (Gc) and amorphous (Ga) components of linear polyethylene calculated
from torsion pendulum data (1 Hz) by Illers (1973) using the Halpin–Tsai equation (Eq. (1.73) exchange E with G).
Drawn after data of Boyd (1979) and a replicate of a graph in Gedde and Hedenqvist (2019b). With permission from
Springer Nature Switzerland AG. The mechanical relaxations designated α, β and γ are shown together with the relaxed
rubber modulus Gr

Fig. 1.44 Argand

diagrams for the α-process
(glass transition) in
amorphous PET (graph a)
and semicrystalline PET
(crystallinity ¼ 50%; graph
b). Drawn after data of
Boyd and Liu (1997)

An interesting observation made by Boyd (1985) was that the dielectric glass transition of a
semicrystalline polymer exhibits only symmetric broadening according to the Cole–Cole equation
(Eq. (1.52)), whereas a fully amorphous polymer shows an asymmetrically broadened glass transition
which is best described by the Havriliak–Negami equation (Eq. (1.57)). Poly(vinyl chloride), for
instance, shows symmetrical broadening, even though its crystallinity is only 5–10%. Figure 1.44
shows Argand diagrams for the α process for amorphous PET (graph a) and for semicrystalline PET
(graph b). The relaxation strength can be expressed by the difference between values of the relaxed
and unrelaxed dielectric constants, i.e. ε´r – ε´u, which for amorphous PET amounts to 2.2 and for the
1.3 Thermal Behaviour of Polymers 41

Fig. 1.45 Relaxation strength (εr – εu) as a function of volume crystallinity (ϕc) of the α process in PET. Drawn after
data of Coburn and Boyd (1986) and Boyd and Liu (1997). The straight (upper) line indicates the expected relaxation
strength for a polymer with an unconstrained amorphous phase. The decrease in relaxation strength is due to the
constraint of the presence of a rigid amorphous fraction (RAF). The magnitude of the mobile amorphous fraction
(MAF) is indicated in the diagram

PET with 50% crystallinity to 0.7. This suggests that the intensity of the α process decreases more
strongly than just merely according to the decrease in the proportion of the amorphous component.
Figure 1.45 shows that the amorphous phase only partially contributes to the intensity of the glass
transition. The straight line in the graphs shows the behaviour of an unconstrained amorphous phase;
the actual relaxation strength of the PET sample with 60 vol.% crystallinity is only half this virtual
value. The idea of dividing the non-crystalline fraction of a semicrystalline polymer into two different
parts, a mobile amorphous fraction (MAF) and a rigid amorphous fraction (RAF), was proposed by
Suzuki et al. (1985). The mass fraction of MAF (wMAF) is calculated as the ratio of the change in heat
capacity associated with the glass transition of the semicrystalline polymer (ΔCp,sc) to that of a fully
amorphous polymer (ΔCp,a):
ΔCp,sc
wMAF ¼ ð1:74Þ
ΔCp,a

The mass fraction of RAF (wRAF) is then given by:

wRAF ¼ 1  wc  wMAF ð1:75Þ

A similar, simple approach is shown in Fig. 1.45 for the relationship between relaxation strength and
volume crystallinity. The relaxation strength of the sub-glass processes does not follow the same
pattern as the glass transition, but it is approximately proportional to the crystallinity (Fig. 1.46).
The kinetics of crystallization, which are of interest for both academia and industry, are preferably
studied under isothermal conditions by DSC, dilatometry or TOA. These methods reveal the overall
crystallinity, volume (ϕc) or mass (wc) based, as a function of time (t). TOA also makes it possible to
42 1 Thermal Analysis of Polymers

Fig. 1.46 Relaxation strength (εr – εu) as a function of volume crystallinity (ϕc) of the β process in PET. (Drawn after
data of Coburn and Boyd (1986) and Boyd and Liu (1997))

assess the linear growth rate of spherulites as viewed in the microscope (cf. Sect. 1.2.5). Further
details, revealing the growth rate of single crystal lamella, can be obtained by atomic force micros-
copy (cf. Sect. 1.2.5).
One of the problems in studying the crystallization of polymers is that the temperature window is
narrow. At high temperatures, crystallization is very slow, and the accuracy is reduced because of the
baseline drift. Another problem is that the time span for the measurements is very long, days or even
weeks, and in many cases, it is not possible to make such long reservations of a highly booked
instrument. The data quality (at high temperatures) is improved significantly by interrupting the
isothermal period with immediate heating at (say) 10 K min1. The melting endotherm is then in most
cases sharp, and an accurate crystallinity value can be determined. The procedure has to be repeated
with different crystallization times in order to obtain a curve of crystallinity vs. crystallization time.
An even more efficient method is to transfer molten samples to an oven and then to take them out after
different periods and cool them rapidly (quenching). An efficient way to keep the temperature
constant is to house the samples in a heavy metal piece. The crystallinity before cooling is obtained
by measuring the high-temperature melting peak only; the low-temperature peak is associated with
the material which crystallize during cooling from the isothermal crystallization temperature. The
following precautions to avoid degradation during long-term high temperature are useful in order to
obtain reliable data: (i) inert atmosphere; (ii) removing dissolved oxygen before the thermal exposure
samples; and (iii) adding extra stabilizer to the polymer sample. In order to extend the crystallization
window towards lower temperatures, high cooling rates are required. The classical DSC instruments
permitted cooling rates of the order of 10 K s1. The recent development of fast scanning calorimetry
with ultra-high cooling rates makes possible studies of crystallization at very low temperatures
(Fig. 1.47). This technique opens new opportunities. The data presented in Fig. 1.47 suggest a change
in the crystallization mechanism at 50  C. Toda et al. (2016) present a modern account about new
findings about polymer crystallization based on fast scanning calorimetry. Another recommended
text is the book by Schick and Mathot (2016).
1.3 Thermal Behaviour of Polymers 43

Fig. 1.47 The

characteristic
crystallization time (τ) as a
function of temperature for
isotactic polypropylene.
Note the two maxima in
crystallization rate at
25 and 75  C. (Drawn after
data of De Santis et al.
(2007) and Silvestre et al.
(2007))

Gedde and Hedenqvist (2019c) have presented a modern text on polymer crystallization, with a
number of thermal analysis landmarks. The overall crystallization rate, especially in the initial stages,
is sometimes analyzed by the general Avrami equation:
ϕc
1 ¼ exp ðKtn Þ ð1:76Þ
ϕ c1

where n and K are constants, ϕc1 is the volume crystallinity at time t and ϕc1 is the maximum
crystallinity attained. The Avrami exponent (n) depends on nucleation type, the geometry of crystal
growth and the kinetics of crystal growth (cf. Gedde and Hedenqvist 2019c). The kinetics at low
degrees of conversion usually follow the Avrami equation but deviate from this trend at higher
degrees of conversion. An excellent text showing in detail how to apply this method is given by
Lorenzo et al. (2007).
A significant part of the work on crystallization kinetics has dealt with the temperature dependence
of the crystallization rate according to the kinetic theory of Lauritzen and Hoffman (cf. Hoffman et al.
1975). The experimental data related to the linear growth rate (Gr) of spherulites (axialites) are
obtained by hot-stage polarized light microscopy at different constant temperatures, and the data are
adapted to the equation:
   
U ∗ K g
Gr ¼ Gr0  exp  exp ð1:77Þ
Rð T c  T 1 Þ T c ΔTf

where Gr0 is a constant; U* is a constant (in joules per mole) for the short-distance diffusion of the
crystallizing segments; R is the gas constant; Tc is the crystallization temperature; T1 is a temperature
which is related to the glass temperature; Kg is the ‘nucleation kinetic constant’ which depends on the
surface energies of the crystals formed, the equilibrium melting point and the mechanism of
crystallization; ΔT is the degree of supercooling (¼ Tm0– Tc, where Tm0 is the equilibrium melting
point and Tc is the crystallization temperature); and f is a correction factor close to unity that corrects
for changes in the heat of fusion with temperature. The factor f depends on the degree of supercooling
(ΔT ). Mathot (1994) presents a detailed discussion of the correction factor. A DSC analogue of
Eq. (1.77) may be written as:
   
1 U ∗ K g
¼ C  exp  exp ð1:78Þ
t0:5 Rð T c  T 1 Þ T c ΔTf

where t0.5 is the time required to attain 50% of the final crystallinity and C is a constant.
44 1 Thermal Analysis of Polymers

1.3.2 Amorphous Polymers

The stiffness of a typical amorphous polymer changes by three orders of magnitude from a few GPa in
the glassy state (low-temperature side) to a few MPa in the rubbery state (high-temperature side). The
glass transition appears to be a second-order phase transition, where volume (V ) and enthalpy (H ) are
continuous functions through the transition temperature range (Fig. 1.48). However, if the amorphous
polymer is cooled from temperature (Ta), a break in the rectilinear H, V ¼ f(T ) curve appears at the
glass transition temperature (Tg). If the cooling is stopped at a temperature close to Tg, at Tb in
Fig. 1.48, and the sample is kept at this temperature, both H and V decrease as a function of time. This
behaviour is often denoted as physical ageing, and it shows that glassy amorphous polymers are
seldom at equilibrium and that the observed glass transition is a kinetic phenomenon rather than a
truly thermodynamic transition. The classical texts by Struik (1978) and Hutohinson (1995) are
recommended reading about physical ageing also including the effects on the properties of glassy
polymers. Physical ageing leads to more densely packed material showing higher stiffness and lower
impact strength. The kinetics of physical ageing can be monitored by dilatometry and DSC.
Figure 1.49 shows how the difference in enthalpy between a young material (rapidly cooled) and a
physically aged material can be calculated based on DSC scans. The calculation is based on the
computing difference in integrated change in enthalpy between two temperatures for the two samples:
T1 being well below and T2 being well above the glass transition region. Some aged glassy polymers
show an endothermic peak at the ΔCp step. This peak is not due to a true first-order phase transition. It
is due to superheating of the glass transition. The enthalpy difference between the reference sample
(e.g. a rapidly cooled sample) and the aged sample originates to a large extent from the area under the
superheating peak for the case shown in Fig. 1.50.
A very important and useful concept is the fictive temperature (Tf), at which the glassy H-line
intersects the liquid H-line. Figure 1.51 shows two different methods for the assessment of Tf. The
bottom graphs show the enthalpy obtained by integration of the upper DSC thermograms; the fictive
temperature of the sample which shows a slow glass transition with superheating (Fig. 1.51a) is
obtained by extrapolating the linear part of the liquid H-line towards lower temperatures and finding
the intersection with the glassy H-line. The graph presented in Fig. 1.51b is simpler without
superheating.
The fictive temperature can also be determined from a conventional DSC thermogram (Fig. 1.51,
upper graphs). In the case of superheating, which is revealed by the peak (Fig. 1.51a), the fictive
temperature is much lower than the step in the DSC curve. The areas A and B are equal when the
lower temperature limit of area A is precisely Tf.. For a conventional glass transition, as shown
Fig. 1.51b, Tf is at the midpoint of the step in the differential heat flow. The fictive temperature

Fig. 1.48 Schematic

representation of the glass
transition in an amorphous
polymer. The decrease in
volume and enthalpy at Tb
is referred to as physical
ageing
1.3 Thermal Behaviour of Polymers 45

Fig. 1.49 (a) Enthalpy (H ) as a function of temperature (T ) for (i) ‘young’ rapidly cooled sample (y) following path
A–B–C–Y; this sample is then heated following the red path; (ii) an ‘aged’ sample following path A–B–C–D–AG: this
sample is then heated following the blue path. (b) DSC thermogram displaying traces of the young sample and of the
aged sample. The difference in enthalpy between the two states at the annealing temperature and at T1 is proportional to
the area shown in the graph

Fig. 1.50 (a) Enthalpy (H ) as a function of temperature (T ) for (i) a ‘young’ rapidly cooled sample (y) following path
A–B–C–Y; this sample is then heated following the red path; (ii) an ‘aged’ sample following path A–B–C–D–AG: this
sample is then heated following the blue path. (b) DSC thermogram displaying traces of the young and of the aged
sample. The difference in enthalpy between the two states at the annealing temperature and at T1 is proportional to the
area A–B in the graph

provides a compact description of the structure of a glassy polymer and is simply related to the
enthalpy (H ) of a glassy polymer at temperature T according to:
 
H  H eq ¼ ðT f  T Þ  Cp,l  Cp,g ð1:79Þ

where Heq is the equilibrium enthalpy at temperature T, Cp,l is the heat capacity of the liquid polymer
and Cp,g is the heat capacity of the glassy polymer. Note that H ¼ Heq if T ¼ Tf.
An amorphous polymer, which is heated at a higher rate than its original cooling rate, shows a
glass transition with a hysteresis peak due to superheating (Fig. 1.52). If the same polymer sample is
46 1 Thermal Analysis of Polymers

Fig. 1.51 Sketches showing two different methods to obtain the fictive temperature (Tf). The upper two graphs are
from Gedde and Hedenqvist (2019e) with permission from Springer Nature

Fig. 1.52 DSC traces of the glass transition region of two samples given the same thermal pre-treatment followed (top)
by rapid heating and (bottom) slow heating. Schematic drawing. The broken lines indicate the thermogram for a sample
heated at the same rate as that at which it was previously cooled

heated much more slowly than its initial cooling rate, the sample undergoes physical ageing during
heating, which is revealed by an exothermal peak just below the glass transition region (Fig. 1.52).
Figure 1.53 shows dependence of the fictive temperature on the cooling rate. The kinetic character
of the glass transition causes the sample to adhere to the equilibrium line at progressively lower
temperatures with decreasing cooling rate. Slow cooling thus promotes a low fictive temperature.
According to the data presented in Fig. 1.53, the change in fictive temperature caused by cooling rate
depends on the actual value of the cooling rate. For cooling rates attainable with conventional DSC
instruments (< 1 K s1), the change in Tf is 3 K if the cooling rate is changed by a factor of
10 (cf. Moynihan et al. 1976; Ferry 1980). The corresponding change in Tf at high cooling rates
1.3 Thermal Behaviour of Polymers 47

Fig. 1.53 The fictive

temperature plotted as a
function of cooling rate for
atactic polystyrene samples
cooled from the melt at
different rates. (Drawn
after data of Schawe
(1998))

Fig. 1.54 Reversing and

non-reversing heat flow
thermograms of
polycarbonate. (Drawn
after data of Hutchinson
(1996))

(1000–10,000 K s1) is 5 K. These data have been obtained in the more recent fast scanning
calorimeters.
Figure 1.54 shows modulated DSC thermograms of the glass transition region of polycarbonate.
The reversing heat flow curve shows an endothermic step in the curve, whereas the sharp minimum
(valley ¼ ‘negative’ peak) in the non-reversing heat flow curve indicates physical ageing. The area
under this negative peak is a measure of the entropy production during heating. The glass transition
temperature measured by the step in the reversing heat flow curve is different from that obtained by
conventional DSC, and it depends on the period of the modulation rather than on the steady heating
rate.
The DSC trace from the glass transition region of highly cross-linked, fully amorphous polymer
shows only a vague step in the curve relating heat flow rate to temperature. It is better to use DMTA
for such an analysis, because the transition is much more prominent. DMTA provides a wealth of
information about the glass transition (cf. Sect. 1.2.4). The glass transition temperature is best
reported as the inflection point of the jump in storage modulus (E’) or as the maximum in the loss
modulus (E”) (cf. Fig. 1.21). Recommended texts about the thermal behaviour of glassy polymers are
those of Boyd and Smith (2007a, b), Gedde and Hedenqvist (2019a) and Haward and Young (1997).
The morphology of amorphous polymer blends is of practical importance. The impact strength of
stiff and brittle glassy polymers (e.g. atactic polystyrene) is improved by including a few percent of an
48 1 Thermal Analysis of Polymers

Fig. 1.55 DSC

thermogram of ABS
showing two glass
transitions. (Drawn after
data by Bair (1970))

Fig. 1.56 Total, reversing

and non-reversing heat
flow thermograms of a
blend of poly(ethylene
terephthalate) and ABS
plastics. (Drawn after data
of TA Instruments (1996))

immiscible elastomer such as polybutadiene. Provided the polymers have different glass transition
temperatures, DSC, TMA, DMTA or DETA are techniques which indicate whether the polymers are
miscible or whether they exist in different immiscible phases. Figure 1.55 shows the thermogram for
an immiscible mixture of styrene-acrylonitrile (SAN) copolymer and polybutadiene. This blend is a
material with a high impact strength material referred to as acrylonitrile butadiene styrene (ABS)
plastic. Two glass transitions are observed, a low-temperature transition associated with polybutadi-
ene and a high-temperature transition associated with the SAN copolymer.
Occasionally, as shown in Fig. 1.56, phase transitions show an overlap of the constituents in
polymer blends. Using modulated DSC, it is possible to separate the irreversible cold crystallization
of poly(ethylene terephthalate), which appears in the non-reversing heat flow thermogram, from the
glass transition of poly(styrene-co-acrylonitrile) in ABS, which the latter appears clearly in the
reversing heat flow thermogram.
Miscible polymer blends are less commonly found. One of the most studied blends is that of
polystyrene and poly(phenylene oxide) (PPO), which is miscible in all proportions and the films of
which are optically clear. Only one glass transition, intermediate in temperature between the Tg’s of
polystyrene and PPO, has been reported. The dependence of the Tg of a miscible binary blend on the
composition of the blend has been described by several equations. Gedde and Hedenqvist (2019a)
present a comprehensive review of this topic.
1.3 Thermal Behaviour of Polymers 49

1.3.3 Liquid Crystalline Polymers

Liquid crystalline polymers attracted considerable interest between 1980 and 2000, and a detailed
presentation is given by Gedde and Hedenqvist (2019f). A liquid crystal possesses a structure
intermediate between a liquid and a crystal, an anisotropic liquid. A conventional liquid exhibits
only local order with a correlation distance of the order of only a few nm. A liquid crystalline polymer
exhibits orientational order over much longer distances, typically longer than 1 μm. The chemical
structure of most liquid crystalline polymers is such that they are often referred to as rod-shaped. A
close packing of rod-shaped molecules (typically a combination of stiff, linear moieties (mesogenic
units) and more flexible groups (spacer groups)) implies that the rods are oriented along a common
director, i.e. the structure is anisotropic. Polymer with disc-shaped molecules forms discotic
structures, i.e. another type of anisotropic structure than the liquid crystalline structures of the
rod-shaped molecules.
Main-chain polymers, with the mesogenic units in the main chain, exhibit a unique combination of
good processing and good mechanical properties and are now used as engineering plastics. Side-chain
polymers, with the mesogens in the side chains, are mainly used in speciality polymers with
applications in electronics and photonics. Liquid crystalline polymers exhibit a number of possible
phase transitions, as shown in Fig. 1.57.
DSC, TMA and hot-stage polarized microscopy are, together with X-ray diffraction, efficient
methods for structural assessment at different temperatures. A combination of these methods is
preferably used. X-ray diffraction of aligned samples is the most reliable method of structural
assessment (cf. Gedde and Hedenqvist 2019f).
When the isotropic melt cools from a high temperature, it is ultimately transformed into a liquid
crystalline phase, i.e. a mesophase. Many different liquid crystalline structures have been reported.
Two major groups differing in degree of order exist: nematics (structures which possess orientational
order only of the mesogens) and smectics (structures which possess both orientational and positional
order of the mesogens). The high-temperature phase transition is revealed by DSC as a first-order
transition (Fig. 1.57) and by hot-stage polarized microscopy from the formation of birefringent
(oriented) structures. The phase assignment of a mesophase can be achieved by polarized light
microscopy according to the detailed scheme presented by Demus and Richter (1978) or by X-ray
diffraction (Gedde and Hedenqvist 2019f). Figure 1.58 shows two examples of structures revealed by
polarized microscopy. Schlieren textures with disclinations of strength 1/2 are characteristic of
nematics (Fig. 1.58a; cf. Gedde and Hedenqvist 2019f). Focal conic structures are found in smectic A
phases (Fig. 1.58b).

Fig. 1.57 Phase transitions (sketched in a diagram of volume (V ) or enthalpy (H ) versus temperature (T )) in
thermotropic liquid crystalline polymers showing the isotropic melt (I), nematic phase (N), smectic phase (S),
semicrystalline state (SC), nematic glass (NG) and smectic glass (SG). The transitions I ! N, N ! S, S ! SC and
N ! SC are first-order transitions. Glass transitions are marked in red. A first-order transition between I and S is also
possible (not shown in the graph)
50 1 Thermal Analysis of Polymers

Fig. 1.58 Polarized photomicrograph of liquid crystalline compounds: (a) nematic structure showing disclinations
with two (disclination strength ¼  ½) or four radiating dark bands; (b) smectic A phase structure. The
photomicrographs were taken by Fredrik Sahlén, KTH Royal Institute of Technology, Stockholm, and have been
displayed earlier in Gedde and Hedenqvist (2019e) with permission of Springer Nature

Fig. 1.59 Dielectric loss as a function of temperature at the frequencies shown in the diagram. The structure of the
repeating unit of the polymer is also shown. (Drawn after data of Gedde et al. (1994))

At lower temperatures, a number of liquid crystalline transitions may occur which can also be
recorded by DSC as first-order transitions (Fig. 1.57). Hot-stage polarized light microscopy and X-ray
diffraction are used to determine the nature of these transitions. At lower temperatures, solid crystals
may be formed. The latter are revealed by DSC as a first-order transition, by TMA as an increase in
sample stiffness and by X-ray diffraction as sharp Bragg reflections. Some liquid crystalline
polymers, e.g. copolyesters, can be supercooled to a glassy state without crystallizing (Fig. 1.57).
DETA provides valuable information about sub-glass processes in liquid crystalline polymers. In
Fig. 1.59, two dielectric sub-glass relaxation processes, denoted β and γ, are displayed. The
low-temperature process (γ) showed an activation energy of 30 kJ mol1 and was assigned to local
motion within the spacer group (Gedde et al. 1994). The high-temperature sub-glass process (β) was
1.3 Thermal Behaviour of Polymers 51

assigned to reorientation of the carboxylic group in the phenyl benzoate moiety. The molecular
interpretation of the different relaxation processes is based on the dielectric spectra of a series of
samples with different structures. For instance, polymers lacking a carboxylic group show no β
relaxation peak. Comparison with NMR data has also proven to be very useful for a molecular
understanding of the relaxation processes.

1.3.4 Polymer Degradation

With very few exceptions, polymers are susceptible to degradation both during the melt processing
and in the subsequent use. Reactions with thermal oxidation and photo-oxidation are for many
polymers the dominating degradation reactions. Stabilizers, e.g. antioxidants, increase the stability
and considerably extend the life of many polymer products. Exposure of a plastic material to heat and
oxygen leads to chemical consumption of the antioxidant. Migration of the antioxidant to the
surrounding media is in some cases, however, the main reason for the loss of stability of the material
(Billingham et al. 1981; Viebke and Gedde 1998). It is therefore important to use efficient methods to
determine the concentration of effective antioxidant in samples exposed to an ageing program. Data
which show the time to reach depletion of antioxidant at different temperatures, can be used to
extrapolate to service conditions; cf. Smith et al. (1992). Thermal analysis, primarily DSC, is
frequently used for this purpose.
The oxidative induction time (OIT) is measured by heating the polymer sample in a nitrogen
atmosphere to a high temperature, typically around 200  C for commodity polyolefins. After a
constant temperature has been established, the atmosphere is switched to oxygen, and the time
(OIT) to the start of an exothermic oxidation reaction is determined (Fig. 1.32). The OIT exhibits
an Arrhenius temperature dependence, and there is a linear relationship between OIT and the content
of efficient antioxidant, since the consumption of antioxidant under these conditions follows zero-
order kinetics:
dc ΔE pffiffiffiffiffiffiffi
¼ k0 e RT pO2
dt ð1:80Þ
dc ΔE  
¼ k00 e RT pO2 ¼ constant
dt

where c is the concentration of efficient antioxidant, t is the time, k0 and k0´ are constants, ΔE is the
activation energy, R is the gas constant, T is the temperature (in kelvin) and pO2 is the partial pressure
of O2. Equation (1.80) can be integrated under isothermal conditions to give:
ΔE
c0 ¼ k00 e RT OIT ð1:81Þ

where c0 is the initial concentration of efficient antioxidant. In order to use OIT data as an absolute
method for the determination of the antioxidant content, the system must first be calibrated to
determine the constants k00 and ΔE. OIT measurements can also be made by TG. The first indication
of oxidation is a small increase in sample mass. At a later stage, when degradation leads to the
formation of volatile products, the sample mass decreases strongly.
When the constants k00 and ΔE have been established, the antioxidant content can be indirectly
determined by recording the thermogram at a constant heating rate in oxygen and determining the
oxidation temperature (Tox) as the temperature of the exothermic deviation from the scanning
baseline. The relationship between Tox and the antioxidant concentration (c0) is obtained by
integrating Eq. (1.80) to give (Fig. 1.60):
52 1 Thermal Analysis of Polymers

Fig. 1.60 (a) Thermograms showing the method for the assessment of oxidation induction time (OIT) at 190  C. The
samples are squalene (hydrocarbon liquid analogue of polyethylene) containing 0.1 and 0.2 wt. % of phenolic
antioxidant (Irganox 1010) and pristine squalene. The switch to oxygen (1 atm) was at time zero. OIT is defined as
the time at which the pre-oxidation baseline intersects the tangent at a point at a deviation of 0.2 W g1 from the
pre-oxidation baseline. (b) OIT (190  C) as a function of the antioxidant concentration. (From Azhdar et al. (2009) with
permission from Elsevier)

Fig. 1.61 Oxidation

temperature as a function
of antioxidant
concentration (Irganox
1010 in polyethylene). The
continuous line is
calculated based on
Eq. (1.76). Drawn after
data of Karlsson et al.
(1990)

Tðox
k00 ΔE
c0 ¼ dT
e RT dT ð1:82Þ
dt
0

This integral can only be solved numerically. An example of the result of this calculation is shown
in Fig. 1.61. Since OIT is proportional to c0, the dynamic method is less suited for the determination
of antioxidant content in highly stabilized systems (high c0), but it is sufficiently sensitive for small
variations in c0 in systems with small c0 values.
OIT/Tox measurements provide rapid and reliable results. Many materials contain a combination of
one or two primary antioxidants (e.g. hindered phenols) and a secondary antioxidant (e.g. phosphite).
Thermal analysis provides no information about the separate concentration of the different
antioxidants but rather an overall assessment of the stability based on the phenolic part of the
stabilizing package. Extraction followed by high-performance liquid chromatography (HPLC) is
one of the main techniques for determining antioxidant concentration, but it is much more time-
1.3 Thermal Behaviour of Polymers 53

Fig. 1.62 The average

OIT through a pipe wall
plotted against the square
root of the exposure time at
95  C (pressure testing).
The inset figure shows a
scanning electron
micrograph of a
precipitated antioxidant
crystal. From Viebke et al.
(1996) with permission of
the Society of Plastics
Engineers, USA

consuming than DSC. Extensive information about the OIT measurements and their interpretation has
been provided by Billingham et al. (1981). In addition, Karlsson et al. (1992); Smith et al. (1992); and
Viebke and Gedde (1998) have used OIT data obtained by DSC to assess the migration of
antioxidants from polyolefins and to predict the lifetime of polyolefin hot water pipes.
Figure 1.62 shows OIT data from samples taken from a polyethylene pipe pressure tested at 95  C.
A linear relationship between OIT and the concentration of this phenolic antioxidant was established
(Viebke et al. 1996). The loss of antioxidant is primarily due to diffusion from the bulk of the pipe to
its outer and inner walls, i.e. migration of antioxidant to the surrounding media. This is the dominant
process after 1000 h exposure (ca. 30 h1/2), i.e. the low slope region denoted Regime B in Fig. 1.62.
During the first 1000 h, internal phase separation of antioxidant occurs (cf. the precipitated crystal
showed in the inset figure in Fig. 1.62). This process, denoted Regime A, follows a very steep line in
the OIT – time1/2 diagram. One lessen to learn is that by extrapolating the Regime A data to OIT ¼ 0,
depletion of the antioxidant system is predicted to occur after 1600 h. The correct prediction based on
Regime B data is ten times longer (i.e. after ca. 16,000 h).

1.3.5 Further Applications of Thermal Analysis in Polymer Science and

Technology

TG is a suitable method for determining the concentration of volatile species in a polymer. A typical
analysis involves heating the sample to a temperature at which the vapour pressure of the sorbed
component is significant and holding the sample at that temperature until the sample mass reaches a
constant value. The relative decrease in sample mass is equal to the mass fraction of the volatile
component. Sorbed water in polyamide 6, sorbed propane in polyethylene and plasticizing dioctyl
phthalate (DOP) in polyvinylchloride can be determined in this way, and TG can accurately
determine the transport properties (solubility and diffusivity) of polymers. A thin film sample with
a width-to-thickness ratio greater than 10 is first saturated with the gaseous or liquid penetrant and
then transferred to the TG. The mass loss can be determined as a function of time at different
temperatures. From the desorption curves, it is possible to determine both the solubility and the
diffusivity, even when the diffusivity is dependent on concentration. The mathematics involved in the
54 1 Thermal Analysis of Polymers

Fig. 1.63 Heating

thermogram of unsaturated
polyester with unreacted
styrene (lower curve) and
after a heating and cooling
scan of the sample (upper
curve)

calculations can be quite advanced, and the interested reader can find useful information in Crank
(1986) and in Chap. 7.
TG is also suitable for the determination of the overall content of fillers in polymers. The fraction
of carbon black can be selectively determined, since it oxidizes completely at elevated temperatures
forming volatile species.
Polymerization and vulcanization, both of which are exothermic reactions important in polymer
technology, are conveniently recorded by DSC. It is, however, difficult in many cases to follow the
kinetics under isothermal conditions. The reactions are initiated very rapidly at elevated
temperatures, and this sets an upper temperature limit for the analysis, while the activation energy
(controlling the temperature shift in the rate constant) and the sensitivity of the calorimeter set the
lower temperature limit. Some reactions can thus only be studied at constant heating rate conditions,
when the method proposed by Barton (1973) can be applied enabling the activation energy of the
process to be obtained.
Another useful application of DSC is in the determination of unreacted functional groups in
thermosets, e.g. unreacted styrene in a cross-linked polyester (Fig. 1.63). In the first heating scan,
an exothermal peak appears above the glass transition (lower thermogram). The increased segmental
mobility above Tg enables the initiation of a post-curing reaction. The exothermal peak is absent in
the second heating scan (upper thermogram in Fig. 1.63). It is possible to determine the molar fraction
(cu) of unreacted species simply from the evolved exothermal enthalpy (ΔHexo) and the molar energy
of polymerization (cross-linking), ΔHcross according to:
ΔH exo
Cu ¼ ð1:83Þ
ΔH cross

1.4 Summary

Thermal analysis involves a group of analytical methods which enable a physical property of a
sample to be measured as a function of temperature or time. The properties that are studied are
typically enthalpy (DSC), dimensions (dilatometry, DMA), anelastic properties (DMTA), mass (TG),
dielectric properties (DETA), optical properties (TOA) and structural parameters (e.g. by AFM and
X-ray scattering using instruments equipped with hot-stages). Reproducible and accurate results are
obtained, and they allow a great number of materials and phenomena involving both physical and
chemical aspects to be studied. All thermal transitions of polymers, including crystallization, melting
and glass formation, are irreversible, and care must be taken in the interpretation of the data. The
thermal analysis family is still growing; new techniques are being developed, where fast scanning
calorimetry and local techniques probing the behaviour on the nanoscale are important developments.
1.5 Exercises 55

1.5 Exercises

1.1 Determine the mass crystallinity from the following DSC data: Δhm ¼ 140 J g1; Δhm0 ¼ (T 0m
 418 K) ¼ 293 J g1; and T1 ¼ 370 K; use the specific data of Wunderlich and Baur (1970);
and calculate the mass crystallinity according to the total enthalpy method.

Temperature (K) cpc (cal mol1 K1)a cpa (cal mol1 K1)a
310 5.40 7.68
330 5.89 7.84
350 6.38 7.99
370 7.02 8.15
380 7.37 8.23
390 7.72 8.31
400 8.14 8.38
410 8.63 8.46
420 – 8.54
a
Per mole of repeating units (14 g mol1)

1.2. Calculate the molar entropy of fusion for linear polyethylene from the following data:
equilibrium enthalpy of fusion Δhm0 ¼ 293 J g1 and equilibrium melting point ¼ 415–418 K
(different values have been reported).
1.3. The entropy of fusion of polyamide 6 is similar to that of polyethylene counted per mole of
main chain atoms, but the equilibrium melting point is 533 K for polyamide 6 compared to
415–418 K for polyethylene. Explain why!
1.4. A medium density polyethylene exhibits two melting peaks, at 385 K and 408 K, respectively,
even after a constant rate cooling from the melt. Give possible reasons!
1.5. Describe a suitable thermal analysis experiment by which the moisture content in polyamide 6
can be determined.
1.6. It is known that a thermoplastic material consists of two fillers: CaCO3 and carbon black.
Design a suitable thermo-analytical experiment by which the separate filler contents can be
determined.
1.7. The thermal expansion data given in Fig. 1.36 were obtained for injection-moulded samples of
binary blends of a liquid crystalline polymer (Vectra) and polyethersulphone (PES). What do
the data tell about the structure of the blends?
1.8. Calculate the volume expansion coefficients for blends of 0, 40 and 80% of Vectra, using the
data shown in Fig. 1.64.

Fig. 1.64 Linear thermal

expansion coefficient
parallel to (II) and
perpendicular to (⊥) to the
moulding directions at
room temperature.
Samples: blends of Vectra
(liquid crystalline polymer)
and PES. Drawn after data
of Engberg et al. (1994)
56 1 Thermal Analysis of Polymers

Fig. 1.65 Oxidative

stability of samples taken
from an unexposed
medium-density
polyethylene pipes as a
function of distance from
the inner wall. Oxidation
temperature measurements
were obtained at a scan rate
of 10  C min1 and OIT at
190  C. (Drawn after data
of Karlsson et al. (1992))

Fig. 1.66 Dielectric loss

as a function of
temperature. Isochronal
curves of a hyperbranched
polyester at different
frequencies as shown in the
graph. (Drawn after data of
Malmstr€om et al. (1994))

1.9. Draw schematic DSC curves for a fully amorphous polymer cooled at two different rates, 1 and
100 K min1. The glass–rubber transition is within the temperature range of the cooling scan.
Draw schematically the heating thermograms from the sample originally cooled at 1 K min1
thereafter heated at 0.1 and 100 K min1.
1.10. Suggest a series of DSC experiments by which the kinetics of physical ageing can be recorded.
1.11. The DSC trace of a polymer shows a large low-temperature first-order transition and a very
small first-order transition at higher temperatures. TOA shows that the major change in light
intensity is associated with the high-temperature transition. Suggest an explanation of
these data.
1.12. The oxidation temperature data shown in Fig. 1.65 show only moderate variations, 236 to
242 K, whereas the OIT data show a considerable variation, from 50 to 100 min. Are the two
sets of data mutually consistent?
1.13. The activation energy for the consumption of antioxidant in the molten state was determined to
be 176 kJ mol1. Calculate the OIT variation (with distance from the inner wall) at 220 and
230  C.
1.14. Calculate the activation energies of the relaxation processes shown in the isochronal scans
displayed in Fig. 1.66.
1.15. A thermodynamics question: Derive Kirchhoff’s law, which relates the change in enthalpy
associated with a chemical reaction or a phase transition to the temperature.
1.16. The thermodynamic quantity enthalpy is a primary concept used in thermal analysis. Why?
1.17. Design an experiment using TG to assess the solubility and diffusivity of a low molar mass
compound in a polymer.
References 57

References

Allen, L. H., Ramanath, G., Lai, S. L., Ma, Z., Lee, S., Allman, D. D. J., & Fuchs, K. P. (1994). Applied Physics Letters,
64, 417.
Aumnate, C., Rudolph, N., & Saronadi, M. (2019). Polymers, 11, 1456.
Azhdar, B., Yu, W., Reitberger, T., & Gedde, U. W. (2009). Polymer Testing, 28, 661.
Bair, H. E. (1970). Polymer Engineering and Science, 10, 247.
Barton, J. M. (1973). Makromolekulare Chemie, 171, 247.
Bekkedahl, N. (1974). Journal of the Research of the National Bureau of Standards, Part A, 78, 331.
Benzinger, T. H., & Kitzinger, C. (1963). In J. Hardy (Ed.), Temperature – Its measurement and control in science and
industry (Vol. 3, Part 3). New York: Reinhold.
Bertilsson, H., Delin, M., Klason, C., Kubát, M. J., Rychwalski, R. W., & Kubát, J. (1994). Journal of Non-Crystalline
Solids, 172–174, 779.
Billingham, N. C., Bott, D. C., & Manke, A. S. (1981). Application of thermal analysis methods to oxidation and
stabilisation of polymers. In N. Grassie (Ed.), Development in Polymer Degradation (Vol. 3). London: Applied
Science Publishers.
Blankenhorn, K., & H€ ohne, G. W. H. (1991). Thermochimica Acta, 187, 219.
Boersma, S. L. (1955). Journal of the American Ceramic Society, 38, 281.
Boltzmann, L. (1847). Annalen der Physik und Chemie, 7(4), 624.
Bolze, J., Kogan, V., Beckers, D., & Fransen, M. (2018). Review of Scientific Instruments, 89, 085115.
Boyd, R. H. (1979). Polymer Engineering and Science, 19, 1010.
Boyd, R. H. (1980). Dielectric constant and loss. In R. A. Fava (Ed.), Methods of experimental physics (Vol. 16, Part C).
New York: Academic Press.
Boyd, R. H. (1985). Polymer, 26, 323.
Boyd, R. H., & Aylwin, P. A. (1984). Polymer, 25, 340.
Boyd, R. H., & Liu, F. (1997). Chapter 4. In J. P. Runt & J. J. Fitzgerald (Eds.), Dielectric spectroscopy of polymeric
materials. Washington, DC: American Chemical Society.
Boyd, R. H., & Smith, G. D. (2007a). Mechanical relaxation (Chap. 1). In Polymer dynamics and relaxation.
Cambridge: Cambridge University Press.
Boyd, R. H., & Smith, G. D. (2007b). Dielectric relaxation (Chap. 2). In Polymer dynamics and relaxation. Cambridge:
Cambridge University Press.
Brown, M., & Gallagher, P. (2007). Handbook of thermal analysis and calorimetry. Oxford: Elsevier.
Cahn, L., & Schulz, H. (1963). Analytical Chemistry, 35, 1729.
Calvet, E., & Prat, H. (1963). In H. A. Skinner (Ed.), Recent progress in microcalorimetry. London: Pergamon Press.
Catto, D., Howard-Knight, J. P., Kohli, P., & Hobbs, J. K. (2011). Review of Scientific Instruments, 82, 043710.
Clausius, R. (1854). Annalen der Physik und Chemie, 93, 481.
Coburn, J. C., & Boyd, R. H. (1986). Macromolecules, 19, 2238.
Cole, K. S., & Cole, R. H. (1941). Journal of Chemical Physics, 9, 341.
Coumans, W. J., & Heikens, D. (1980). Polymer, 21, 957.
Crank, J. (1986). The mathematics of diffusion (2nd ed.). Oxford: Clarendon Press.
Davidson, D. W., & Cole, R. H. (1950). Journal of Chemical Physics, 18, 1417.
Davidson, T., & Wunderlich, B. (1969). Journal of Polymer Science, Part A-2: Polymer Physics, 7, 377.
Delin, M., Rychwalski, R. W., Kubát, M., & Kubát, J. (1995). Rheologica. Acta, 34, 182.
Demus, D., & Richter, L. (1978). Textures of liquid crystals. VEB Deutscher Verlag für Grundstoffindustrie: Leipzig.
De Santis, F., Adamovsky, S., Titomanlio, G., & Schick, C. (2007). Macromolecules, 40, 9026.
Efremov, M. Y., Olson, E. A., Zhang, M., Schiettekatte, F., Zhang, Z., & Allen, L. H. (2004a). Review of Scientific
Instrument, 75, 179.
Efremov, M. Y., Olson, E. A., Zhang, M., Zhang, Z., & Allen, L. H. (2004b). Macromolecules, 37, 4607.
Engberg, K., Str€ omberg, O., Martinsson, J., & Gedde, U. W. (1994). Polymer Engineering and Science, 34, 1336.
Feldman, Y. D., Zuer, Y. F., Polygalov, E. A., & Fedotov, V. D. (1992). Colloid and Polymer Science, 270, 768.
Ferry, J. D. (1980). Viscoelastic properties of polymers (3rd ed.). New York: John Wiley & Sons.
Fulcher, G. S. (1925a). Journal of the American Chemical Society, 8, 339.
Fulcher, G. S. (1925b). Journal of the American Chemical Society, 8, 789.
Gabbott, P. (2008). Principles and applications of thermal analysis. Oxford: Blackwell Publishing.
Gedde, U. W., Terselius, B., & Jansson, J.-F. (1981a). Polymer Engineering and Science, 21, 174.
Gedde, U. W., Eklund, S., & Jansson, J.-F. (1981b). Polymer Bulletin, 4, 717.
Gedde, U. W., & Jansson, J.-F. (1983a). Polymer, 24, 1521.
Gedde, U. W., & Jansson, J.-F. (1983b). Polymer Bulletin, 9, 90.
Gedde, U. W., Liu, F., Hult, A., Sahlén, F., & Boyd, R. H. (1994). Polymer, 35, 2056.
58 1 Thermal Analysis of Polymers

Gedde, U. W., & Hedenqvist, M. S. (2019a). The glassy amorphous state (chap. 5). In Fundamental polymer science.
Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Morphology of semicrystalline polymers (Chap. 7). In Fundamental
Polymer Science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019c). Crystallization kinetics (Chap. 8). In Fundamental Polymer Science.
Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019d). Introduction to polymer science (chap. 1). In Fundamental Polymer
Science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019e). Solution to problems given in exercises (Chap. 11). In Fundamental
Polymer Science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019f). The molten state (Chap. 6). In Fundamental Polymer Science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W. (2020). Essential classical thermodynamics. Cham: Springer Nature Switzerland AG.
Geil, P. H., Anderson, F. R., Wunderlich, B., & Arakawa, T. (1964). Journal of Polymer Science, Part A: General
Papers, 2, 3707.
Gillen, K. T. (1978). Journal of Applied Polymer Science, 22, 1291.
Graewert, M. A., & Svergun, D. I. (2013). Current Opinions in Structural Biology, 23, 748.
Gray, A. P. (1970). Thermochimica Acta, 1, 563.
Hager, N. E. (1964). Review of Scientific Instruments, 35, 618.
Hannay, J. B. (1877). Journal of the Chemical Society, 32, 381.
Haward, R. N., & Young, R. J. (Eds.). (1997). The physics of glassy polymers (2nd ed.). Berlin: Springer.
Havriliak, S., & Negami, S. (1967). Polymer, 8, 161.
Hellmuth, E., & Wunderlich, B. (1965). Journal of Applied Physics, 36, 3039.
Hemminger, W. (1994). Calorimetric methods (Chap. 2). In V. B. F. Mathot (Ed.), Calorimetry and thermal analysis of
polymers. Munich: Hanser.
Hemminger, W. (1998). Recommendations of the ICTAC Nomenclature Committttee. ICTAC News, December,
pp. 106–122.
Hemminger, W., & Sarge, S. W. (1998). Chapter 1. In M. E. Brown (Ed.), Handbook of thermal analysis and
calorimetry (Vol. 1). Amsterdam: Elsevier.
Hoffman, J. D., Frolen, L. J., Ross, G. S., & Lauritzen, J. I., Jr. (1975). Journal of the Research of the National Bureau of
Standards, Part A, 6, 671.
Honda, K. (1915). Science Reports of Tohoku University, 4, 97.
Hutchinson, J. M. (1995). Progress in Polymer Science, 20, 703.
Hutchinson, J. M. (1996). Paper presented at Mettler-Toledo User Meeting, Stockholm.
H€ohne, G. W. H. (1994). Fundamentals of differential scanning calorimetry and differential thermal analysis (Chap. 3).
In V. B. F. Mathot (Ed.), Calorimetry and thermal analysis of polymers. Munich: Hanser.
Illers, K. H. (1973). Kolloid-Zeitschrift & Zeitschrift für Polymere, 251, 394.
Karlsson, K., Assargren, C., & Gedde, U. W. (1990). Polymer Testing, 9, 421.
Karlsson, K., Smith, G. D., & Gedde, U. W. (1992). Polymer Engineering and Science, 32, 649.
Kohlrausch, R. (1854). Annalen der Physik und Chemie, 167(2), 179.
James, J. (2017). Thermomechanical analysis and its applications (chap. 7). In S. Thomas, R. Thomas, A. Zachriah, &
R. Kumar (Eds.), Thermal and rheological measurement techniques for nanomaterial characterization (Vol. 3).
Amsterdam/Oxford/Cambridge: Elsevier.
Kane, W. J., Young, R. J., Ryan, A. J., Bras, W., Derbyshire, G. E., & Mant, G. R. (1994). Polymer, 35, 1352.
Kremer, F., & Sch€ onhals, A. (Eds.). (2003). Broad-band dielectric spectroscopy. Berlin/Heidelberg: Springer.
Lavoisier, A. (1789). Elements of chemistry, Part 3, Volume 3, Paris.
Le Chatelier, H. (1887). Bulletin de la Societe Francaise de Mineralogie et de Cristallographie, 10, 204.
Liberti, F. N., & Wunderlich, B. (1968). Journal of Polymer Science, Part B: Polymer Physics, 6, 833.
Loganathan, S., Valapa, R. B., Mishra, R., Pugazhenthi, G., & Thomas, S. (2017). Thermogravimetric analysis for
characterization of nanomaterials (Chap. 4). In S. Thomas, R. Thomas, A. Zachriah, & R. Kumar (Eds.), Thermal
and rheological measurement techniques for nanomaterial characterization (Vol. 3). Amsterdam, Oxford and
Cambridge: Elsevier.
Lorentz, H. A. (1881). Annalen der Physik, 248, 127.
Lorenz, L. (1870). Det Konglige Danske Videnskabernes Selskabs Skrifter, 8, 205.
Lorenzo, A. T., Arnal, M. L., Albuerne, J., & Müller, A. J. (2007). Polymer Testing, 26, 222.
Malas, A. (2016). Rubber nanocomposites with graphene as the nanofiller (Chap. 6). In S. Thomas, S. and H. J. Maria
(Eds.) Progress in Rubber Nanocomposites, Woodhead Publishing Series in Composite Science and Engineering,
No 75. Cambridge: Woodhead Publishing.
Malmstr€ om, E., Liu, F., Boyd, R. H., Hult, A., & Gedde, U. W. (1994). Polymer Bulletin, 32, 679.
References 59

Marcus, S. M., & Blaine, R. L. (1994). Thermochimica Acta, 243, 231.

Mathot, V. B. F. (1994). Thermal characterization of states of matter (Chap. 5). In V. B. F. Mathot (Ed.), Calorimetry
and thermal analysis of polymers. Munich/Vienna/New York: Hanser.
Mathot, V. B. F., Scherrenberg, R. L., Pijpers, M. F. J., & Bras, W. (1996). Journal of Thermal Analysis, 46, 681.
McCrum, N. G., Read, B. E., & Williams, G. (1967). Anelastic and dielectric effects in polymeric solids. New York:
John Wiley & Sons.
Menard, K. P., & Menard, N. (2017). Dynamic mechanical analysis. In M. RA (Ed.), Encyclopedia of analytical
chemistry: Applications, theory and instrumentation. New York: John Wiley & Sons.
Moraglio, G., Polizzotti, G., & Danussi, F. (1965). European Polymer Journal, 1, 183.
Moynihan, C. T., Easteal, A. J., DeBolt, M. A., & Tucker, J. (1976). Journal of the American Ceramic Society, 59, 12.
Mullin, N., & Hobbs, J. K. (2011). Physics Review Letters, 107, 197801.
Núnez, E., Ferrandi, C., Malmstr€om, E., Claesson, H., & Gedde, U. W. (2004). Journal of Macromolecular Science.
Part B: Physics, 43, 1143.
Núnez, E., Vancso, G. J., & Gedde, U. W. (2008). Journal of Macromolecular Science. Part B: Physics, 47, 589.
Pethrick, R. A. (2013). Non-destructive evaluation (NDE) of composites: Dielectric techniques for testing partially or
non-conducting composite materials (Chap. 6). In V. M. Karbhart (Ed.), Non-Destructive Evaluation (NDE) of
polymeric matrix composites. Cambridge: Woodhead Publishing.
Pijpers, M. F. J., Mathot, V. B. F., Goderis, B., Scherrenberg, R., & van der Vegte, E. (2002). Macromolecules, 35,
3601.
Portale, G., Cavallo, D., Alfonso, G. C., Hermida-Merino, D., van Drongelen, M., Baltzano, L., Peters, G. W. M.,
Goussens, J. G. P., & Bras, W. (2013). Journal of Applied Crystallography, 46, 1681.
Prud’homme, R. E. (2016). Progress in Polymer Science, 54–55, 214.
Ramos, J., Vega, J. F., & Martinez-Salazar, J. (2018). European Polymer Journal, 99, 298.
Ramsay, W. (1877). Journal of the Chemical Society, 32, 395.
Ramsey, W. (1918). The life and letters of Joseph Black. London: Constable & Company.
Rego Lopez, J. M., Conde Brana, M. T., Terselius, B., & Gedde, U. W. (1988). Polymer, 29, 1045.
Richardson, M. J. (1976). Plastics and Rubber: Materials and Applications, 1, 162.
Schawe, J. E. K. (1998). Journal of Polymer Science, Part B: Polymer Physics, 36, 2165.
Schawe, J. E. K., & Pagatscher, S. (2016). Material characterization by fast scanning calorimetry: Practice and
applications (Chap. 1). In C. Schick & V. B. F. Mathot (Eds.), Fast scanning calorimetry. Cham: Springer
International Publishing.
Schick, C., & Mathot, V. (Eds.). (2016). Fast scanning calorimetry. Cham: Springer International Publishing.
Silvestre, C., Cimmino, S., Duraccio, D., & Schick, C. (2007). Macromolecular Rapid Communication, 28, 875.
Smith, G. D., Karlsson, K., & Gedde, U. W. (1992). Polymer Engineering and Science, 32, 658.
Smith, T. L. (1959). Transactions of the Society of Rheology, 3, 113.
Stone, R. L. (1951). Bulletins of the Ohio State University, Engineering Experimental Station Station, 146, 78.
Struik, L. C. E. (1978). Physical aging in amorphous polymers and other materials, Amsterdam/Oxford/New York:
Elsevier.
Suzuki, H., Grebowicz, J., & Wunderlich, B. (1985). British Polymer Journal, 17, 1.
Suurkuusk, J., & Wads€o, I. (1982). Chimica Scripta, 20, 155.
TA Instruments. (1996). Modulated DSC Compendium. Basic Theory & Experimental Considerations.
Tammann, G., & Hesse, G. (1926). Zeitschrift für Anorganische und Allgemeine Chemie, 156, 245.
Tence-Girault, S., Lebreton, S., Bunau, O., Dang, P., & Bargain, F. (2019). Crystals, 9, 271.
Toda, A., Androsch, R., & Schick, C. (2016). Polymer, 91, 239.
Ungar, G. (1990). American Chemical Society: Polymer Preprints, 31, 108.
Wagner, M. (2017). Thermal analysis in practice: Fundamental aspects (p. 224). Cincinnati: Hanser Gardner
Publications, Chapter 12.
Vancso, J. G., & Sch€onherr, H. (2010). Scanning force microscopy (Springer laboratory series). Berlin: Springer
Nature.
Watson, E. S., O’Neill, M. J., Justin, J., & Brenner, N. (1964). Analytical Chemistry, 7, 151.
Wiberg, G., & Gedde, U. W. (1998). Thermochimica Acta, 319, 123.
Viebke, J., Hedenqvist, M., & Gedde, U. W. (1996). Polymer Engineering and Science, 36, 2896.
Viebke, J., & Gedde, U. W. (1998). Polymer Engineering and Science, 38, 1244.
Wiedemann, H. G. (1964). Chemie Ingenieur Technik, 36, 1105.
Williams, G., & Watts, D. C. (1970). Transactions of the Faraday Society, 66, 80.
Williams, M. L., Landel, R. F., & Ferry, J. D. (1955). Journal of the American Chemical Society, 77, 3701.
Vogel, H. (1921). Physikalische Zeitschrift, 22, 645.
Wunderlich, B. (1964). Polymer, 5, 611.
Wunderlich, B. (1980). Macromolecular physics, volume 3. Crystal melting. New York/London: Academic Press.
60 1 Thermal Analysis of Polymers

Wunderlich, B. (1995). ATHAS Data Bank. Knoxville: University of Tennessee. Available at http://funnelweb.utcc.utk.
edu/~athas/databank/intro.htlm
Wunderlich, B. (1990). Thermal analysis. Boston: Academic Press.
Wunderlich, B., & Arakawa, T. (1964). Journal of Polymer Science Part A: General Papers, 2, 3697.
Wunderlich, B., & Baur, H. (1970). Advances in Polymer Science, 7, 151.
Zoller, P. (1990). Dilatometry. In J. I. Kroschwitz (Ed.), Polymers: Polymer characterization and analysis (Encyclo-
pedia reprint series). New York: John Wiley & Sons.
Zoller, P., Bolli, P., Pahud, V., & Ackermann, H. (1976). Reviews of Scientific Instruments, 47, 948.
om, K. (1895). Kongliga Svenska Vetenskaps Akademiens Handlingar, 643.
Ångstr€
Chapter 2
Microscopy of Polymers

2.1 Introduction

Microscope (New Latin: microscopium) means ‘an instrument for viewing what is small’, and
microscopy is the name given to a group of experimental methods which make possible the magnifi-
cation of morphological structures to make details visible. Initially, it was confined to optical or light
microscopy (synonymous terms). Optical microscopy soon developed into several different
techniques. Modern optical microscopes are capable of performing a range of different optical
microscopy methods. Later developments led to electron microscopy and atomic force microscopy,
both of which are in fact families, each comprising a range of different techniques. Let us begin with a
historical background describing the development of the different microscopy techniques and also
telling about the inventors and when it happened.
The Dutch opticians Hans and Zacharias Janssen discovered the optical microscope in 1590
(Turner 2012). The microscope of father (Hans) and son (Zacharias) Janssen consisted of a 45 cm
long tube with two lenses (objective and eyepiece). This optical microscope was a force in the rapid
development of both physical and biological sciences, and the optical microscope led to a shift from
metaphysics to a more mechanistic view of life and nature. In the seventeenth century, microscopes
were used to study details of the human body, animals and plants. Marcello Malpighi discovered,
during this time, the capillary blood vessels from studies of frog lungs, and he discovered the blood
corpuscles using optical microscopy. His observations were confirmed by the work of the Dutch self-
trained scientist Antonie van Leeuwenhoek (Baumel and Berger 1973). Robert Hooke, another
pioneer microscopist, revealed cells in plants and tissues (Baumel and Berger 1973). Some
philosophers such as John Locke meant that microscopy is meaningless and potentially dangerous.
He argued that the microscopist performed an act that contradicted God’s purpose with man by seeing
things that are not meant to be seen (Locke 1824). These ideas and the poor optics of the microscopes
during the eighteenth century made people rather suspicious of microscopy in general. Respect for
microscopy increased however during the nineteenth century with the important discovery of bacteria
by Louis Pasteur (Feinstein 2008). The instrumental development was strong during the nineteenth
and early twentieth centuries. Ernst Abbe gave optical microscopy a theoretical foundation,
explaining the limitation in resolution (minimum separation of two adjacent points that can be seen
as two different items) as being due to diffraction (Abbe 1873). Abbe (1874) also constructed the first
objective lens with a correction for chromatic aberration. Ernst Ruska constructed the first transmis-
sion electron microscope (TEM) in 1931 (Knoll and Ruska 1932; Ruska 1980). Ruska received the
Nobel Prize in Physics for this accomplishment in 1986, the resolution of his instrument being several
tens of nm. This development was preceded by the realization that particles, like electrons, have an
associated wave nature (Louis de Broglie 1925) and by the discovery by Hans Busch (1926) that

# Springer Nature Switzerland AG 2021 61

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_2
62 2 Microscopy of Polymers

axially symmetric magnetic fields can focus electrons (Ruska 1980). Several commercial instruments
with a resolution claimed to be 0.2 nm were developed in Europe, America and Japan on the basis of
the principles outlined by Ruska.
The principles of the scanning electron microscope (SEM) date back to Max Knoll (1935) and
Manfred von Ardenne (1938a, b), but the first SEM instrument was not built until the 1950s, an
accomplishment due to the pioneering work of Charles Oatley and co-workers in Cambridge, UK
(McMullan 1952). The scanning transmission electron microscope was further developed and
marketed in the beginning of the 1970s.
The scanning probe microscopy (SPM) techniques with the potential to resolve individual atoms
are an even later development. Gerd Binnig, Heinrich Rohrer and co-workers invented the scanning
tunnelling microscope (STM) in 1982 (Binnig et al. 1982a, b), and the same research group invented a
closely related instrument, which is much more important for polymer research, the atomic force
microscope (AFM) a few years later (Binnig et al. 1986). Binnig and Rohrer received the Nobel Prize
in Physics in 1986 (together with Ruska). AFM is nowadays a standard instrument in many research
laboratories. The SPM technique uses a fine-tip probe that scans over a surface, revealing its surface
topography (STM, AFM), mechanical properties (scanning force microscope, SFM), thermal
properties (scanning thermal profiler, STP) and magnetic properties (magnetic force microscope,
MFM).
The morphology of polymer blends, block copolymers, semicrystalline polymers, liquid crystal-
line polymers and composites can be directly observed by microscopy, and the structure can in most
cases be assessed without the need for any sophisticated model. Small-angle X-ray scattering
provides information about the long period (i.e. the thickness of the crystalline and amorphous layers)
in a semicrystalline polymer (cf. Gedde and Hedenqvist 2019a and Chap. 3), but structural informa-
tion cannot be gained from the scattering pattern without the use of a model. Transmission electron
microscopy of samples treated with etchants or with staining compounds makes direct observation of
the crystal lamellae possible (Fig. 2.1).
What must be questioned is whether the image obtained by microscopy is free from distortions and
artefacts. It is, for instance, possible that radiation or mechanical stress affects the observed structure.
Radiation effects are rare in optical microscopy, but they are common in electron microscopy. Some
polymers are very sensitive to the electron beam and rapidly degrade leaving a hole in the specimen,
whereas other polymers like polyethylene become cross-linked. More moderate damage,
e.g. shrinkage of the specimen, may also occur.
The purpose of the sample preparation and subsequent irradiation is to obtain a contrast-rich
representation of the ‘true’ structure, and if contrast is not naturally present in the specimen, it may be

Fig. 2.1 (a) Transmission electron micrograph of chlorosulphonated polyethylene. (b) Small-angle X-ray scattering
pattern of the same polyethylene sample
2.1 Introduction 63

Table 2.1 Characteristic features of optical microscopy (OM), scanning electron microscopy (SEM), transmission
electron microscopy (TEM) and atomic force microscopy (AFM)
OM SEM TEM AFM
Size of studied objects 0.2–200 μm 0.004–4000 μm 0.0002–20 μm 0.0002–10 μm
Depth of field Small Large Large Large
Objects Surface or bulk Surface Bulk Surface
Specimen environment Ambient or fluid High vacuum High vacuum Ambient, high vacuum or fluid
Radiation damage None Some Severe None
Specimen preparation Easy Easy Difficult Easy
Can detect chain orientation Yes No Yes No
Chemical analysis No Yes Yes No
Source: Sawyer and Grubb (1996)

created by staining, i.e. by adding a substance with a characteristic colour or high (electron) density
selectively to a specific phase. Contrast may also be achieved by selective etching leaving a
topography indicative of the phase structure, although it is possible that the preparation method
used may produce artificial structures.
A third problem originates from the fact that only a small region of the whole object is studied. Is
the region viewed representative of the object? The amount of information obtained from an
observation is substantial, and the microscopist has to select a few features which are typical of the
specimen. It is however possible that non-typical features may be presented.
The result of a microscopy study is basically a two-dimensional picture of the true three-
dimensional morphology. A frequently asked question about a polymer blend is whether a polymer
constituent forms a continuous phase, since it is very likely that a cut through the structure with a
continuous phase will give a phase that appears to be dispersed (i.e. interrupted). A good analogy is a
tree, which is definitely a continuous structure. What do you see if you make a cut through the tree
with all its branches? Semicrystalline polymers are treated with staining chemicals to make crystals
visible in the electron microscope, but it is only possible to see the crystals when their fold surfaces
are parallel to the electron beam. Also, in the case, a two-dimensional cut with the aforementioned
preconditions is revealed. Are the crystal lamellae continuous or not? This is a very difficult problem
to solve based on observations of electron micrographs of a stained section.
Microscopy of primary importance for polymer science may be divided into four main groups:
optical microscopy (OM), scanning electron microscopy (SEM), transmission electron microscopy
(TEM) and atomic force microscopy (AFM), each of which should be considered as a family
consisting of several individual techniques. Several new techniques have also recently been devel-
oped: infrared microscopy and Raman microscopy provide the infrared (Raman) spectra of fine
morphological details; UV microscopy enables the assessment of local variations in UV absorption
and has been used to assess the content of UV-absorbing additives with reference to morphology; and
acoustic microscopy renders information about local variations of the mechanical properties in a
sample and may thus reveal local variations in chain orientation and composition. These special
techniques are not however further discussed in this text.
It is possible to use both optical and electron microscopes for scattering/diffraction. The method of
observing the interference pattern at the back focal plane of the objective in the optical microscope by
using an extra lens (Bertrand lens) is referred to as conoscopy, and this provides information about the
optical anisotropy (the uniaxial or biaxial symmetry; the direction of the principal axes of the
refractive index). McCrone, McCrone and Delly (1980) give details about conoscopy. It is also
possible by closing the field iris and the aperture iris to obtain a small-angle light scattering pattern
(cf. Gedde and Hedenqvist 2019a). Electron diffraction using electron microscope is a classical
method. The scattering methods are presented in Chap. 3.
Table 2.1 presents a summary of the general features of OM, SEM, TEM and AFM.
64 2 Microscopy of Polymers

2.2 Optical Microscopy

2.2.1 Fundamentals

The fundamental principles of optical microscopy are here briefly recapitulated. Detailed
presentations of the fundamentals of optical microscopy are available in textbooks on optics and in
specialized volumes on microscopy (Sawyer et al. 2008; Hemsley 1989a, b; and McCrone et al.
1980). The magnification of the modern type of optical microscope (compound microscope) is
given by two lens systems referred to as the objective and the eyepiece (Fig. 2.2). Both lenses are
positive, i.e. they focus a beam of parallel rays to a focal point. The objective generates a magnified
real image of the specimen, and the eyepiece further magnifies the real image, and a magnified real
image is formed on the retina of the eye. The maximum magnification achieved by OM is about
2000x.
A modern research optical microscope can be modified in various ways. The surface topography of
a specimen is studied in the reflected light mode, whereas the bulk structure is studied with the light
transmitted through the specimen. Bright-field microscopy is microscopy in which the specimen is
illuminated essentially parallel to the main optical axis in either the reflected or the transmission
modes (Fig. 2.3), whereas in dark-field microscopy, the illumination is at a very large angle to the
main optical axis. Both techniques can be applied in either the reflected or the transmission mode
(Fig. 2.3). The term ‘dark field’ originates from appearance of the image, where the phase boundaries
appear light against a black background, which is the inverse of the appearance of the bright-field
image.
The eye can detect variations in both light intensity and colour. Light with different states of
polarization cannot be distinguished with the unaided eye, but there are several techniques by which a
variation in the optical path length (OP) can be converted to a variation in light intensity (Fig. 2.4).
The variation in the optical path length can be due to a variation in refractive index, to a variation in

Fig. 2.2 Magnification of

an object in the optical
microscope obtained by
objective, eyepiece and
human eye

Fig. 2.3 Bright-field and

dark-field microscopy
2.2 Optical Microscopy 65

Fig. 2.4 The principle

behind the techniques that
reveal variations in optical
path. A variation in optical
path length (OP) between
the two beams
(OP1 6¼ OP2) leads to a
lowering of the intensity of
the outgoing light
(Iout < Iin)

Fig. 2.5 Definition of

numerical aperture (NA) of
an objective lens

specimen thickness and/or to a variation in surface topography in the case of reflected light. The
techniques using this principle are polarized microscopy, phase-contrast microscopy and differential
interference-contrast microscopy. These techniques are presented in Sect. 2.2.2.
The resolution is defined as the least distance between two ‘points’ that can be recognized as two
separate items. Lord Rayleigh (Strutt 1879) showed that two points can be resolved if the light
intensity at a point intermediate between the points is 85% or less of the maximum intensity at the
centres of the points. The resolution (d ) that is dependent on diffraction and is given by:
λ
d¼ ð2:1Þ
NAobj þ NAcond

where λ is the wavelength of the light and NAobj and NAcond are the numerical apertures of the
objective and the condenser, defined as:
NA ¼ n sin α ð2:2Þ

where n is the refractive index and α is half the acceptance angle (Fig. 2.5). Equation (2.1) is valid
only if NAobj  NAcond. High resolution is thus obtained when components with a high numerical
aperture are used. The resolution increases with decreasing wavelength of the light.
All optical components are imperfect. Lenses suffer from various aberrations: chromatic (light of
different wavelengths is refracted differently), spherical, coma, astigmatism, field curvature (the sharp
image falls on a curved surface) and distortion. Modern lenses correct for some of these aberrations:
achromats and apochromats correct for chromatic aberration. Lenses corrected for field curvature are
denoted ‘plano’, but the aberrations reduce the resolution of the microscope, and the value obtained
from Eq. (2.1) should be considered to be the best possible resolution of the microscope.
The components in a transmission optical microscope are schematically shown in Fig. 2.6. The
light source in a modern microscope is of the filament type, the commonest type being of tungsten and
producing a continuous spectrum from 300 nm to 1500 nm (wavelength). Colour correction filters are
66 2 Microscopy of Polymers

Fig. 2.6 Schematic

representation of the
compound microscope,
including the conjugate
planes

used to obtain a narrower spectrum. The intensity of the light should not be adjusted by changing the
voltage but rather by inserting a neutral filter. The collector lens is placed so that the focal point of the
lens is at the filament of the bulb. The field iris adjusts the size of the viewed field, and the aperture
iris adjusts the part of the lamp filament that illuminates the specimen.
The condenser concentrates the light on the specimen, providing a uniform illumination, and
matches the aperture of the illuminating cone with the numerical aperture of the objective. A common
bright-field condenser consists of a top lens that can be swung into the light beam to work with higher
magnification objectives, whereas the lower condenser lens is used alone with lower magnification
objectives. The condenser may provide special illumination. Dark-field condensers illuminate the
specimen at a large angle to the main optical axis, while special condensers are used for phase-
contrast and differential interference-contrast microscopy. More details of these techniques are
presented in Sect. 2.2.2. The condensers are also characterized according to their numerical aperture
and optical correction.
The objective creates most of the magnification, and it also controls the aperture, which in turn
controls the resolution. The maximum objective magnification is about 100x. The objective is
characterized by magnification, numerical aperture and optical correction (achromate, fluorite,
apochromate, plano, etc.). Objectives marked with ‘oel’ are oil immersion objectives, and they
should always be used with immersion oil between the cover glass and the objective lens. The
working distance, i.e. the distance between the objective lens and the specimen, is very short and
usually less than 1 mm for high magnification objectives. Special objectives have been built for
hot-stage work with a magnification of 25–35x and with a long working distance (ca. 10 mm). Special
objectives with stress-free glass often denoted ‘POL’ are used for polarized microscopy, phase-
contrast microscopy and differential interference-contrast microscopy. The eyepiece is used to
magnify the primary image further and to enable scales and pointers to be introduced into the primary
image plane.
2.2 Optical Microscopy 67

The conjugate planes are defined as the planes in the optical system that are equivalent, so that an
object placed in one of the planes will appear as a sharp image in the subsequent planes of the series.
The compound microscope contains two separate series of conjugate planes (Fig. 2.6):
• I consists of the lamp filament, the front focal plane of the condenser (which is located at the
aperture iris), the back focal plane of the objective and the exit pupil of the eye.
• II consists of the lamp iris diaphragm (field iris), the specimen plane, the primary image plane and
the retina of the eye.
Series II thus includes the specimen plane. The field iris controls the diameter of the area that is
viewed. Pointers or scales must be inserted in the primary image plane. Series I is related to the
filament. An image of the filament is created at the back focal plane of the objective. An important
principle behind uniform and intense illumination of the specimen is the K€ ohler illumination, named
after August K€ohler (K€ohler 1893). It must be possible to focus the illuminator, with an adjustable
field iris, and the lamp filament position must be adjustable in all directions. The correct illumination
is obtained by adjusting the collector lens and the condenser. Detailed instructions are provided by
McCrone et al. (1980) and by Hemsley (1989a).

2.2.2 Polarized Microscopy and Related Techniques

Polarized microscopy requires a microscope equipped with two crossed polarizers. A polarizer is
basically a component (filter or prism) that permits transmission only of light with the electrical field
vector pointing along a line. The out-coming light is said to be linearly polarized. The first, the
polarizer, is located in front of the specimen stage, and the second, the analyzer, is located at the back
of the objective. The light entering the specimen is linearly polarized, and, if the sample is optically
isotropic, the light passes through it without any change of polarization, and the crossed analyzer then
extinguishes the light. If, however, the sample is birefringent, the light is elliptically polarized after
the passage through sample. The analyzer reduces the intensity of the light, but an appreciable
fraction of the incoming light is transmitted.
It is customary to define a refractive index ellipsoid for a given specimen (or region within a
specimen), which becomes an ellipse in the plane. The lengths of the principal axes of the refractive
index ellipse are proportional to the maximum and minimum refractive indices in the plane. The
polarization of the light remains unchanged as long as the refractive index ellipse has its principal
axes parallel to the polarizer/analyzer pair. All other orientations of the ellipse lead to a situation in
which the light is partially transmitted through the analyzer (Fig. 2.7).
These statements may be explained by the following discussion, where an electrical field vector
E represents the polarization of the light. If the incoming light is linearly polarized, E may be divided

Fig. 2.7 The influence on

the light transmitted
through the analyzer of the
orientation of the refractive
index ellipses with respect
to the polarizer (P) and
analyzer (A)
68 2 Microscopy of Polymers

into two components, E1 and E2, where ‘1’ denotes the direction of the maximum refractive index and
‘2’ the direction of the minimum refractive index. The two components E1 and E2 are initially in
phase, but the difference in refractive index causes a retardation of the ‘1’ component with respect to
the ‘2’ component, and the outgoing light is then a combination of the two linearly polarized
components that are out of phase, i.e. an elliptically polarized light. If the phase difference is a
multiple of the wavelength, the two components combine to give a linearly polarized light. If ‘1’ is
parallel to either the polarizer or the analyzer, the partitioning of E results in only E1 or E2, the other
being perpendicular and hence zero and the outgoing light is linearly polarized and blocked by the
analyzer (Fig. 2.7).
The λ-plate and other compensators are accessories commonly used in polarized microscopy. The
λ-plate retards the ‘slow’ light by one wavelength, i.e. 550 nm, with respect to the ‘fast’ light. In older
literature, the λ-plate is called the gypsum plate. Another fixed compensator of common use is the λ/4-
plate (mica plate), which shifts the phase of the slow light by λ/4 with respect to the fast light. The
mica plate produces circular polarized light from plane polarized light if the direction of vibration of
the incident light is at an angle of 45 to the principal axes of the mica plate. A detailed presentation of
compensators and their use for the assessment of in-plane birefringence is described by Gedde and
Hedenqvist (2019b). A recommended text about compensators is that of Harding (1986). It is
impossible to distinguish the two differently birefringent objects shown in Fig. 2.8 in a conventional
polarized light microscope by inserting a λ-plate at an angle of 45 to the polarizer/analyzer pair
(Fig. 2.8c). The reason for the presence of colours (presented in the Michel-Levy birefringence chart,
named after the French geologist Auguste Michel-Levy) is explained by Gedde and Hedenqvist
(2019b). When the phase difference between the rays of the two orthogonal directions is greater than
λ (the high refractive index is parallel to the slow direction of the λ-plate), the colour is blue. If the
direction of the high refractive index is transverse to that of the slow direction of the λ-plate, the phase
difference is less than λ and the colour is yellow. Superstructures present in semicrystalline polymers
are conveniently studied with polarized light microscopy, and the optical features of spherulites are
discussed by Gedde and Hedenqvist (2019a).
Figure 2.9 shows negative spherulites (nt > nr, where nt is the refractive index along the tangential
direction and nr is the refractive index along the spherulite radius) in a semicrystalline polymer. The
image obtained by conventional polarized microscopy cannot distinguish between negative and
positive spherulites; the latter show nt < nr. By inserting a λ-plate at a 45 angle with respect to the
polarizer/analyzer pair, different colours appear in the quadrants of the spherulite: a yellow colour
appears in quadrants 1 and 3 and a blue colour in quadrants 2 and 4. A positive spherulite shows the
reverse scheme, i.e. quadrants 1 and 3 are blue and quadrants 2 and 4 are yellow.
Phase-contrast microscopy converts differences in refractive index to variations in light intensity.
Light that is scattered by the specimen travels in a direction different from that of the undeflected
light, and it is also shifted in phase (retarded) by λ/4. The scattered light is phase-shifted by another λ/
4, and destructive interference takes place when the undiffracted and diffracted beams recombine to
form the image. Contrast is obtained when the relative intensities of the scattered and the unscattered
(transmitted) light vary from one location in the sample to another.
Figure 2.10 shows schematically the Zernike phase-contrast microscope (named after Fritz
Zernike, Dutch physicist; Zernike 1942a, b, 1955), which uses a special condenser with an annular
stop. The condenser blocks all light except for that going through the ring. A matching phase ring is
placed in the back focal plane of the objective. The unscattered light alone travels through the
absorbent phase ring, where the intensity is reduced by as much as 75%. A dark-field image would be
obtained if the phase ring were to absorb all the undiffracted light. The scattered light travels through
other parts of the phase plate, and it is retarded by a further λ/4 with respect to the direct beam. The
combination of scattered and unscattered light leads to a change in light intensity depending on the
variation in refractive index. The phase-contrast image shows typically bright halos around the fine
structure. Further information about phase-contrast microscopy is given by Hemsley (1989b).
2.2 Optical Microscopy 69

Fig. 2.8 (a) Images of birefringent objects (indicatrices are displayed) in the polarized light microscope with the
orientations of the polarizer (P) and of the analyzer (A) shown. (b) A negative spherulite (nt > nr) and a positive
spherulite (nt < nr) as displayed in the polarized light microscope. Note the Maltese cross pattern. (c) Left: Michel-Levy
colour chart showing the first order colours. Bottom: Different colours representing two identical objects with different
orientations of the indicatrices in crossed polarizers with an inserted λ-plate; note the orientation of the slow direction of
the λ-plate

Differential interference-contrast (DIC) microscopy is suitable for studying specimens with

optical path length difference less than λ/2, but it works best at values close to λ/10. The DIC
microscope splits the incoming beam into two components; the Nomarski DIC microscope (Nomarski
1952, 1955) has crossed polarizers and two Wollaston prisms. The first Wollaston prism located in
front of the specimen produces two polarized rays, which are displaced with a mutually perpendicular
orientation of their vibrations. The second Wollaston prism behind the objective recombines the two
components. The analyzer makes the vibration direction parallel so that the light rays can interfere at
the primary image plane, and the close spacing of the displaced beams causes the image to appear in
relief. A region of constant refractive index and constant specimen thickness appears without any
contrast. However, any sudden change in refractive index gives a strong contrast. DIC microscopy is
mostly used in the reflective mode. Recommended further reading about DIC and modulation contrast
is given by Hoffman (1989).
70 2 Microscopy of Polymers

Fig. 2.9 (a) Polarized micrograph of a sample containing negative spherulites. The polarizer–analyzer pairs were
oriented horizontally and vertically. (b) The sample in the same microscope but with the addition of a λ-plate. Note the
orientation of the slow direction of the λ-plate. The blue colour indicates that the high refractive index is along the slow
direction of the λ-plate

Fig. 2.10 Phase-contrast

microscope according to
Zernike. Drawn after
McCrone et al. (1980)

2.3 Electron Microscopy

Very fine details of the microstructure cannot be resolved by optical microscopy, but an electron
microscope can resolve details smaller than 1 nm. The interested reader will find very useful
information about electron microscopy in Vadimsky (1980), Vaughan (1993) and Sawyer et al.
2.3 Electron Microscopy 71

Fig. 2.11 Schematic description of conventional TEM and SEM

(2008). The electro-optical imaging can be achieved in different ways. Lenses consisting of magnetic
fields control the electron beam. Rotational symmetric electromagnets focus the electron beam in the
same way as convex lenses focus the beam in optical microscopes. The transmission electron
microscope is built according to the same principle as the optical microscope with a condenser
lens, an objective lens and a projector lens (analogue of the eyepiece) (Fig. 2.11). A magnified image
is obtained on a fluorescent screen or on a photographic film. An electron beam is commonly emitted
by a tungsten hairpin filament or a lanthanum hexaboride (LaB6) filament heated with a low voltage
source. The potential of the filament is highly negative, and the electrons are accelerated towards an
anode held at a small positive potential. A typical voltage used in transmission electron microscopes
(TEM) is 100 kV. The acceleration voltage used in a scanning electron microscope (SEM) is lower,
typically about 15 kV.
The resolution of the electron microscope depends on the acceleration voltage. The wavelength of
the accelerated electrons is, according to de Broglie (1925), equal to:
h
λ¼ ð2:3Þ
mv

where h is Planck’s constant, m is the mass of the electron and v is the velocity of the electron. For an
electron with charge e subjected to a voltage U, the following expression holds, provided that the
relativistic effects are negligible:

mv2
eU ¼ ð2:4Þ
2

Insertion of Eq. (2.4) in Eq. (2.3) yields:

rffiffiffiffiffiffiffiffiffiffiffiffi
h2
λ¼ ð2:5Þ
2meU
72 2 Microscopy of Polymers

The high resolution (d ) of the electron microscope is due to the short wavelength of the electrons,
i.e.:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 ffi
0:61λ 2
d¼ þ ðCs α3 Þ ð2:6Þ
α

where α is the half-angle aperture (in radians) and Cs is the spherical aberration coefficient. The first
term is the resolution given by diffraction, and the second term is that due to chromatic aberration.
The wavelength of the accelerated electrons in a 100 kV TEM is, according to Eq. (2.5), equal to
0.0037 nm. Insertion of this value and of α ¼ 6.103 rad and Cs ¼ 1.6 mm in Eq. (2.6) yields the
resolution d ¼ 0.5 nm, but this value is in practice higher due to other aberrations of the magnetic
lenses. The resolution increases with increasing voltage (cf. Eqs. 2.5 and 2.6). The diffraction-
dependent resolution of a 20 kV SEM is approximately 2 nm. However, most modern SEM
instruments have a secondary electron image resolution of 10–30 nm. The depth of field is given
by 2d/α. The aperture is normally very small (α is 102 to 103 rad), and the depth of field is a few μm
for both SEM and TEM, which is significantly larger than for OM.
The electron microscope operates under high vacuum. The electron beam in a conventional TEM
is stationary. By appropriate adjustment of the condenser lenses, the beam is highly collimated
(diameter of 5 nm), and it is allowed to scan over the specimen surface under the control of scanning
coils in both SEM and scanning TEM (Fig. 2.11). The advantage of scanning TEM over conventional
TEM is that no image-forming lenses are required after the specimen. This eliminates chromatic
aberration which otherwise arises due to multiple scattering in the sample. Thicker samples and lower
acceleration voltages for the same resolution can consequently be used in the scanning TEM. It is also
possible to achieve electron diffraction on areas as small as 3 nm in a scanning TEM. There are
nevertheless drawbacks with scanning TEM compared with conventional TEM: space-charge effects
from the small-diameter and high-current beam and poorer contrast in the diffraction images.
SEM provides information about the topography of a specimen. The electrons that strike the
surface of the specimen yield secondary electrons (used for the image), back-scattered electrons (used
for the image), Auger electrons (50–2000 eV) and X-rays. The secondary electrons have relatively
low energies, less than 50 eV with an average value close to 4 eV, and they cannot escape from depths
greater than 10 nm in the specimen. The wavelength of the X-ray emitted by a given element is
unique. Some scanning electron microscopes are equipped with an X-ray microanalyzer, and these
instruments provide information not only by the topography but also about the elemental composition
of the surface material. Energy dispersive X-ray photometers (EDS) are capable of simultaneously
detecting all elements of atomic numbers (Z ) greater than 11. With a special window, even lighter
elements can be detected. EDS is fast, whereas the spatial resolution in SEM is poor, and the results
may be ambiguous due to the overlap of X-rays emitted from different elements. The wavelength
dispersive X-ray spectrometer (WDS) analyzes only one element at a time. The detection limit for
WDS is Z > 3. WDS is slow, the spatial resolution is poor but the energy resolution is very good.
Figure 2.12 shows schematically the origin of diffraction contrast in TEM. A crystalline sample
diffracts the electrons, and it is possible to intercept all diffracted electrons using a small objective
aperture (Fig. 2.12, left). The final image is formed from only the undiffracted electrons in the bright-
field case. Dark-field images are obtained by adjusting the aperture of the objective so that only
diffracted electrons pass through the aperture (Fig. 2.12, right). It is possible to select specific
crystalline diffraction spots and to obtain dark-field images of crystals. The dark-field image is thus
the complement to the bright-field image. Contrast may also be obtained in a fully amorphous sample
because heavy atoms scatter electrons more than light atoms. A variation of the heavy atoms
(i.e. density) in the specimen gives rise to so-called density contrast. A region of high density scatters
more electrons than a region of low density, so that regions of low density appear bright, whereas
2.4 Atomic Force Microscopy and Related Techniques 73

Fig. 2.12 Bright-field and

dark-field adjustment of the
objective aperture

regions with high density appear dark in a bright-field image because the aperture of the objective
blocks the scattered electrons. It should be noted that the naturally occurring variations in density in
polymers are mostly insufficient and staining is often used to create a contrast. Staining involves the
selective addition of heavy elements to one of the phases, making it denser than other structural
regions.
Phase contrast also occurs in electron microscopy. The passage of the wave front through a
specimen causes a phase shift, which can be converted to image intensity by interference between
the retarded waves with another wave. Components, which cause phase contrast, have been tried in
the transmission electron microscope, but although they work, they are impractical.

2.4 Atomic Force Microscopy and Related Techniques

Atomic force microscopy (AFM) has become extremely popular among polymer scientists, since
AFM can reveal the surface profile and the mechanical performance down to an almost atomic
resolution of non-conductive materials such as polymers. The effort required in specimen preparation
is also much less than that required for TEM. Both solid and liquid samples can be studied, even under
non-vacuum conditions, and this is an advantage compared with electron microscopy. AFM is readily
equipped with a hot-stage which makes it possible to study processes like melting and crystallization
at different temperatures.
Figure 2.13 shows the basic components of the microscope. The cantilever (often made of silicon
or silicon nitride) and the sharp tip (with a radius of the order of nanometres) scan over the surface of
the specimen being studied in one of three different modes: (i) the non-contact mode, where the tip
does not touch the specimen surface at any time; (ii) the contact mode, where the tip touches the
specimen surface all the time; and (iii) the intermittent contact mode, where the tip oscillates and
touches the surface for part of the time. The latter is often referred to as the tapping mode. Basically,
the tip, which is solidly connected with the cantilever, is subjected to different forces: a normal force
and a lateral force. The cantilever is made of a hookean elastic material, and the extent of deforma-
tion of the cantilever is thus proportional to the acting force. The deflection of the cantilever is usually
assessed by a detector, which as shown in Fig. 2.13a can be an optical device, but deflection can be
determined by other means (Sch€onherr and Vancso 2010). The force giving rise to the deflection of
the cantilever originates from the interaction between the tip and the specimen. The potential energy
of atoms approaching a surface depends strongly on the distance between the two objects (Fig. 2.14).
More information about van der Waals forces is provided by Gedde and Hedenqvist (2019c). At
distances greater than the minimum, the force between the atom probe and the surface is attractive. At
shorter distances, the force is repulsive.
The first AFM instruments recorded the static deflection of the cantilever (static AFM) which, by
holding a set value for the deflection (i.e. the force), makes it possible to generate a topographical
image. The dynamic AFM instruments became dominant since the first development of static AFM.
74 2 Microscopy of Polymers

Fig. 2.13 (a) Sketch showing some of the components used in AFM instruments. (b) Cantilever oscillation in free
space and in close contact with a polymer sample. Note that free oscillation is in phase with the exciting force
originating from the piezoelectric component, whereas the contact oscillation is reduced in amplitude and shifted in
phase (by ϕ) with respect to that of the free oscillation

In these dynamic instruments, the spring-like cantilever oscillates (vibrates) at its eigenfrequency by
the action of a piezoelectric component (Fig. 2.13b). A piezoelectric material undergoes deformation
by an applied electric field; the piezoelectric effect was discovered by Jacques and Pierre Curie
in 1880.
In the non-contact mode, the cantilever is vibrating at its eigenfrequency (typically 100 kHz) with
an amplitude of less than 10 nm. The non-contact mode operates outside the minimum, and weak
attractive forces prevail. The forces between the sharp tip and specimen surface (intermolecular
forces of the van der Waal type) reduces the resonance frequency of the cantilever, which in turn is
counteracted by a feedback loop system to adjust the distance between tip and sample. This provides
information about the specimen height at each recorded x, y-point, i.e. the topography of the
specimen. One particular advantage of this technique is that both the specimen and the tip remain
untouched and thus undamaged. A drawback is that the liquid contamination of the surface may cause
capillary forces to act on the tip and dominate over the van der Waal forces, thus giving false results.
In the contact mode, the tip is pulled across the specimen in good physical (repulsive) contact with the
2.4 Atomic Force Microscopy and Related Techniques 75

Fig. 2.14 Potential energy and normal force of an atom probe approaching a surface as a function of the distance of
separation. From Sawyer and Grubb (1996). With permission from Chapman and Hall, London

surface. This method is also denoted the static mode, and it is a brutal process with some drawbacks:
(i) causing damage to both the sample and the tip and (ii) generating false results (artefacts). The
intermittent contact mode (tapping mode) avoids these problems, and it is the most suitable method
for the study of polymers (cf. Figure 2.13b). The cantilever is set to vibrate at its eigenfrequency when
the tip is far from the specimen surface. When the tip is moved close to the surface and in occasional
physical contact with the specimen, the amplitude is reduced (Fig. 2.13b). The topography is revealed
by the fact that the measured vibrational amplitude is dependent on the average distance between the
tip and specimen surface during the vibrational cycle. Furthermore, the vibration is also shifted in
phase with reference to the free vibration, i.e. a phase lag is recorded between the response of the
cantilever and the exciting force (Fig. 2.13b), and leads to phase images that reflect differences in
anelastic behaviour, adhesion and chemical composition between different structural regions on the
specimen.
When the tip is scanned over a surface (contact mode AFM), frictional forces cause bending of the
cantilever (Fig. 2.13a). In conventional AFM, the cantilever used is wide in order to minimize the
twisting, whereas in the frictional force microscope (FFM), the cantilever is thinner and the frictional
force can be accurately determined. FFM provides information not only about the surface topography
but also about the chemical adhesion.
AFM provides valuable information about morphology and about the kinetics of phase
transformations in polymers. The topographic information provides some insight about the morphol-
ogy. The chemical composition of different regions on a very fine scale (nm-range) provides even
more conclusive facts. This can be achieved using chemical force microscopy, which measures
frictional force or, in general, the force between a sample and a functionalized tip and thus provides
information about the local chemical composition. Charles Lieber and co-workers (Frisbie et al.
1994; Noy et al. 1997) pioneered this branch of AFM. There are several methods to prepare
the tip. One is to coat the tip with a thin layer of gold and a self-assembled monolayer. This
76 2 Microscopy of Polymers

technique has been developed and applied to polymers by Vancso and co-workers (Sch€ onherr and
Vancso 1997; Sch€onherr et al. 1998; Vancso et al. 2005). Other research groups have used silanized
tips (Tsukruk et al. 1998). Recommended further reading on this topic is Sch€ onherr and
Vancso (2010).
The macroscopic properties of complex polymer multiphase systems depend on the fractions of
the different phases and their phase morphology, both of which can be revealed by AFM and a few
other classical methods like DSC. Information about the phase properties is however missing. The
development of AFM methods, which probe properties on a very fine scale (nm-range), was triggered
by this lack. Mechanical, thermal, electrical and magnetic properties have been probed on the
nm-scale (cf. Sect. 2.6).
A wealth of information about the basic AFM methods is provided by Sarid (1991), Sawyer et al.
(2008), Sch€onherr and Vancso (2010) and Voigtl€ander (2015).

2.5 Novel Techniques in Polymer Microscopy

The local probing of properties on a very fine scale using AFM-based methods has been developing
since the 1990s. In Sect. 2.5, the development of chemical force microscopy by modification of the
chemistry of the probing tip is described. The local probing of mechanical properties by harmonic
imaging is one such development. The basic results of tapping mode AFM is an image based on the
phase shift between the response of the cantilever and the exciting force, but more information is
hidden in the deviation from the simple sinus wave of the vibration signal of the cantilever. The
motion of the cantilever can be divided into the fundamental motion and the higher harmonic modes
by Fourier analysis. It is in principle possible from this analysis to extract information about the
viscoelastic properties of the polymer. This line of research was started by Jamayo and Garcia (1997)
and Hillenbrand et al. (2000). More recent important contributions are by Haviland and co-workers
(Platz et al. 2008; Thorén et al. 2018). The book by Cappella (2016) provides comprehensive
information on how to obtain modulus values from AFM data and is a highly recommended text.
Thermal properties such as heat conductivity, heat capacity, enthalpy changes associated with first-
order phase transitions and glass transition temperature can be assessed on a sub-micrometre scale by
AFM-based methods using tip probes that are temperature sensing (resistance thermometer) and heat
injecting (e.g. a resistive probe). The injection of heat at a point-like region followed by the recording of
the change in temperature of the surrounding region provides the means for the thermal property
imaging, a technique referred to as scanning thermal microscopy. The pioneering work was carried out
by Williams and Wickramasinghe (1986); Majumdar et al. (1993); and Pylkki et al. (1994). Early
reviews are by Majumdar (1999) and Pollock and Hammiche (2001). Recommended texts for further
details are Sawyer et al. (2008) and a recent review by Zhang et al. (2019).
Other novel AFM-based methods are magnetic force microscopy (cf. review by Passeri et al. 2017)
and intermodulation electrostatic force microscopy (Borgani et al. 2014, 2016). In addition to the
AFM world, IR and Raman microscopy are fairly recent contributions to the field of polymer analysis.
Recommended readings about these methods are Sawyer et al. (2008) and Toporski et al. (2018).
Figure 2.15 shows a scanning electron micrograph of a hole in a polyethylene insulation caused by
high electric AC field. The structure formed is called an electrical tree, and the hole is part of one of
the channels of the tree. The tree structure is difficult to study by high-resolution techniques, but
nanotomography can make it possible to reveal details about the tree structure. Ptychographic X-ray
nanotomography is a coherent diffraction imaging technique that can provide a method to approach
nm-resolution. Coherent X-ray is used for illumination, and a coherent diffraction is recorded in the
far field, and several algorithms enable a 3D map of the electron density distribution to be
constructed, i.e. a three-dimensional density distribution of the sample. Further information about
2.5 Novel Techniques in Polymer Microscopy 77

Fig. 2.15 Scanning electron micrograph of a sample of polyethylene insulation that had been subjected to a high
electric field. Micrograph taken by Love Pallon, Fibre and Polymer Technology, KTH Royal Institute of Technology,
Stockholm

Fig. 2.16 Specimens prepared by serial focused ion beam SEM (FIB-SEM) used for ptychographic nanotomography.
Note the small size of the specimen. From Pallon et al. (2017) with permission from the American Chemical
Society, USA

this technique is provided by Faulkner and Rodenburg (2004), Rodenburg et al. (2007), Dierolf et al.
(2010), Diaz et al. (2012) and Holler et al. (2014). Figure 2.16 shows the sample used for the electrical
tree analysis. The first step was to use optical microscopy to find the region which contained the trees.
The final preparation step was to generate the tiny pillar specimen using FIB-SEM.
Figure 2.17 shows details of the tree structure in a 3D view with information about the magnitude
of the density fluctuations, which are especially clear in the 2D cuts.
78 2 Microscopy of Polymers

Fig. 2.17 (a–d) Three-dimensional view of the isosurface of the electrical tree; the red regions have a density of
600 kg m3. The diameters of the left (L) and right (R) branches are ca. 1 μm. The growth of the electrical tree was
downwards in the direction of the electrical field (marked E). (e–g) Two-dimensional density cuts through the structures
with paths of reduced density between the prestep structures and the main channels, as indicated with red arrows. From
Pallon et al. (2017) with permission from the American Chemical Society, USA
2.6 Preparation of Specimens for Microscopy 79

2.6 Preparation of Specimens for Microscopy

A general rule in microscopy is that the preparation of a specimen takes a much longer time than the
actual examination. The difficulty in preparing good samples is substantial, particularly for transmis-
sion electron microscopy. More detailed descriptions of the preparation methods can be found
in Patel (1980), Curson (1989), Woodward (1989), Sawyer and Grubb (1996) and Sawyer
et al. (2008).

2.6.1 Preparation of Samples for Optical Microscopy

Specimens for OM should be thin (1–20 μm) and without scratches or zones of plastic deformation on
the surface. Thin films are readily obtained by compression moulding or by solution casting of 1%
solutions. Small drops of the solution are placed on an object glass kept on a hot plate. The
temperature of the plate should be 30–40  C above the melting point of the polymer. The solvent is
evaporated, and the sample is covered with a cover glass. Solution casting is particularly suitable for
preparing thin specimens (1 μm). Very thin specimens can be produced by spin coating.
Samples can also be obtained by sectioning from thicker objects using a microtome. This is a
difficult procedure, which requires both skill and experience. A correct combination of pre-specimen
geometry, knife and temperature has to be used. Soft polymers should be sectioned at cryogenic
temperatures. The small section obtained is inserted between object and cover glasses. It may be
helpful to use a refractive index matching immersion oil to avoid disturbing scattering from surface
scratches.

2.6.2 Preparation of Samples for Scanning Electron Microscopy

Polymer specimens to be examined by SEM are first coated with a thin layer of an electrically
conductive material. The electrons that strike the specimen are partly scattered and subsequently
recorded by the detector and partly conducted to the grounded electrode. The coating is achieved by
vacuum evaporation or sputtering of a heavy metal (e.g. Au–Pd) or carbon. The specimen is often
glued to a metal-base specimen holder with an electrically conductive adhesive to achieve good
electrical contact with the grounded electrode. Samples prepared for elemental analysis using the
X-ray microanalyzer should be coated with carbon. SEM can be used for many different purposes.
One is fractography, which require no pre-treatment of the sample other than a coating with metal or
carbon. Another is the study of fibre orientation in fibre composites, which requires grinding and
polishing using an abrasive paste. The phase morphology of polymer blend can be studied on samples
prepared by different methods: solvent or degrading extraction of a polished sample to remove one of
the components or fractography of samples cracked at cryogenic temperatures. Spherulites in
semicrystalline polymers can be studied on solvent-etched samples. Degrading etchants can also
be used.

2.6.3 Preparation of Samples for Transmission Electron Microscopy

The preparation of specimens for TEM is both difficult and time-consuming. The natural variation in
density is seldom sufficient to achieve adequate contrast, and contrast must be obtained by staining or
80 2 Microscopy of Polymers

by etching followed by replication. Staining involves a chemical reaction between one of the
components and the staining reagent. Atoms of higher elements are added selectively to this phase,
and regions of different electron density are achieved in the stained specimen (Fig. 2.18).
Treatment of polydienes with osmium tetroxide (OsO4) was introduced by Andrews (1964) and
Andrews and Stubbs (1964). A later development by Kato (1965, 1966, 1967) showed that OsO4 was
able to stain the rubber component in polymer blends. Osmium tetroxide adds to the double bonds,
and the density of the rubber phase is increased. Since the OsO4 cross-links the rubber polymer by
adding to two double bonds located in two different polymer chains (Fig. 2.19), the rubber is fixed,
and sectioning can be carried out without smearing. Osmium tetroxide staining has been used on
ABS, high-impact polystyrene, other rubber-containing plastics, styrene–butadiene–styrene
copolymers and poly(ethylene terephthalate). The morphology of semicrystalline polyisoprenes has
also been studied. The osmium tetroxide diffuses only into the amorphous component leaving the
crystals unaffected. Osmium tetroxide has a high vapour pressure making vapour staining possible.
Osmium tetroxide should be treated with great care since it adds not only with double bonds in
polymers but also with the membranes of the eye.
Ruthenium tetroxide is a stronger oxidizing staining agent than OsO4, and RuO4 can add also to
phenyl groups, e.g. in polystyrene. Ruthenium tetroxide has also been used to stain polyethylene and
polypropylene in addition to the list of polymers stained by OsO4.
The ability of chlorosulphonic acid to stain the amorphous component of polyethylene was first
reported by Kanig (1973). The strong acid diffuses and reacts almost exclusively with the amorphous
component leaving the crystals unaffected. Sulphur, chlorine and oxygen are introduced into the
amorphous component, and the density of the amorphous component increases to a value above that

Fig. 2.18 The effect of staining. The heavy element is added only to the discrete phase

Fig. 2.19 Reaction

scheme for the addition of
OsO4 to unsaturated
polymer molecules
2.6 Preparation of Specimens for Microscopy 81

Fig. 2.20 Penetration

depth (h) as a function of
chlorosulphonic acid
treatment time: LDPE filled
circles, HDPE open circles,
PP filled squares. (From
Kalnins et al. (1992). With
permission from Elsevier,
UK)

of the crystalline component. Elemental analysis reported by Kalnins et al. (1992) showed that
chlorosulphonation causes a major reduction in hydrogen content from about 14% to only 4.3%
and that the chlorosulphonated material contains 38.5% of C, 25.4% of O, 25.2% of S and 6.2% of
Cl. The O:S ratio is about twice that in the SO2Cl group, which indicates that the polymer is oxidized
by the chlorosulphonic acid treatment. The chlorosulphonic acid cross-links the polymer, thus fixing
the stained material to be sectioned in a manner similar to that achieved with OsO4 in unsaturated
rubber. The contrast is further enhanced by treatment with uranyl acetate. Chlorosulphonation of
polyethylene makes it possible to assess the lamellar structure, although only crystals with their fold
surfaces parallel to the electron beam are clearly revealed. More details about the chlorosulphonation
of polyethylene are presented in Gedde and Hedenqvist (2019a). A polyethylene sample treated with
chlorosulphonic acid turns black, and the thickness of the black skin layer increases approximately
linearly with time (Fig. 2.16).
The rate of staining decreases with increasing crystallinity of the polymer. Polypropylene also
reacts with chlorosulphonic acid, but this polymer disintegrates rather than cross-links, and the black
layer is very thin even after prolonged treatment (Fig. 2.20).
Table 2.2 presents a summary of commonly used staining agents. The list is not complete, but the
more commonly used staining agents are presented.
A typical specimen that is examined by TEM consists of a series of thin (50–100 nm) sections of
stained polymer on a microscopy grid. Thin sections are produced in an ultramicrotome. Sectioning
requires great care, skill and experience and involves specimen mounting, optional embedding
(epoxy, polyesters and polymethacrylates are commonly used), curing, trimming and finally section-
ing. Glass knives and diamond knives are mostly used. Glass knives are cheap and easy to make but
only remain sharp for less than 1 h. Glass knives can be used for polymers softer than glass. Diamond
knifes are expensive but sharp, and they can also be re-sharpened. Figure 2.21 shows the knife, the
trough and how the sectioned samples float in a row in the trough. Soft polymers cannot be sectioned
at room temperature without a fixing (staining) agent. They require cryo-ultramicrotomy,
i.e. sectioning at low temperature. Several problems arise when sectioning at a lower temperature
such as freezing of the trough fluid and frost build-up. Reid (1975) is recommended reading on this
subject.
Etching involves the selective removal of a specific component from the structure. The topography
of an etched sample reflects the distribution of the component removed from the structure. There are
two main types of etching: solvent etching and chemical etching. It is important to confirm the
efficiency of the etching by techniques such as DSC or spectroscopic methods.
Solvent etching removes whole molecules from the sample without any chemical reaction. It is
possible to remove one of the components from an immiscible polymer blend by this method. A low
melting species can also be selectively removed from a semicrystalline polymer. The solvent and the
treatment temperature are the two factors that are changed in order to achieve the best treatment.
82 2 Microscopy of Polymers

Table 2.2 Staining agents used for TEM

Functional
groups Polymers Staining agents
–CH2–CH2– Polyethylene Chlorosulphonic acid
–CH2–CHCH3– Polypropylene Ruthenium tetroxide phosphotungstic acid
–CH¼CH– Polydienes, etc. Osmium tetroxide ruthenium tetroxide
–OH;–COH Polyalcohols, Osmium tetroxide, ruthenium tetroxide
polyaldehydes
–O– Polyethers Osmium tetroxide, ruthenium tetroxide
–COOR Polyesters Hydrazine followed by osmium tetroxide phosphotungstic acid
silver sulphide
–CONH- Polyamides Phosphotungstic acid tin chloride
–Ph; –Ph– Ruthenium tetroxide silver sulphide mercury trifluoroacetate
Source: Sawyer and Grubb (1996)

Fig. 2.21 Ultramicrotomy

using a glass knife

Blends of amorphous polymers can be etched with good solvents of the minor component leaving
holes in the etched surface. Crystalline polymers require etching at elevated temperatures. Low
melting species of polyethylene can be selectively removed by etching with p-xylene. The tempera-
ture chosen for the treatment should be approximately 30 K lower than the maximum melting point
of the species to be dissolved and removed from the sample (Fig. 2.22). This relationship is
applicable to linear polyethylene of low and intermediate molar mass, but high molar mass polyeth-
ylene and slightly branched polyethylene cannot be etched efficiently at low temperatures (Gedde
et al. 1983).
Chemical etching is accomplished by the degradation and removal of one of the components from
the sample. The classical techniques are etching with plasma and ions, but both these techniques tend
to produce artefacts. Acids, alkalis and n-alkyl amines are used to etch polyesters. Nitric acid
degrades the amorphous component of polyethylene leaving the crystals intact, but the degradation
is so complete that the sample disintegrates into a powder.
The most important development in this area is due to Olley and Bassett (1982) who invented
permanganic acid etching. The original recipe consisted of 7% potassium permanganate in
concentrated sulphuric acid. Later recipes contained orthophosphoric acid, and the amount of
potassium permanganate was reduced to 1–2%. Etching using the original recipe led in some cases
to artificial structures, particularly evident at low magnifications. The permanganic acid degrades the
2.6 Preparation of Specimens for Microscopy 83

Fig. 2.22 Cumulative

melting and dissolution
(in p-xylene) of a linear
polyethylene crystallized at
401 K to completeness and
then rapidly cooled to room
temperature. Drawn after
data of Gedde et al. (1983)

Table 2.3 Etching techniques

Polymer Etchant Comments
Polyethylene Hot p-xylene, toluene and Low melting species are removed
CCl4
Polyethylene, isotactic Permanganic acid Lamellar morphology and superstructure are
polypropylene, isotactic revealed. Artefacts have been reported in low
polystyrene, polyether ether ketone magnification images
Polyesters including PET n-Alkylamines Superstructures and lamellar morphology are
revealed. Artefacts have been reported
ABS 10 M chromic acid, Degrades the rubber phase
sulphuric acid, chromic
acid and water
PMMA in blends Electron irradiation in PMMA is degraded and contrast develops
electron microscope
Polyamides Aromatic and chlorinated Low melting species are removed
hydrocarbons
Source: Sawyer and Grubb (1996)

amorphous more readily than the crystalline component, and the lamellar structure of the semicrys-
talline polymer is revealed. The lamellar structures of several semicrystalline polymers, including
polyethylene, isotactic polypropylene, poly(butene-1), poly(4-methylpentene-1), isotactic polysty-
rene and poly(etherether ketone), have been studied. One advantage of the permanganic acid method
over the staining methods (e.g. chlorosulphonation) is that the crystal lamellae are made visible,
regardless of their orientation. It is possible to assess their thickness from a ‘side view’ and their
lateral shape from a ‘top view’. The staining methods reveal only crystallites viewed from the side.
Table 2.3 presents a summary of etching methods, but not a complete list.
The surface topography of etched samples can be studied directly with SEM (after metal or carbon
coating) or with TEM after replication. The replica technique has the advantage that the specimen
examined contains no beam-sensitive polymer. Single-stage or direct replicas show the best resolu-
tion and are fast although difficult to make. The sample is coated with carbon, shadowed with a heavy
metal (e.g. Pt) at an angle of between 30 and 45 . The replica is then stripped off from the sample and
transferred to the microscopy grid. The stripping is the difficult step. The preparation of a two-stage
replica involves the formation of a replica using a polymer solution, stripping off the replica from the
84 2 Microscopy of Polymers

specimen, shadowing with a heavy metal, carbon coating, placing the shadowed replica on the grid
and dissolution of the plastic film. Sawyer and Grubb (1996) provide more information.
A number of special preparation techniques have been developed for single polymer crystals. The
thickness of a single crystal is of the order of 10–20 nm, and it can be determined by shadowing,
i.e. evaporation of a heavy metal at a known angle (cf. Gedde and Hedenqvist 2019a). Gold
decoration is a technique which reveals very fine surface steps on crystals. The specimen is coated
with a very thin coating of gold (< 1 nm) and the metal nucleates on edges of steps. The specimen is
then coated with carbon, and the carbon film with the gold is stripped off and examined in a
microscope.
The direction and regularity of the folds in the fold surface can be revealed by decoration of single
crystals with low molar mass polyethylene. Wittmann and Lotz (1985) showed that the low molar
mass polyethylene crystallizes epitaxially in rods on the fold surface, where the long axis of the rods
is oriented perpendicular to the direction of the folds. The method has been applied to a range of
different polymers, e.g. polyethylene, isotactic polypropylene and polyoxymethylene (Wittmann and
Lotz 1985), and a good example of the potential of this method is shown in Fig. 2.23.

2.6.4 Preparation of Samples for Atomic Force Microscopy

It is much easier to prepare specimens for AFM than for TEM. Single crystals, semicrystalline
polymers solidifying with a ‘free’ surface, fracture surfaces or etched samples can be studied directly
without any further treatment. Some polymers like silicone rubber may however contain uncross-
linked species which migrate to the surface to form a liquid coating, and it is beneficial to remove the
liquid surface layer by extraction prior to the AFM studies.

Fig. 2.23 Transmission electron micrograph of a single crystal of polyoxymethylene decorated with polyethylene
vapour. With permission from Wiley (Wittmann and Lotz 1985)
2.7 Applications of Microscopy in Polymer Science and Engineering 85

2.6.5 Artificial Structures

Artificial structures or simply artefacts are not true features of the structure of a material. Artefacts are
created by the preparation method or during examination such as radiation damage in the case of
TEM and SEM and mechanical damage in the case of AFM. No observed structure should be taken
for granted until the possibility of artefact has been considered and eliminated. Another problem with
microscopy lies in the choice of typical structures of a given sample, and this depends strongly on the
experience of the microscopist. The following are a few of the reported artefacts:
• Etching with permanganic acid using the original recipe (7% KMnO4) in concentrated H2SO4
leads to the formation of large regular (hexagonal in some cases) structures that in some reports on
polyethylene were mistaken for spherulites. It was later shown that these were artefacts formed by
precipitation of a manganese salt.
• In some cases, solvent etching causes in some cases local swelling of the polymer and the
formation of regular artificial topographical structures: CCl4 vapour etching of cross-linked
polyethylene led to large (10–100 μm) quilt-like patterns falsely reported to be spherulites,
whereas the ‘true’ spherulites had a diameter of 1–2 μm.
• Freeze-fracturing may lead to the formation of spherulite-shaped crack patterns without any
resemblance to the true spherulitic structure.
• Chlorosulphonated sections of polyethylene may shrink considerably in the electron beam due to
thermal effects. The shrinkage is more pronounced in the case of samples treated with
chlorosulphonic acid for a short period of time. The more numerous cross-links present in samples
subjected to prolonged acid treatment prevent shrinkage of the sections. As a result of the
shrinkage, the dimensions of morphological features apparently become smaller.
• Artefacts concerned with AFM may arise from the deformation of the specimen surface or from
the tip shape. A surface with sharp steps cannot be accurately imaged. A blunt tip will give a
smeared out image of a sharp structural feature. The shape of the tip can be ‘calibrated’ by
preparing an image of latex particles of known diameter.

2.7 Applications of Microscopy in Polymer Science and Engineering

The list of applications of microscopy in polymer science and engineering is almost infinite.
Microscopy is useful in all fields, and the understanding of many phenomena is based on data
generated by microscopy, but a number of reports have nevertheless obviously presented data of
doubtful character. The possibility of obtaining artefacts is not negligible. It is also possible that
non-representative features have been reported. We shall here present a few areas in which micros-
copy has played a vital role. The list of good examples can however be made significantly longer. The
beautiful atlas of polymer morphology of Woodward (1989) is highly recommended. A very
comprehensive collection of polymer applications, with many beautiful optical, electron and atomic
force micrographs, has been presented by Sawyer et al. (2008).

2.7.1 Semicrystalline Polymers

The brilliant Andrew Keller discovered chain folding in polymer crystals in 1957 through TEM
studies of solution-grown polyethylene single crystals. Replicates of the single crystals were exam-
ined in the microscope, and the peculiar geometry of the crystals and the electron diffraction pattern
86 2 Microscopy of Polymers

Fig. 2.24 Transmission electron micrograph of replicate of single crystals of linear PCL. With permission from
Elsevier, UK (Núñez and Gedde 2005)

Fig. 2.25 Transmission

electron micrograph of a
replica of a single crystal of
polyethylene decorated
with polyethylene vapour.
With permission from
Wiley, New York
(Wittmann and Lotz 1985)

led Keller (1957) to the surprising conclusion that the chains must be folded. Figure 2.24 presents an
electron micrograph of a single crystal based on linear PCL prepared by metal shadowing as done by
Keller.
The sectorization was obvious to Bassett et al. (1963) at that time, but the decoration techniques
developed by Wittmann and Lotz (1985) provided further evidence of the regularity and direction of
the chain folds in the different sectors (Fig. 2.25); the long axes of the deposited rods are perpendicu-
lar to the direction of the folds. A detailed treatise on chain folding in solution-grown single crystals
and in melt-crystallized samples is presented in Gedde and Hedenqvist (2019a).
The AFM technique has been used more recently to study polymer single crystals, but the AFM
images seldom reveal molecular folds. Frictional anisotropy in the single crystals was discovered
by Vancso and co-workers (Nisman et al. 1994; Smith et al. 1994; Pearce and Vancso 1998), and
this was attributed to sectorization and to preferential chain folding parallel to the growth surface
of the crystals. Figure 2.21a and b shows an AFM height image and a surface profile along a line
of poly(4-methyl-1-pentene). The sector boundaries protrude 1 nm from the other parts of the single
crystals. The frictional force anisotropy is also evident in the lateral force images (Fig. 2.21c).
2.7 Applications of Microscopy in Polymer Science and Engineering 87

Fig. 2.26 (a) AFM height image of a single crystal of (poly(4-methyl-1-pentene). (b) Lateral force image of a single
crystal of the same polymer. With permission from Elsevier, UK (Pearce and Vancso 1998)

If chain folding were random as suggested by Paul Flory in his famous random switchboard model,
the frictional force anisotropy would be largely absent (Fig. 2.26).
Gedde and Hedenqvist (2019a) have shown that there is a structural hierarchy in semicrystalline
polymers and that the superstructures, spherulites, axialites and the other organizations are conve-
niently assessed by polarized light microscopy, although possible difficulties may arise from the
problems in making the specimen sufficiently thin, particularly in the case of very small
superstructures. Spherulites as small as 1 μm in diameter have been reported, and they cannot be
resolved by conventional polarized light microscopy. Small-angle light scattering is the ideal method
to assess the superstructure of samples containing small spherulites (cf. Gedde and Hedenqvist
2019a), and it is also possible to obtain meaningful results from solvent-etched samples examined
with SEM. The scanning electron micrograph shown in Fig. 2.27 is from a polyethylene sample
crystallized under isothermal conditions and rapidly cooled to room temperature. The segregated
material that crystallized during the rapid cooling period has been removed by solvent etching
making it possible to observe spherulites.
The lamella organization is preferably studied with TEM. Let us consider polyethylene, for which
two preparation methods are commonly used: staining with chlorosulphonic acid/uranyl acetate and
etching with permanganic acid.
Figure 2.28 shows a transmission electron micrograph of two chlorosulphonated samples, where
the crystal lamellae appear white and the amorphous component dark. Crystals with sharp contrast
have their fold surfaces parallel to the direction of viewing, the latter being along [uv0]. It is typical to
find ‘blank’ areas with no lamellar contrast on the micrographs, because tilted lamellae show no
contrast. The crystal lamellae are, however, seen only when they are approximately parallel to the
electron beam, which is a clear disadvantage of the method. A major advantage of the chlorosul-
phonation method is that the thickness of both the crystal lamellae and the amorphous interlayer can
be determined with high precision.
The location of the sections is placed at random with respect to the centres of the spherulites
(axialites). From simple geometrical considerations, it can be shown that the average distance
between the section and the spherulite centre is R/31/2 where R is the radius of the spherulites. It is
known that [010] is parallel to the radius of a mature spherulite. Hence, sharply appearing crystal
lamellae are dominantly viewed along [010]: 40% of the surface is within an angle of 20 from [010]
and 60% within an angle of 30 from [010].
The linear polyethylene displays relatively straight and wide crystal lamellae. The amorphous
interlayers are generally very thin, and occasional roof-ridged lamellae are found, while the number
88 2 Microscopy of Polymers

Fig. 2.27 Scanning electron micrograph of p-xylene-etched samples of linear polyethylene crystallized at 401.2 K and
then quenched. With permission from Elsevier, UK (Gedde and Jansson 1984)

Fig. 2.28 Transmission electron micrographs of chlorosulphonated samples of (a) linear polyethylene and (b)
branched polyethylene. Micrographs of Maria Conde Braña, Polymer Technology, KTH Royal Institute of Technology,
Stockholm

of lamellae per stack is high. The branched polyethylene displays wide S-shaped lamellae
surrounding significantly more narrow lamellae of the same or lower thickness.
Voigt-Martin and Mandelkern (1989) compared crystal thickness data obtained for several linear
polyethylenes by TEM (chlorosulphonated thin sections), Raman spectroscopy and small-angle
X-ray scattering (SAXS) on several linear polyethylenes. TEM, Raman spectroscopy and SAXS
data for a sample with a narrow thickness distribution ranging from 10 to 30 nm, and an average
crystal thickness of 21.5 nm, obtained by crystallization at relatively low temperature exhibited
perfect agreement in crystal thickness as obtained by the different methods. In samples exhibiting
pronounced bimodal distributions or very broad distributions in crystal thickness, TEM clearly
underestimated the thicker crystals. In a sample having a bimodal distribution with maximum values
2.7 Applications of Microscopy in Polymer Science and Engineering 89

Fig. 2.29 Transmission electron micrographs of replicas of polyethylene etched with permanganic acid. Micrographs
of Maria Conde Braña, Polymer Technology, KTH Royal Institute of Technology, Stockholm

at 10 and 15 nm, TEM overemphasized the thinner crystals. Both populations appeared in the
histograms but with incorrect weights. Another sample that was crystallized at 403 K exhibited a
very broad crystal thickness distribution with maxima at 10, 29, 36 and 40 nm. The crystals with
maxima at 10, 29 and 36 nm were revealed by TEM but with different weightings, and the 40 nm
crystals were detected only by Raman spectroscopy and not by TEM.
It has been argued on several occasions that the crystal thickness values obtained by TEM on
chlorosulphonated specimens are too low. Hill et al. (1992) showed that the sections may shrink when
they are exposed to the electron beam, even at low intensities, particularly if the samples had been
chlorosulphonated for only a short period of time.
Figure 2.29 shows transmission electron micrographs of replicas of samples etched with
permanganic acid. One of the advantages is directly noted. There are no blank regions. All crystals,
irrespective of orientation, are revealed. It is possible to find regions where the crystals are seen from
the side with the chain axis in the sample plane, and it is also possible to find regions that reveal the
lateral shape of the crystals, i.e. with the chain axis perpendicular to the sample plane. Another
advantage of permanganic acid etching is that replicas are not beam-sensitive. A disadvantage is that
it is impossible to measure the thickness of the amorphous interlayer. It is possible, however, to make
a fairly accurate assessment of the long period.
A very important question addressed by polymer morphologists concerns the continuity of crystals
and the mechanism of lamellar branching in spherulites (cf. Gedde and Hedenqvist 2019a). The
permanganic acid etching technique has provided new insight into these problems. Bassett and
co-workers (1988) pointed out the importance of screw dislocations for the spreading of the crystal
lamellae in three-dimensional space (Fig. 2.30a). In a more recent study, White and Bassett (1998)
presented electron micrographs of the α-crystals of isotactic polypropylene treated with permanganic
acid. Row structures were grown from thin samples in which a glass fibre was pulled through the melt.
According to these authors, it is easier to measure and identify lamellar orientation and how it
changes as a function of distance from the nucleation site for a linear nucleus than for a normal
spherulitic crystallization. Figure 2.30b shows a nice example of dominant radial lamellae growing
almost at right angles to the central row together with daughter crystal lamellae at an angle of 80 to
the dominant lamellae.
Real-time crystallization experiments were classically done by hot-stage polarized light micros-
copy, the growth of the bright spherulites (axialites) against a dark background (molten, isotropic
material) being readily followed in the microscope (Fig. 2.31). The linear growth rate of the
90 2 Microscopy of Polymers

Fig. 2.30 (a) Transmission electron micrograph of replica of linear polyethylene etched with permanganic acid. The
micrograph shows the spiral development around a screw dislocation in a sample crystallized at 130  C. With
permission from Elsevier, UK (Bassett et al. 1988). (b) Transmission electron micrograph of replica of isotactic
polypropylene crystallized from a fibrous nucleus and etched with permanganic acid. With permission from Elsevier,
UK (White and Bassett 1998)

Fig. 2.31 Polarized photomicrograph of growing spherulites: (a) ‘perfect’ spherulite of star-branched PCL; (b)
irregular spherulite in linear PCL. The bars represent 50 μm. With permission from Marcel Dekker (Núñez et al. 2004)

spherulites (axialites) at a constant temperature is readily and accurately determined from a series of
polarized photomicrographs taken after different times. The detailed lamellar morphology of both
growing spherulites and axialites can be captured with AFM (Pearce and Vancso 1997). Figure 2.32
shows the growth of PEO spherulites in deflection images. The linear growth rate as obtained by AFM
was comparable with data obtained by conventional hot-stage polarized light microscopy. The molten
region in between two growing spherulites forms a valley (Fig. 2.32c), which suggests that the
amorphous chains are pulled into the growing crystals.
2.7 Applications of Microscopy in Polymer Science and Engineering 91

Fig. 2.32 (a) AFM deflection images of poly(ethylene oxide) spherulites at 57  C. The elapsed time between the
images was 3 min. (b) and (c) Height profile across the crystalline/melt interface (along x) for a growing poly(ethylene
oxide) spherulite. With permission from the American Chemical Society (Pearce and Vancso 1997)

2.7.2 Liquid Crystalline (LC) Polymers

Polarized microscopy has been one of the most important methods for the structural assessment of LC
polymers. It goes back to the work by Demus and Richter (1978) who presented a great many
photomicrographs of small-molecule liquid crystals. Figure 2.33 shows a series of photomicrographs
of different liquid crystalline and crystalline monomers and polymers. The schlieren texture
(Fig. 2.33a) is typical of a nematic organization. The presence of disclinations of strength 1/2 is
unique for the nematics. The focal conic texture evident in Fig. 2.33b is typical of a smectic A phase.
Polarized microscopy can be used as one of several methods for the assessment of liquid
crystalline structure. The optical textures displayed by LC polymers are smaller than those observed
in small-molecule liquid crystals. Overlapping structures are often a problem, making observation of
the characteristic structural features difficult. Polarized light microscopy is simple to use and thus
deserves to be used, but X-ray diffraction, particularly in the case of aligned samples, provides more
definite answers about the nature of the liquid crystal phase.
TEM has been used to study the morphology of thermotropic LC polymers (Hedmark et al. (1987).
Figure 2.34 shows the morphology of X7G (a copolyester based on p-hydroxybenzoic acid and
ethylene terephthalate). The disperse phase is softer and richer in ethylene terephthalate and can be
etched with n-alkyl amines. Injection-moulded specimens of the same polymer show pronounced
skin-core morphology as shown in Fig. 2.35. This scanning electron micrograph of the n-
92 2 Microscopy of Polymers

Fig. 2.33 Polarized photomicrographs of liquid crystalline compounds: (a) schlieren texture, which indicates a
nematic structure; (b) focal conic texture, which indicates sA structure. Photomicrographs taken by Fredrik Sahlén,
Polymer Technology, KTH Royal Institute of Technology, Stockholm

Fig. 2.34 Transmission electron micrograph of poly( p-hydroxybenzoic acid-co-ethylene terephthalate) with 60 mole
% p-hydroxybenzoic acid. With permission from Wiley (Hedmark et al. 1987)

Fig. 2.35 Scanning electron micrograph of half cross-section of a 3 mm thick ruler etched with n-propylamine. With
permission from the Society of Plastics Engineers, USA (Hedmark et al. 1988)
2.7 Applications of Microscopy in Polymer Science and Engineering 93

Fig. 2.36 Transmission electron micrograph of an injection moulded specimen of a liquid crystalline polymer (Vectra
A; a copolyester based on p-hydroxybenzoic acid and hydroxy naphthoic acid) treated with permanganic acid. With
permission from the Society of Plastics Engineers, USA (Hudson et al. 1994)

propylamine-etched sample shows that the skin consists of aligned fibres and that the core is
essentially non-oriented. These findings have been further substantiated by X-ray scattering studies.
Hudson et al. (1994) used TEM to detect molecular orientation and disclinations in injection-moulded
LC polymers. They found that the skin was highly oriented with a very low concentration of
disclination defects. The less oriented core showed a disclination density which was six orders of
magnitude greater than that of the skin. Etching with permanganic acid, which removed disordered
material between the crystallites, revealed the disclinations. Figure 2.36 shows a transmission
electron micrograph of a region with two disclination pairs.

2.7.3 Polymer Blends

The morphology of polymer blends is of significant scientific and technological importance. Tough-
ened glassy polymers and self-reinforced blends are good examples of technically important systems,
where it is important to be able to assess the miscibility of the constituents. A detailed discussion of
the thermodynamics of polymer blends and about the currently available techniques for the assess-
ment of miscibility has been presented by Gedde and Hedenqvist (2019d). A few micrographs
showing the result of different preparation methods are presented here. Polyethylene and polystyrene
are incompatible, but it is possible to increase the interfacial adhesion and to reduce the size of the
polystyrene domains in the polyethylene matrix by adding a few percent of a styrene-b-ethylene-co-1-
butene-b-styrene (SEBS) tri-block polymer. Figure 2.37 shows scanning electron micrographs of a
blend of polyethylene-polystyrene-SEBS. Two different preparation techniques have been employed:
solvent etching using chloroform, which is a good solvent for polystyrene, and fracturing at a
cryogenic temperature.
The solvent-etched specimen was first polished. The spherical holes indicate the presence of
spherical polystyrene particles (Fig. 2.37a). The fracture surface contains many small spherical
particles, which consist of polystyrene. The polyethylene component had undergone some plastic
deformation, which is seen in the microfibrils appearing in the micrograph (Fig. 2.37b).
Finer details can be resolved by TEM either on stained samples (Fig. 2.38) or on replicas of etched
samples. The classical osmium tetroxide treatment has proven to be very useful. Apart from the
94 2 Microscopy of Polymers

Fig. 2.37 Scanning electron micrographs of blends of polyethylene (50 parts), polystyrene (50 parts) and SEBS
(5 parts): (a) chloroform-etched sample; (b) fractured sample at 23  C, with permission from Society of Plastics
Engineers, USA (Gustafsson et al. 1993)

Fig. 2.38 Transmission electron micrograph of a sample of OsO4-stained PVC/poly(ethylene-co-vinyl acetate) with
45 wt.% vinyl acetate. The poly(ethylene-co-vinyl acetate) phase was first hydrolyzed by reflux boiling in a solution of
sodium hydroxide. The dark regions are the OsO4-stained poly(ethylene-co-vinyl acetate) phase. Micrograph of Bj€ orn
Terselius, Polymer Technology, KTH Royal Institute of Technology, Stockholm

staining of unsaturated rubber polymers, it effectively stains structures containing vicinal hydroxyl
groups. The vinyl-acetate groups of EVA-45 are, for example, stained by OsO4 after hydrolysis to
vinyl alcohol (Fig. 2.38).
Polymer blends based on liquid crystalline polymers, and flexible-chain polymers are readily
processed due to their low viscosity, and after injection moulding, they possess anisotropic properties
with a Young’s modulus along the major flow direction typically in the range of 10 GPa (Engberg
et al. 1994). The scanning electron micrographs of fracture surfaces of Vectra (LC polymer) and poly
(butylene terephthalate) show a change in morphology from non-reinforcing globules of LC polymer
to highly reinforcing LC polymer fibres when the content of LC polymer in the blends is increasing
(Fig. 2.39).
2.7 Applications of Microscopy in Polymer Science and Engineering 95

Fig. 2.39 Scanning electron micrographs of fracture surfaces of injection moulded blends of Vectra A (a copolyester
based on p-hydroxybenzoic acid and hydroxy naphthoic acid) and poly(butylene terephthalate); (a) 28 vol.% LC
polymer; (b) 58 vol.% LC polymer. From Engberg et al. (1994) with permission from the Society of Plastics
Engineers, USA

2.7.4 Composites Including Nanocomposites

Short-fibre composites consist of sub-mm- and mm-long fibres with different orientations. The elastic
modulus of the fibres is approximately two orders of magnitude greater than that of the polymer
matrix. The fibre orientation distribution is the most important factor determining the mechanical
performance of the composite. Other factors that have an impact on the mechanical properties of
these materials are the fibre length distribution and the volume fraction of fibres. Figure 2.40 shows a
photomicrograph taken of a cross-section of an injection-moulded ruler. The specimen had been
polished to smoothness before examination in the optical microscope. It is also possible to use SEM
after polishing and metal coating. Contact microradiography can also be used to assess fibre orienta-
tion and fibre length (Darlington et al. 1976). A circular cross-section indicates that the fibre is
parallel to the normal to the surface. Highly elongated fibre cross-sections are obtained if the fibres
have their long axes in the specimen plane. It is, however, impossible from a single micrograph to
distinguish a fibre oriented from bottom left to top right from a fibre oriented from top left to bottom
right.
A strong interface between the fibres and the polymer is very important to ensure good adhesion
and efficient stress transfer between the two phases. This is commonly accomplished by applying a
surface coating to the fibre. Polymer nanocomposites consist of a polymer and particles with at least
one of its dimensions in the nanometre range (< 50 nm). The particles can be equiaxed like the one
shown in Fig. 2.41, where the aluminium oxide nanoparticles are spherical with a typical diameter of
20–30 nm (Fig. 2.41a). Figure 2.41b and c shows transmission electron micrographs of aluminium
oxide nanoparticles coated in an aqueous solution with two different n-alkyl triethoxysilanes; the
coating layer is revealed as a ca. 5 nm thick layer on the denser (darker) spherical particles. This
hydrophobic coating layer makes these particles compatible with hydrophobic polymers like poly-
ethylene (Liu et al. 2015, 2017) (Fig. 2.42). Composites with uncoated nanoparticles show
agglomerates of nanoparticles (Fig. 2.42a, c) and also a wider inter-particle distribution compared
to the composite which contained the coated nanoparticles (Fig. 2.42b).
Well-dispersed nanoparticles also have an impact on the mechanical performance of the polyeth-
ylene nanocomposites. In samples drawn beyond the yield point (necking), scanning electron
microscopy revealed the presence of cavities adjacent to the nanoparticles. More details about the
cavitation were obtained by X-ray ptychographic tomography. Figure 2.43b, c and e shows the
96 2 Microscopy of Polymers

Fig. 2.40 Photomicrograph

of a polished surface of
short glass fibre reinforced
PA 6,6. Micrograph of Per-
Olof Andersson, light-
weight structures, KTH
Royal Institute of
Technology, Stockholm

Fig. 2.41 Transmission electron micrographs of aluminium oxide nanoparticles: (a) pristine, uncoated nanoparticles,
(b) nanoparticles coated by reaction with octyltriethoxysilane and (c) nanoparticles coated by reaction with octadecyl-
triethoxysilane. From Liu et al. (2015) with permission from Elsevier, UK

presence of low-density regions surrounding some of the nanoparticles. X-ray ptychographic tomog-
raphy is an excellent method to reveal structures into the bulk of a sample (cf. Sect. 2.6).
A more traditional composite, a carbon black-filled rubber (NBR), is shown in Fig. 2.44. Informa-
tion about the effect of biodiesel on the rubber was obtained by scanning electron microscopy on
samples fractured in liquid nitrogen. The fresh polymer composite showed that there was a thin
rubber layer coating on every carbon black particle (Fig. 2.44a), whereas the carbon black particles in
the sample that had been exposed to biodiesel had no such bound rubber layer. This explained the
pronounced decrease in fracture toughness of the NBR samples exposed to biodiesel (Akhlaghi et al.
2016).
2.7 Applications of Microscopy in Polymer Science and Engineering 97

Fig. 2.42 Scanning electron micrographs together with plots of the distribution of interparticle distance of fracture
surfaces of specimens cooled to liquid nitrogen temperature of the following polyethylene nanocomposites: (a) pristine
(uncoated) Al2O3 nanoparticles (3 wt.% in composite); (b) Al2O3 nanoparticles coated by reaction with
octyltriethoxysilane (3 wt.% in composite). Pictures a and b are from Liu et al. (2015) with permission from Elsevier,
UK. (c) Large agglomerate of nanoparticles shown in a nanocomposite, which contains 3 wt.% pristine (uncoated)
Al2O3 nanoparticles. From Li, Hillborg and Gedde (2015) with permission from IEEE
98 2 Microscopy of Polymers

Fig. 2.43 (a) Three-dimensional view (by X-ray ptychographic tomography) of a strained section of a polyethylene
nanocomposite containing 3 wt.% aluminium nanoparticles coated by reaction with octyltriethoxysilane. The blue parts
indicate the aluminium oxide nanoparticles and the red parts low-density regions formed by stress-induced cavitation.
The inset figure shows the samples studied, generated by FIB-SEM. (b) and (c) show magnified details of nanoparticles
surrounded by low-density material generated by cavitation. (d) A slice through a nanoparticle which generated (e) a
density vs. x diagram. Note the low-density region surrounding the nanoparticle. From Liu et al. (2017) with permission
from Elsevier, UK

Fig. 2.44 Scanning electron micrographs of freeze-fractured NBR samples containing carbon black (spherical
particles) (a) fresh (unexposed) sample; (b) sample after exposure to biodiesel at 80  C for 34 days. Courtesy of
Shahin Akhlaghi and Amir Pourrahimi, Fibre and Polymer Technology, KTH Royal Institute of Technology,
Stockholm
2.8 Summary 99

2.7.5 Native Polymers and Polymeric Biomaterials

Biological samples were among the first to be studied by optical microscopy during the eighteenth
century. The introduction of electron microscopy and more lately atomic force microscopy has
provided unique tools to study the complex biological samples including many native polymers.
The biological samples have many features in common with polymers: (i) they are sensitive to the
electron beam; (ii) they have low contrast, and in order to reveal true structures, staining is required;
and (iii) many of them are soft matter. Recommended texts on microscopy on biomaterials and
biology are Dombrowski (2003), Dykstra and Reuss (2003), Hibbs (2004) and Cai (2018).

2.8 Summary

Optical microscopy, transmission electron microscopy, scanning electron microscopy and atomic
force microscopy are powerful techniques for the study of the morphology of complex polymer
systems such as semicrystalline polymers, liquid crystalline polymers, polymer blends and polymer
composites. The list of applications in other areas of polymer science and technology is extensive.
Optical microscopy can resolve features in a specimen as small as 1 μm, but the depth of field is
relatively small, making the assessment of topographical features difficult in many cases. Sample
preparation is generally simple. Variations in absorption coefficient, sample thickness, refractive
index and birefringence can be converted to contrast (light intensity) in the final image with the
optical microscopy techniques available. Polarized microscopy, phase-contrast microscopy and
differential interference-contrast microscopy convert differences in optical path (sample thickness,
refractive index and birefringence) to variations in light intensity. Assessments of superstructures,
such as spherulites, and axialites of semicrystalline polymers and mesomorphic structures (such as
nematic and smectic) of liquid crystalline polymers are important applications of optical microscopy.
Scanning electron microscopy produces detailed topographical images by recording the scattered
secondary electrons. The resolution is typically between 10 and 30 nm. The depth of field is very
large, and sharp images can be obtained for specimens with large topographical variations. Informa-
tion about morphology is obtained from a topographical analysis of fracture surfaces and etched
specimens. The sample preparation is relatively simple. All polymeric samples have to be coated with
a thin layer of a conductive material prior to examination in the scanning electron microscope. Some
scanning electron microscopes are equipped with an X-ray microanalyzer, and these instruments can
detect additives and particles containing heavier elements.
Transmission electron microscopy produces detailed images of the density variation and topo-
graphical variations in bulk samples. The resolution is below 1 nm and the depth of field is large, but
the sample preparation is difficult. The naturally occurring density variations in polymers are
generally small, and, in order to achieve contrast, it is necessary to add heavier elements selectively
to one of the phases. This procedure is referred to as staining. Osmium tetroxide, adding to unsatu-
rated polymers, and chlorosulphonic acid, adding selectively to the amorphous component of
polyethylene, are two important staining chemicals. Another preparation technique used on
specimens for transmission electron microscopy is etching, which may further be divided into solvent
etching and chemical etching. Olley and Bassett invented etching with permanganic acid, a mixture
of concentrated sulphuric acid, orthophosphoric acid and potassium permanganate. This mixture has
proven to be a most useful etching agent for a great variety of semicrystalline polymers. The replica
technique makes it possible to assess the topography of the etched samples.
Atomic force microscopy and related techniques have become very popular during the past
5 years. The surface profile can be obtained down to atomic resolution without the need for difficult
100 2 Microscopy of Polymers

and time-consuming specimen preparation. Another attractive feature of atomic force microscopy is
that it can be applied to both solid and liquid samples. Atomic force microscopy and related
techniques have been developed from the pure imaging of the surface profile towards the assessment
of local properties and more recently in situ studies of physical and chemical processes in polymers.

2.9 Exercises

2.1. Suggest suitable methods for determining the spherulite structure of two polyethylene samples.
One is known to have large spherulites and the other to have very small spherulites.
2.2. How can the sign of spherulites be determined by polarized microscopy?
2.3. Suggest suitable staining or etching methods for the following polymer blends: PVC/NR, ABS,
PE/PS, PET/LCP and PMMA/PVC.
2.4. Suggest another method (based on microscopy) for assessing miscibility in the polymer blends
of Question 2.3.
2.5. Compare the two scanning electron micrographs shown in Fig. 2.37. What information can be
obtained from the two images?
2.6. Discuss how to determine by microscopy whether a polymer in a polymer blend is a disperse or
continuous phase.
2.7. The lamellar structure of polyethylene can be studied after chlorosulphonation. What is the
typical view on the micrographs?
2.8. Suggest a method to study the overall shape of the crystal lamellae in melt-crystallized
polyethylene.
2.9. Explain the term ‘dark field’ in optical microscopy and electron microscopy.
2.10. Fractography, to study a fracture surface, can be carried out using various microscopy
techniques. Which is the preferred method? Motivate!
2.11. How would you measure the linear growth rate of spherulites in isotactic polypropylene?
Suggest a second method.
2.12. Estimate the fibre content (in volume percent) in the short-fibre composite shown in Fig. 2.40.
2.13. Identify disclinations of strengths 1/2 in the polarized photomicrograph shown in Fig. 2.33a.

References

Abbe, E. (1873). Archiv für Mikroskopische Anatomie, 9, 413.

Abbe, E. (1874). Jenaische Zeitschrift für Naturwissenschaft, 8, 96.
Akhlaghi, S., Pourrahimi, A. M., Hedenqvist, M. S., Sj€ostedt, C., Bellander, M., & Gedde, U. W. (2016). Polymer
Degradation and Stability, 123, 69.
Andrews, E. H. (1964). Proceedings of the Royal Society (London), A227, 562.
Andrews, E. H., & Stubbs, J. M. (1964). Journal of the Royal Microscopy Society, 82, 221.
Bassett, D. C., Frank, F. C., & Keller, A. (1963). Philosophical Magazine, 8, 1739.
Bassett, D. C., Olley, R. H., & Al Raheil, I. A. M. (1988). Polymer, 29, 1539.
Baumel, H. B., & Berger, J. J. (1973). Biology . . . its people and papers. Washington, DC: National Science Teachers
Association.
Binnig, G., Rohrer, H., Gerber, C., & Weibe, E. (1982a). Applied Physics Letters, 40, 178.
Binnig, G., Rohrer, H., Gerber, C., & Weibe, E. (1982b). Physical Review Letters, 49, 57.
Binnig, G., Quate, C. F., & Gerber, C. (1986). Physical Review Letters, 56, 930.
Borgani, R., Forschheimer, D., Bergqvist, J., Thorén, P.-A., Ingan€as, O., & Haviland, D. B. (2014). Applied Physics
Letters, 105, 141113.
Borgani, R., Pallon, L. K. H., Hedenqvist, M. S., Gedde, U. W., & Haviland, D. B. (2016). Nano Letters, 16, 5934.
References 101

Busch, H. (1926). Annalen der Physik, 81, 974.

Cai, J. (Ed.). (2018). Atomic force microscopy in molecular and cell biology. Singapore: Springer Nature Singapore.
Cappella, B. (2016). Mechanical properties of polymers measured through AFM force-distance curves. Cham: Springer
Nature.
Curson, A. D. (1989). Specimen preparation. In D. A. Hemsley (Ed.), Applied polymer light microscopy.
London/New York: Elsevier Applied Science Publishers.
Darlington, M. W., McGinley, P. L., & Smith, P. F. (1976). Journal of Materials Science, 11, 877.
De Broglie, L.-W. (1925). Ph.D. thesis: On the theory of Quanta, Annales de Physique, 10 Series, Tome 3. Paris:
Masson & Co.
Demus, D., & Richter, L. (1978). Textures of liquid crystals. Leipzig: VEB Deutscher Verlag für Grundstoffindustrie.
Diaz, A., Trtik, P., Guizar-Sacairos, M., Menzel, A., Thibault, P., & Bunk, O. (2012). Physics reviews B: Condensed
matter and material. Physics, 85, 020104.
Dierolf, M., Menzel, A., Thibault, P., Schneider, P., Kewish, C. M., Wepf, R., Bunk, O., & Pfeiffer, F. (2010). Nature,
467, 436.
Dombrowski, R. T. (2003). Microscopy techniques for analyzing the phase nature and morphology of biomaterials. In
M. Jaffe, W. Hammond, P. Tolias, & T. Arinzen (Eds.), Characterization of biomaterials (2nd ed.). Cambridge:
Woodhead Publisher.
Dykstra, M. J., & Reuss, L. E. (2003). Biological electron microscopy: Theory, techniques and trouble shooting (2nd
ed.). New York: Springer.
Engberg, K., Ekblad, M., Werner, P.-E., & Gedde, U. W. (1994). Polymer Engineering and Science, 34, 1346.
Faulkner, H. M. L., & Rodenburg, J. M. (2004). Physics Reviews Letters, 93, 023903.
Feinstein, S. (2008). Louis Pasteur: The father of microbiology. Berkeley Heights: Enslow Publishers.
Frisbie, D., Rozsnyai Noy, A., Wrighton, M. S., & Lieber, C. M. (1994). Science, 265, 2071.
Gedde, U. W., & Hedenqvist, M. S. (2019a). Morphology of semicrystalline polymers (Chap. 7). In Fundamental
polymer science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Chain orientation (Chap. 9). In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019c). Introduction to polymer science (Chap. 1). In Fundamental polymer
science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019d). Polymer solutions (Chap. 4). In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W., & Jansson, J.-F. (1984). Polymer, 25, 1263.
Gedde, U. W., Eklund, S., & Jansson, J.-F. (1983). Polymer, 24, 532.
Gustafsson, A., Salot, R., & Gedde, U. W. (1993). Polymer Composites, 14, 421.
Harding, G. F. (1986). Double refraction in polymers: Measurements and interpretation. In G. H. Meeten (Ed.), Optical
properties of polymers. London/New York: Elsevier, Applied Science Publishers.
Hedmark, P. G., Jansson, J.-F., Hult, A., Lindberg, H., & Gedde, U. W. (1987). Journal of Applied Polymer Science, 34,
743.
Hedmark, P. G., Rego Lopez, J. M., Westdahl, M., Werner, P.-E., Jansson, J.-F., & Gedde, U. W. (1988). Polymer
Engineering and Science, 30, 1248.
Hemsley, D. A. (Ed.). (1989a). Applied polymer light microscopy. London/New York: Elsevier Applied Science
Publishers.
Hemsley, D. A. (1989b). Basic light microscopy and the phase contrast microscope. In D. A. Hemsley (Ed.), Applied
polymer light microscopy. London/New York: Elsevier, Applied Science Publishers.
Hibbs, A. R. (2004). Confocal microscopy for biologists. New York: Springer.
Hill, M. J., Bradshaw, D. G., & Chevili, R. J. (1992). Polymer, 33, 874.
Hillenbrand, R., Stark, M., & Guckenberger, R. (2000). Applied Physics Letters, 76, 3478.
Hoffman, R. (1989). Modulation contrast and differential interference contrast techniques. In D. A. Hemsley (Ed.),
Applied polymer light microscopy. Elsevier Applied Science: London/New York.
Holler, M., Diaz, A., Guizar-Sicairos, M., Karvonen, P., F€arm, E., H€ark€
onen, E., Ritala, M., Menzel, A., Raabe, J., &
Bunk, O. (2014). Scientific Reports, 4, 3857.
Hudson, S. D., Lovinger, A. J., Venkataraman, S. K., Liu, C., & Manzione, L. T. (1994). Polymer Engineering and
Science, 34, 1327.
Jamayo, J., & Garcia, R. (1997). Applied Physics Letters, 71, 2394.
Kalnins, M., Conde Braña, M. T., & Gedde, U. W. (1992). Polymer Testing, 11, 139.
Kanig, G. (1973). Kolloid Zeitschrifte und Zeitschrifte für Polymere, 251, 782.
Kato, K. (1965). Journal of Electron Microscopy, 14, 20.
Kato, K. (1966). Journal of Polymer Science, Polymer Letters, 4, 35.
Kato, K. (1967). Polymer Engineering and Science, 7, 38.
102 2 Microscopy of Polymers

Keller, A. (1957). Philosophical Magazine, 2, 1171.

Knoll, M. (1935). Zeitschrift für Technische Physik, 16, 407.
Knoll, M., & Ruska, E. (1932). Zeitschrift für Physik, 78, 318.
K€ohler, A. (1893). Zeitschrift für Wissenschaftliche Mikroskopie und für Mikroskopische Technik, 10, 433.
Li, W., Hillborg, H., & Gedde, U. W. (2015). IEEE Transaction on Dielectrics and Electrical Insulation, 22, 3536.
Liu, D., Pourrahimi, A. M., Olsson, R. T., Hedenqvist, M. S., & Gedde, U. W. (2015). European Polymer Journal, 66,
67.
Liu, D., Pallon, L. K. H., Pourrahimi, A. M., Zhang, P., Diaz, A., Holler, M., Schneider, K., Olsson, R. T., Hedenqvist,
M. S., Yu, S., & Gedde, U. W. (2017). European Polymer Journal, 87, 255.
Locke, J. (1824). The work of John Locke (12th ed.). London: C. and J. Rivington.
Majumdar, A. (1999). Annual Review of Material Science, 29, 505.
Majumdar, A., Carrejo, J. P., & Lai, J. (1993). Applied Physics Letters, 62, 2501.
McCrone, W. C., McCrone, L. B., & Delly, J. G. (1980). Polarized light microscopy. Ann Arbor: Ann Arbor Science
Publisher.
McMullan, D. (1952). Investigation relating to the design of electron microscope, Ph.D. Thesis (supervisor: C. W.
Oatlay). University of Cambridge, UK.
Nisman, R., Smith, P., & Vancso, G. J. (1994). Langmuir, 10, 1669.
Nomarski, G. (1952). Interferometre a Polarisation. Brevet Francais No. 1 059 123, 1952.
Nomarski, G. (1955). Journal de Physique et la Radium, 16, S9.
Noy, A., Vezenov, D. V., & Lieber, C. M. (1997). Annual Reviews of Material Science, 27, 381.
Núñez, E., & Gedde, U. W. (2005). Polymer, 46, 5992.
Núñez, E., Ferrando, C., Malmstr€om, E., Claesson, H., & Gedde, U. W. (2004). Journal of Macromolecular Science,
Part B. Physics, 43, 1143.
Olley, R. H., & Bassett, D. C. (1982). Polymer, 23, 1707.
Pallon, L. K. H., Nilsson, F., Yu, S., Liu, D., Diaz, A., Holler, M., Chen, X. R., Gubanski, S., Hedenqvist, M. S., Olsson,
R. T., & Gedde, U. W. (2017). Nano Letters, 17, 1402.
Passeri, D., Angeloni, L., Reggente, M., & Rossi, M. (2017). Magnetic force microscopy. In C. S. S. R. Kumar (Ed.),
Magnetic characterization techniques for Nanomaterials. Berlin/Heidelberg: Springer.
Patel, G. N. (1980). Chemical methods in polymer physics. In R. A. Fava (Ed.), Methods of experimental physics
(Part B, Polymers: Crystal structure and morphology) (Vol. 16). New York: Academic Press.
Pearce, R., & Vancso, G. J. (1997). Macromolecules, 30, 5843.
Pearce, R., & Vancso, G. J. (1998). Polymer, 39, 6743.
Platz, D., Tholén, E. A., Pesen, D., & Haviland, D. B. (2008). Applied Physics Letters, 92, 153106.
Pollock, H. M., & Hammiche, A. (2001). Journal of Physics D: Applied Physics, 34, R23.
Pylkki, R., Moyer, P., & West, P. (1994). Japanese Journal of Applied Physics, 33, 3785.
Reid, N. (1975). Ultramicrotomy. In A. M. Glaubert (Ed.), Practical methods in electron microscopy. Amsterdam:
Elsevier.
Rodenburg, J. M., Hurst, A. C., Cullis, A. G., Dobson, B. R., Pfeiffer, E., Bunk, O., David, C., Jefimovs, K., & Johnson,
I. (2007). Physics Review Letters, 98, 034801.
Ruska, E. (1980). The early development of electron lenses and electron microscopy. Stuttgart: S. Hirzel Verlag.
Sarid, D. (1991). Scanning force microscopy. Oxford: Oxford University Press.
Sawyer, L. C., & Grubb, D. T. (1996). Polymer microscopy (2nd ed.). London: Chapman & Hall.
Sawyer, L. C., Grubb, D. T., & Meyers, G. F. (2008). Polymer microscopy (3rd ed.). New York: Springer.
Sch€onherr, H., & Vancso, G. J. (1997). Macromolecules, 30, 6391.
Sch€onherr, H., & Vancso, G. J. (2010). Scanning force microscopy of polymers. Berlin/Heidelberg: Springer.
Sch€onherr, H., Hruska, Z., & Vancso, G. J. (1998). Macromolecules, 31, 3679.
Smith, P. F., Nisman, R., Ng, C., & Vancso, G. J. (1994). Polymer Bulletin, 33, 459.
Strutt, J. W. (3rd Baron Rayleigh) (1879). Philosophical Magazine, 8, 261.
Thorén, P.-A., Borgani, R., Forchheimer, D., Dobryden, I., Claesson, P. M., Kassa, H. G., Leclère, P., Wang, Y., Jaeger,
H. M., & Haviland, D. B. (2018). Physical Review Applied, 10, 024017.
Toporski, J., Dieng, T., & Hollricher, O. (Eds.). (2018). Confocal Raman microscopy (Springer series on surface
sciences 66) (2nd ed.). Cham: Springer.
Tsukruk, V. V., Bliznyuk, V. N., & Wu, Y. (1998). ACS Symposium Series, 694, 321.
Turner, G. L.’. E. (2012). Animadversions on the origin of the microscope. In J. D. North & J. J. Roche (Eds.), The light
of nature: Essays in the history and philosophy of science presented to A. C. Crombie. Berlin: Springer.
Vadimsky, R. G. (1980). Electron microscopy. In R. A. Fava (Ed.), Methods of experimental physics (Part B: Polymers:
Crystal structure and morphology) (Vol. 16). New York: Academic Press.
Vancso, G. J., Hillborg, H., & Sch€onherr, H. (2005). Advances in Polymer Science, 182, 55.
References 103

Vaughan, A. S. (1993). Polymer microscopy. In B. J. Hunt & M. I. James (Eds.), Polymer characterisation. London:
Blackie Academic.
Voigtl€ander, B. (2015). Scanning probe microscopy: Atomic force microscopy and scanning tunneling microscopy.
Berlin/Heidelberg: Springer.
Voigt-Martin, I. G., & Mandelkern, L. (1989). Journal of Polymer Science, Polymer Physics Edition, 27, 967.
von Ardenne, M. (1938a). Zeitschrift für Physik, 109, 553.
von Ardenne, M. (1938b). Zeitschrift für Technische Physik, 19, 407.
White, H. M., & Bassett, D. C. (1998). Polymer, 39, 3211.
Williams, C. C., & Wickramasinghe, K. (1986). Applied Physics Letters, 49, 1587.
Wittmann, J. C., & Lotz, B. (1985). Journal of Polymer Science, Polymer Physics Edition, 23, 205.
Woodward, A. E. (1989). Atlas of polymer morphology. Munich/New York: Hanser.
Zernike, F. (1942a). Physica, 9, 686.
Zernike, F. (1942b). Physica, 9, 974.
Zernike, F. (1955). Science, 121, 345.
Zhang, Y., Zhu, W., Hui, F., Lanza, M., & Muñoz Rojo, M. (2019). Advanced Functional Materials, 30, 1900892.
Chapter 3
Spectroscopy and Scattering of Radiation by Polymers

3.1 Introduction

Spectroscopy and scattering methods have always been important in polymer science and engineer-
ing, and they constitute wide-ranging groups consisting of a variety of individual methods, and the
objectives and the outcomes of the studies based on these methods are also diversified. The main
spectroscopic methods include vibrational spectroscopy (infrared (IR) and Raman spectroscopy),
nuclear magnetic resonance (NMR) spectroscopy and a number of other techniques such as electron
spin resonance (ESR) spectroscopy, ultraviolet (UV) spectroscopy, visible light (VIS) spectroscopy
and X-ray photoelectron (XPS) spectroscopy. Spectroscopy is defined as the study of the interaction
between electromagnetic radiation and matter over a defined spectral range. The absorption of
radiation in two different spectral regions is shown in Fig. 3.3b and c. In the case of IR radiation,
absorption occurs when the frequency of radiation is the same as the frequency of a vibration
(cf. Sect. 3.2.1). Some of the components of visible light are absorbed every time you look at a
coloured object. Raman spectroscopy is dependent on the exchange of radiation energy and vibra-
tional energy resulting in inelastic scattering (Fig. 3.3c). NMR and ESR spectroscopy involve the
simultaneous application of a magnetic field. The wavelength (λ) of the radiation used in the
spectroscopy family varies by 11 orders of magnitude from 0.1 nm to 10 m (Fig. 3.1). This broad
wavelength band corresponds also to a large variation in the energy per photon (U ): U ¼ hc/λ, where
h is Planck’s constant and c is the velocity of the radiation. The relationship between photon energy
and wavelength is U ¼ 1.2410–6/λ, where U is given in eV and λ in meter. In XPS, U  1240
eV/photon  120 000 kJ mol–1 (1 eV  96.5 kJ mol–1), which is much greater (240 times greater) than
the dissociation energy of a covalent bond, which is typically 500 kJ mol–1. UV radiation covers a
wide range of energy, from 400 to 12 000 kJ mol–1, whereas the energy of visible light with
wavelengths from 380 to 740 nm ranges from 315 to 160 kJ mol–1. IR covers a range between 0.75
μm and 1000 μm, but conventional IR spectrophotometers cover the mid-IR range (2.5–50 μm),
which corresponds to moderate of energies, 2.4–48 kJ mol–1. IR spectra can be recorded in the near-
IR range (0.75–2.5 μm) using most UV-VIS spectrophotometers. ESR and NMR instruments operate
with very low energies of radiation: 0.03 m (410–3 kJ mol–1) for ESR and 1–10 m (110–4–110–5 kJ
mol–1) for NMR.
The spectroscopic methods provide detailed information about the chemical structure such as the
repeating unit structure, the local distribution of repeating units in copolymers and the tacticity. The
presence of distinct functional groups, such as carbonyl groups of different types, is revealed. Some
information about molecular architecture is obtained, in particular about short-chain branching.
Spectroscopy also provides data about the physical structure, e.g. chain orientation (cf. Gedde and
Hedenqvist 2019a), segmental mobility and crystallinity. Some of the methods provide data with

# Springer Nature Switzerland AG 2021 105

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_3
106 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.1 Spectral

wavelength range of
different spectroscopy
methods

Fig. 3.2 The early development/progress within different spectroscopic methods

spatial resolution, e.g. IR and Raman microscopy and magnetic resonance imaging (MRI). The
chemical structure of the immediate surface structure is revealed by XPS. ESR detects free radicals,
and it can be used to study polymerization; degradation induced by, e.g. oxidation or mechanical
stress; and molecular mobility. NMR is an extremely powerful tool which can reveal structure at
different levels (from repeating units to morphological features) and molecular mobility on different
length and time scales. This means that the areas of application are extremely versatile, as pointed out
by Spiess (2017a, b) in a recent review: materials, micro-imaging, toxicology, body fluids, food,
membranes, nanogram spectroscopy, screening applications, high throughput and proteomics. Mod-
ern technology has made spectrometers affordable and available for a great many companies and
institutions. Figure 3.2 presents a brief account of the early progress of the different spectroscopic
families. Noteworthy is that ten of the displayed scientists mentioned are Nobel Laureates, which
3.1 Introduction 107

Fig. 3.3 Examples of interaction between radiation and matter: (a) refraction; a beam changes direction when it enters
a medium with a different refractive index; (b) absorption of IR by transfer of energy to vibrations, normal modes; (c)
elastic and inelastic scattering from two particles; note the change of radiation energy between the beams in the inelastic
case; (d) absorption of radiation in the UV/visible light spectral range where a chromatophoric molecule which absorbs
in this spectral range is shown; (e) a circular electromagnetic wave emerging from the scattering point (molecule); (f)
interference between two scattering points emitting synchronized circular waves with the directions of the waves of
constructive interference is displayed by the three vectors

indeed indicates the importance of spectroscopy as a means to achieve results for the benefit of
humanity.
The scattering methods use various types of radiation, which include visible light, UV radiation,
X-rays and neutrons. The basic mathematics is the same for all types of scattering. Scattering is
confined to the wave character of electromagnetic radiation: visible light, X-ray, electrons and
neutrons. The motion of waves is one of two different models for the transport of energy from one
spatial point to another, the other being material transport. The propagation of waves (e.g. light or
acoustic waves) involves no material transport. For instance, the motion of a wave on a water surface
involves motion of water molecules in the direction of the field of gravity, but no transport of water
molecules along the direction of motion of the wave. Waves can be mechanical (requiring the
presence of a medium) or electromagnetic (requiring no medium). A characteristics of waves,
expressed by the Huygens principle (after the Dutch physicist Christiaan Huygens; 1678), is that
they can go around corners (obstacles), bending or diffracting, which is not possible for material
transport. In contrast, transport by diffusion through condensed matter occurs by a random walk
mechanism, where the motion of the diffusing molecules occurs as the result of a series of short
random jumps (cf. Chap. 7). This makes possible the spread of matter in any direction along a
concentration gradient.
The wavelengths of the radiation types used vary between 0.1 nm and 1000 nm (approximate
values), i.e. a variation by four orders of magnitude. Thermal neutrons typically have a wavelength of
0.45 nm. Scattering methods use monochromatic radiation, and the scattered radiation can either have
the same wavelength as the incoming radiation, which is denoted elastic scattering, or a longer
wavelength which is denoted inelastic scattering (Fig. 3.3c). Elastic scattering is simple, and as
indicated by Fig. 3.3d, the scattered wave is circular emerging from the centre of the scattering point,
and coherent radiation is diffracted by the two adjacent scattering points producing areas of
108 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.4 Some key events in the early development of scattering methods

constructive interference (Fig. 3.3f). Light scattering (wavelength: 380–740 nm) assesses the molar
mass and radius of gyration of polymer molecules in dilute solution (cf. Gedde and Hedenqvist
2019b). X-ray scattering uses radiation of very short wavelength (ca. 0.1 nm) and assesses crystal unit
cells with a side length of less than a nm by recording the diffraction pattern at high scattering angles
(wide-angle X-ray scattering, WAXS). Small-angle X-ray scattering (SAXS) can assess structures
with repeating distances ranging from 5 to 50 nm, e.g. stacks of crystal lamellae in melt-crystallized
polymers (cf. Gedde and Hedenqvist 2019c). Small-angle neutron scattering in special chemically
labelled sample solutions provides information about the radius of gyration of polymer molecules
including solid polymers such as those with a semicrystalline structure.
This early history of diffraction, i.e. the non-linear propagation of a wave passing a small opening,
an edge or around both sides of a small obstacle, started with Huygens’ observations of the propaga-
tion of water waves (Fig. 3.4). He noticed that a series of planar water waves meeting a small opening
gives rise a series of secondary scattered waves with a semi-circular shape. The opening acts as a
sender of circular waves provided that the opening is much smaller than the wavelength. The
Huygens’ principle based on these observations was formulated as each point which is struck by a
wave front becomes a centre for a secondary (scattered) wave. The envelope of this secondary wave
yields the wave front at a later time. Huygens assumed that the wave was fully circular with one part
moving in the forward direction and the other propagating backwards. This assumption was essen-
tially not validated by any observation. Fresnel, who continued the work much later, dealt with this
problem empirically and assumed that the amplitude of the secondary wave decreases with increasing
angle between the directions of propagation of the primary and secondary waves. This assumption
was validated by Kirchhoff based on the fundamental wave equation, which is equally valid for both
mechanical and electromagnetic waves. Strutt explained the blue sky by elastic light scattering
(referred to as Rayleigh scattering) of the small gas molecules (primarily N2) which are much smaller
3.2 Vibrational Spectroscopy 109

(a few Ångstr€om) than the wavelength of the light (~5000 Å). The equation derived by Strutt is
helpful for the interpretation of the light scattering of polymer molecules in dilute solution (cf. Sect.
3.4.2). Strutt showed that the extent of this scattering is dependent on the wavelength (λ) and that it is
proportional to λ–4. This law explains the blue sky in daytime and the red colour of the low sun.
Later come the discovery of X-rays and the understanding of its electromagnetic wave nature and
the interpretation of X-ray diffraction (WAXS) based on crystallography (Ewald, von Laue and Bragg
& Bragg). SAXS was coined 25 years later by Guinier, and this made possible the study of colloidal
systems and the assessment of the long period of stacks of crystal lamellae in semicrystalline
polymers. WAXS makes possible the most direct determination of crystallinity (according to Ruland)
and also the crystal thickness (Scherrer), but SAXS provides an even more reliable determination of
crystal thickness. The discovery of synchrotron radiation in 1949 made possible some year later the
use of this type of intense radiation for WAXS and SAXS. This new era started around 1970, and
diffraction moved to some extent from small university departments to very large facilities. Neutron
scattering requires an advanced infrastructure (the generation of a stream of neutrons), and it started
in the early 1970s. By using labelled polymer molecules, hydrogen being replaced by deuterium, the
radius of gyration of polymer molecules can be determined in the condensed state (e.g. rubbery,
glassy and semicrystalline states). The Flory theorem, for instance, was confirmed by small-angle
neutron scattering (cf. Gedde and Hedenqvist 2019b).

3.2 Vibrational Spectroscopy

3.2.1 Fundamentals

The atoms in a molecule vibrate leading to changes in bond length, bond angle or torsion angle. The
molecule consists of a set of harmonic oscillators, analogous to classical mechanics, where the
frequency of the oscillation is proportional to the square root of the force constant and inversely
proportional to the square root of the reduced mass, the latter being equal to m1m2/(m1+m2), where
m1 and m2 are the masses of the objects which are connected via the spring. This implies that the
frequency of a vibration involving heavy elements is low and that if the force constant for the
oscillation is high, as in a case of a simple stretching of a covalent bond, then the frequency of the
oscillation is high. The disturbance of a molecule from its equilibrium causes a motion which is a
combination of a number of simple harmonic vibrations referred to as normal modes. The frequency
of the motion of the atoms in each of the normal modes is the same, and all the atoms pass the zero
position simultaneously.
Certain frequencies of infrared radiation are absorbed, and the energy is transferred to a particular
vibration, which may in turn be either transferred to the surrounding molecules, i.e. converted to heat,
or re-emitted as radiation. The principle of infrared spectroscopy is thus that the absorption
frequencies correspond to the frequencies of the normal modes. It should be noted that not all the
normal modes are infrared active. An oscillating dipole can absorb the infrared light only if the
frequencies are matching. The intensity of the absorption (I) is proportional to the square of the
change in dipole moment (μ):

2
dμ
I/ ð3:1Þ
dQ

where Q is the displacement coordinate of the motion. Some vibrations of groups of atoms are not
infrared active even though that the individual bonds participating in the vibration are dipoles. The
symmetry of the vibrations in these cases counteracts the change in dipole moment associated with
110 3 Spectroscopy and Scattering of Radiation by Polymers

the individual bonds (cf. Gedde and Hedenqvist 2019e). The net change in dipole moment of the
whole group becomes zero, and hence the vibration is not infrared active. An additional requirement
for infrared absorption originates from the vector character of the transition moment (M) and the
electrical field vector (E) associated with the infrared radiation:

I / ðE  MÞ2 ¼ ðE  M cos ðθÞÞ2 ð3:2Þ

where θ is the angle between the vectors E and M. The transition moment vector is oriented along the
net change of dipole moment associated with the vibration. The magnitude of absorption of the
radiation depends on the state of polarization of the incoming radiation if the polymer chains have a
preferential orientation. This is the basis of infrared dichroism measurements, a method used to assess
chain orientation in polymers (cf. Gedde and Hedenqvist 2019a).
The harmonic vibrations, which are the common vibrations of molecules at room temperature, are
within the spectral range (λ ¼ 2.5–25 μm) of mid-infrared (MIR) spectroscopy (Siesler 2011, Bec´
and Huck 2019). Anharmonic vibrations occur if the vibrational amplitudes are extensive, especially
if the masses of the connected two atoms are very different. These vibrations give rise to overtones
that appear in a spectral range corresponding to λ ¼ 0.6–2.5 μm, and they are studied by near-infrared
(NIR) spectroscopy. In this spectral range, combinations of vibrational transitions/modes are also
active. The probability of overtones and vibrational combination transitions is generally low, which
leads to a low IR absorption, and larger samples can thus be analyzed which simplifies the sample
preparation for NIR spectroscopy. In addition, the low absorption in the NIR region makes it possible
to analyze aqueous solutions. The specificity is nevertheless significantly lower in NIR spectroscopy
than in conventional IR spectroscopy, due to the fact that the absorption bands originate from
overlapping overtones, combination transitions and to a requirement of large difference in mass for
the anharmonic oscillations, the NIR region mainly showing features of X–H groups (C–H, O–H and
N–H). Far-infrared (FIR) spectroscopy (20–100 < λ < 1000 μm) has proven useful for, e.g. the
analysis of collective motions in networks based on inter- and intramolecular hydrogen bonds
(El Khoury and Hellwig 2017).
Irradiation of matter causes scattering including both elastic scattering (Rayleigh scattering) and
inelastic scattering such as Raman scattering. Only a few photons, typically one out of a total of 108,
undergo Raman scattering. The radiation energy is transferred to vibrations (normal modes), and the
change in frequency from that of the illuminating radiation is equal to the frequency of the normal
mode. The strongest inelastic scattering (Stokes Raman scattering) has a frequency that is lower than
that of the frequency of the incoming radiation. Scattered radiation with a higher frequency than that
of the incoming light is called anti-Stokes Raman scattering. Not all of the normal modes are Raman
active. The requirement for Raman scattering is that the normal mode involves a change in polariz-
ability (cf. Gedde and Hedenqvist 2019e), and it is necessary to illuminate the sample with highly
monochromatic (laser) light in order to detect the low-intensity Raman scattering. For a molecule
with n atoms, there are 3n – 6 vibrational degrees of freedom. If this principle held for polymers, there
would be an enormous number of vibrational modes, but fortunately most of the vibrations are the
same. The IR spectrum of a polymer is in practical terms independent of molar mass. Figure 3.5
shows some of the vibrational modes of polyethylene.
The Raman shift in frequency corresponds in many cases to the IR absorption spectrum, but some
vibrational modes appear only in the IR spectrum, while other modes appear only in the Raman
spectrum. Bands with strong IR absorption originate from highly polar groups, whereas strong Raman
scattering arises from highly polarizable groups. The strongest absorption bands in the IR spectrum of
poly(ethylene terephthalate) are due to the carbonyl groups, whereas the strongest peaks in the Raman
spectrum is due to the stretch of the carbon–carbon bonds in the phenylene groups.
Figure 3.6 presents a condensed summary of the absorption bands (group frequencies) associated
with important molecular groups. This scheme is useful to remember to facilitate chemical analysis.
3.2 Vibrational Spectroscopy 111

Fig. 3.5 Vibrational

modes in polyethylene

Fig. 3.6 Group frequencies. Drawn after Bower and Maddams (1989)
112 3 Spectroscopy and Scattering of Radiation by Polymers

3.2.2 IR and Raman Spectrophotometers

Current IR instruments are either a dispersive instrument or a Fourier transform IR spectrometer.

Both techniques use a polychromatic light source, giving a continuous spectrum. The light source for
MIR is usually a bar-shaped thermal light source (λ ¼ 4–15 μm) made of silicon carbide referred to as
Globar (glow bar). For near-IR a tungsten halogen lamp is used. Far-IR radiation is given by Globar,
but for the lowest wavenumber range, mercury-vapour lamps are better and can be added to a
standard MIR/FIR instrument. Synchrotron light and quantum cascade lasers are two other light
sources which are favourable when the highest requirements have to be met (El Khoury and Hellwig
2017; Wrobel and Bhargava 2018). In the dispersive technique, a grating is used to disperse the IR
radiation geometrically. A scanning mechanism passes the dispersed radiation over a slit system and
isolates the frequency falling on the detector.
Fourier transform infrared (FTIR) spectrometers use a Michelson interferometer to provide a
spectrum. From the interferogram (I(x)), the frequency-domain spectrum (G(υ)) is obtained by a
mathematical procedure (Fourier transform) expressed according to:

ð
þx

GðυÞ ¼ I ðxÞ cos ð2πυxÞ dx ð3:3Þ

x

In the mid-IR range (energy-limited region), the FTIR spectrometer is clearly advantageous over a
dispersive IR spectrometer and is today the main technique being used. It is more time-efficient and
has a higher signal-to-noise ratio and greater accuracy in this spectral region. However, spectrometers
operating in the NIR range are usually grating spectrometers which are more efficient and more
accurate than Fourier transform instruments, especially for the analysis of clear liquids and gases.
Infrared spectroscopy is sometimes performed in the transmission mode, but several reflection
techniques are available. Internal reflection spectroscopy (IRS), which is also called attenuated total
reflection (ATR), is a useful technique for analysing samples with low transmission, which are thus
difficult to study in the transmission mode. ATR-IR spectroscopy can also be considered as a surface
characterization method, because the assessed material is that in the top few micrometres of the
specimen. This technique is today the most frequently used mid-IR technique because of its flexibility
and convenient sample preparation. Figure 3.7 shows the arrangement of the sample and how the IR
radiation is reflected many times from the surfaces. In order for the ATR crystal to work as a
waveguide, it has to be IR transparent and have a refractive index higher than that of the surroundings.
The sampling depth (d ) into the specimen is:

Fig. 3.7 Schematic

description of the working
principle of ATR-IR
spectroscopy. The central
waveguide is made of,
e.g. thallium bromoiodide
(KRS5) or germanium. For
hard samples and those that
may degrade the normal
waveguide, it is possible to
use a diamond waveguide
3.2 Vibrational Spectroscopy 113

λ
d¼
 2 0:5 ð3:4Þ
2πn1 sin θ 
2 n2
n1

where λ is the wavelength of the radiation in air, n1 is the refractive index of the ATR crystal, n2 is the
refractive index of the sample and θ is the angle of incidence. Typical values for the refractive indices
are n1 ¼ 4.0 (germanium) and n2 ¼ 1,5. The sampling depth is close to λ/10. It should be noted that
the sampling depth depends on the wavelength, which means that the absorption peaks have different
weightings in an ATR spectrum than in the transmission spectrum of the same substance.
There are a number of accessories, interfaces and variants of measuring IR spectroscopy in
addition to transmission and ATR, and IR spectroscopy is also used in combination with other
analytical techniques such as gas chromatography (GC), thermogravimetry equipped with a differen-
tial scanning calorimeter (TG/DSC) and mass spectrometry (MS), as indicated in Chaps. 1 and 4. IR
spectrometers are also available as hand-held devices enabling non-destructive measurements to be
made directly on products in service.
For powder samples and samples that reflect only a little light, a diffuse reflectance interface is
preferred. To measure the absorbance in a thin coating on a reflective metal surface, a specular
reflectance tool is desired where the radiation angle of incidence is 45 , whereas a grazing angle
specular interface is ideal for very thin films and the analysis of submicron surface layers, the angle of
incidence being very high, 82 . In NIR measurements, probes are available that can make
measurements on liquids, powders and solid material several metres away from the main equipment.
Combinations of different analytical techniques assembled in a single unit are now commercially
available, such as TG-IR spectroscopy and GC-MS-IR spectroscopy. These combinations provide a
comprehensive set of data describing complex materials and complex processes such as the degrada-
tion of polymeric material: the temperature dependence of the reaction enthalpy and of the mass, as
well as the identity of the released substances as revealed by MS and IR spectroscopy after GC
separation.
IR imaging is a very useful technique for investigating local variations in the composition of a
sample. With today’s rapid spectral sampling techniques obtaining hundreds of spectra per second
and high-resolution detectors, it is possible to rapidly obtain a ‘map’ with a spatial resolution of a few
μm in both transmission and reflection (ATR) modes.
The challenge of Raman spectroscopy is the weakness of the Raman scattering. Typically only one
of 108 incident photons is inelastically scattered (Butler et al. 2016). Fluorescence and sample heating
due to absorbance of the laser energy may cause a problem. As with IR spectroscopy, the Raman
instruments can be divided into those based on a dispersive/grating technique and those based on a
Fourier transform/interferometer technique. The light sources in current use are lasers within the UV,
visible or NIR spectral ranges. Lasers emitting radiation with a longer wavelength reduce the
fluorescence; the Nd:YAG laser with a NIR wavelength (1064 nm) is one such commonly used
source. A laser emitting longer wavelength radiation, however, leads to a lower Raman intensity
(it scales with the fourth power of the frequency (Siesler 2011)), which in FT-Raman spectroscopy
can be counteracted by adding more scans. To reduce the fluorescence, lasers are used in the deep-UV
region (λ < 260 nm), where the fluorescence is outside the Raman spectral region used (Zhu et al.
2014). The laser beam has to be focused accurately on the specimen surface, and the scattered
radiation is recorded at 90 to the incident beam. Pulsed lasers, rather than continuous wave lasers,
have the advantage of reducing unwanted visible light and long-lived fluorescence. Among the
advantages of FT-Raman compared to dispersive Raman are their better wavelength accuracy and
larger spectral range. The advantage of dispersive systems is that they are faster, because of their short
spectral acquisition times. This is especially the case with the recently developed 2D/double grating
114 3 Spectroscopy and Scattering of Radiation by Polymers

technique. In dispersive Raman spectroscopy, a lower laser power can be used with less risk of
damaging the sample.
Other techniques are available to improve the weak spontaneous Raman scatter, apart from
increasing the frequency of the laser radiation. Techniques with greater sensitivity, spatial resolution
and/or rate of analysis include surface-enhanced Raman spectroscopy (SERS), tip-enhanced Raman
spectroscopy (TERS), resonance Raman spectroscopy (RRS), coherent anti-Stokes Raman spectros-
copy (CARS) and stimulated Raman spectroscopy (SRS). Detailed descriptions of these are given by
Mitsutake et al. (2019), but a few highlights are presented below. In SERS, due to the plasmonic
effect, single-molecule detection levels can be reached. In TERS, the chemical detection level of
SERS is achieved but with a spatial resolution down to the nm-range by the use of an AFM tip. In
RRS, a resonance effect occurring when the laser light has a wavelength close to that of the electronic
absorption band increases the Raman intensity. CARS and SRS are two non-linear Raman techniques
where the signal intensity is non-linearly dependent on the excitation intensity and yield Raman
spectra ultra-rapidly.
As with IR, hand-held Raman devices are available, and it is also possible to record Raman spectra
relatively far from the main equipment using quartz fibre optics.
Imaging techniques are available for Raman spectroscopy. In confocal Raman microscopy, the
laser light is focused on the sample with an optical lens, and the resulting backscattered light is
refocused through a pinhole aperture which removes intensity from these parts of the sample that are
outside the sample region in focus. This permits the analysis of submicron-sized regions as well as the
possibility of obtaining a depth profile of the Raman scatter by refocusing along the thickness of a
semitransparent sample.
Using a scanning probe microscope, it is possible to combine atomic force microscopy with
Raman spectroscopy (cf. TERS). This provides high-resolution information about the chemistry
together with information about characteristics such topography and mechanical properties. Raman
spectroscopy has also been combined with, e.g. laser-induced breakdown spectroscopy (LIBS), laser-
induced fluorescence spectroscopy (LIF), photoluminescence (PL) and rheometry.

3.2.3 Quantitative Analysis

An infrared spectrum is presented in the form of a graph of transmittance (T ) or absorbance (A)

against frequency (υ) or wavenumber (υ in cm-1). These quantities are defined as follows:

  104
υ cm1 ¼ ð3:5Þ
λðμmÞ

I
T¼ ð3:6Þ
I0
 
I0
A ¼ log ð3:7Þ
I

where λ is the wavelength, I is the intensity of the light at depth l in the medium and I0 is the intensity
of the incident light. Quantitative IR analysis is based on the Beer-Lambert law (note that in the case
of Raman scattering, the intensity of the scattered light is directly proportional to the concentration of
the species considered):
3.2 Vibrational Spectroscopy 115

Fig. 3.8 Graphical

description of the
assessment of the
absorbance value of a well-
separated peak

A ¼ εcl ð3:8Þ

where ε is the absorptivity or extinction coefficient, c is the concentration and l is the thickness of the
specimen. The absorbance is thus proportional to the concentration of a given substance or group in a
molecule. The absorbance of a given peak is obtained by defining a baseline in the A-υ spectrum and
then measuring the area under the peak or, more simply but less accurately, by merely recording the
height of the peak (Fig. 3.8). The baseline construction is often a difficult task because it is not
uncommon for nearby peaks to overlap. In some cases, it is advisable to draw a sloping baseline. In
other cases, the composite spectrum must first be resolved into its components according to the
equation:
X
LðυÞ ¼ Li ðυÞ ð3:9Þ
i

where each LðυÞ may be described by a Gaussian or Lorentzian distribution function, the latter being
defined as:
ω i 2
I i0
Lð υ Þ ¼ 2
ωi 2 ð3:10Þ
ðυ  υi0 Þ2 þ 2

where υi0 is the centre wavenumber of the peak, Ii0 is the peak intensity value and ωi is the breadth of
the peak. In practice it is useful to know the υi0 values of the absorption peaks involved. This can be
accomplished by analysing simpler substances showing only single absorption peaks. Equations (3.9)
and (3.10) can then be fitted to the experimental data to enable the adjustable parameters Ii0 and ωi to
be determined.

3.2.4 Sample Preparation

Liquids studied by IR spectroscopy are examined as thin films between two IR transparent plates or as
solutions in IR transparent solvents, e.g. carbon tetrachloride, cyclohexane or chloroform. Solid
polymers may be studied in solution using one of these solvents. It may also be useful to cast a film
from such a solvent onto a crystal of NaCl. The optimum thickness for a transmission measurement
depends on the IR absorption of the polymer. Specimens of weakly absorbing polymers such as
polyethylene can be relatively thick, approximately 100 μm, whereas specimens of a strongly
absorbing polar polymers have to be thinner, typically between 10 and 30 μm. It is possible to
116 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.9 Arrangement of

incident beam in Raman
spectroscopy for (left) a
thin and (right) a thick
specimen

prepare samples of suitable thickness by cutting with a microtome. Potassium bromide discs with
approximately 1 wt.% polymer are also used. Polymers for certain rubbers or other very tough
polymers have to be ground at cryogenic temperatures. In the ATR technique, the sample can be in
either liquid or solid form. Care should be taken when using rough materials, such as hard powder
particles, that can damage the equipment.
One of the advantages of Raman spectroscopy is the ease of sample preparation. Almost any
sample shape or form is acceptable. Thin films should be oriented almost parallel to the incident beam
to allow a long contact length and hence a large surface area that will Raman scatter (Fig. 3.9).
Thicker transparent specimens can be illuminated from their end surface, and, in essence, the
specimen then acts as a waveguide (Fig. 3.9). Powder samples and fibres are more difficult to study
due to the elastic scattering of the exciting radiation. Solutions and melts can be studied in glass
containers.
It should be stressed that IR and Raman spectroscopy are two of a number of methods that can be
used to monitor events continuously in a sample experiencing changes in the aggregational state, for
instance, polymerization involving a change from a liquid to a solid state.

3.2.5 Applications of Vibrational Spectroscopy

The list of useful applications of IR and Raman spectroscopy in polymer science and physics is
almost endless. A few illustrative examples are given in this section. The characterization of
molecular structure in polymers is one of the important uses. It is possible to measure the concentra-
tion of end groups and hence assess the number average molar mass in polymers such as poly
(butylene terephthalate). The – COH end groups absorb at 3535 cm–1 and the – COOH end groups at
3290 cm–1. End group analysis to assess molar mass is not, however, applicable to high molar mass
polymers. One advantage of the IR method is that the polymer needs not be dissolved prior to the
analysis. This method is obviously not suited for branched polymers.
Stereoregularity (tacticity) can also be assessed by IR spectroscopy for some polymers. The IR
spectra of isotactic and syndiotactic poly(methyl methacrylate) are different, and the same
observations has been made for polystyrene and polypropylene. Solid isotactic polypropylene
shows several absorption peaks (805, 840, 898, 995 and 1100 cm–1) that are absent in the spectrum
of atactic polypropylene. The spectra of the two polymers in the molten state are however identical,
and this suggests that the absorption peaks present only in semicrystalline isotactic polypropylene are
due to vibrations of the particular helical structure (31 helix) characteristic of the crystalline phase.
The difference in configuration thus causes a difference in conformation which is revealed by IR
spectroscopy.
3.2 Vibrational Spectroscopy 117

Tacticity is, however, ideally assessed by 13C–NMR spectroscopy (Sects. 3.3.1 and 3.3.3). Chain
branching in polyethylene is another historically important field. The CH3 groups, which in a polymer
of medium to high molar mass are almost exclusively the end groups of the branches, absorb radiation
at 1379 cm–1, and their concentrations can be determined according to the equation:


A1379
%CH3 ¼ K ð3:11Þ
A1465

where K is a constant which depends on the branch type·, A1379 is the absorbance at 1379 cm–1
(methyl groups) and A1465 is the absorbance at 1465 cm–1 (assigned to methylene groups). There are
several methylene absorption bands interfering with the 1379 cm–1 peak, and these can be subtracted
by taking the difference spectrum between the sample and a perfectly linear polyethylene. The
proportionality constant K can be obtained by calibration using polyethylenes with known degrees
of branching, as assessed by 13C-NMR spectroscopy.
The concentration of unsaturated groups in polymers can be determined both by IR and Raman
spectroscopy. These groups are susceptible to various chemical reactions, desirable in same cases and
hazardous in other cases. The C¼C stretch band appears at about 1650 cm–1, and a series of
absorption peaks should appear at 965 cm–1 (trans-vinylene), 910 cm–1 (vinyl end group) and
730 cm–1 (cis-vinylene) in IR spectroscopy. Raman spectroscopy is, however, more suitable for the
quantitative analysis of unsaturation, since the Raman peaks are more intense than the IR absorption
peaks. Polybutadiene shows Raman peaks at 1650 cm–1 (cis-vinylene), 1655 cm–1 (vinyl end group)
and 1665 cm–1 (trans-vinylene).
The carbonyl-stretch vibration is commonly used to assess the degree of oxidation of polyolefins.
A number of overlapping absorption peaks, essentially between 1700 and 1800 cm–1, appear in
oxidized polyolefins. They are assigned to various oxidation products: ketones, aldehydes, esters and
carboxylic acids. A quantitative analysis can be made only by resolution of the ‘composite’ spectrum
and after proper calibration with pure low molar mass ketones, aldehydes, esters and carboxylic acids.
The spatial distribution of oxidation in polymer products can be determined by IR microscopy.
Figure 3.10 shows the oxidation profiles of polyamide 6,6 samples exposed to air at 180  C for
different periods of time; the arrow in the graph indicates how the oxidation profiles change as a
function of ageing time. Diffusion-limited oxidation (DLO) is a well-known phenomenon occurring
at high temperature, where the oxygen is consumed before it reaches the core of the sample. This give
rise to a pronounced oxidation profile where the oxidation is concentrated to the surface region. The
oxidation profiles are obtained by cutting through the cross-section with a microtome and then
assessing the carbonyl content as a function of depth using an IR microscope equipped with a
liquid-nitrogen-cooled mercury-cadmium-telluride (MCT) detector (Wei et al. 2018).
With confocal Raman microscopy, it is possible to obtain spectral information through the
thickness of a sample without the need for physical cutting, which means that this method is
non-destructive. This is only possible if the sample is Raman-(semi)transparent. Figure 3.11 shows
the result of an analysis of a two-layer material with a 110 μm polyethylene layer on a several mm
thick poly(methyl methacrylate) base. The data are obtained with a He-Ne laser (λ ¼ 632.8 nm) and a
holographic grating of 1800 lines mm–1, and the depth profile is obtained by the vertical displacement
of the microscope stage using a μm-screw. Note the interface broadening between the two materials
and the fact that the interface is not located precisely at the 110 μm position. This discrepancy is
ascribed to the laser refraction at the air/sample interface, which is also responsible for the attenuation
of the Raman signal along the thickness of each layer.
The chain conformation can be studied with IR spectroscopy. Syndiotactic polyvinylchloride is a
well-known example. The trans-planar zigzag conformation in which the C–Cl bond is trans to the
C–H bond shows stretching vibrations at 603 and 639 cm–1, whereas both trans-gauche and gauche-
118 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.10 Diagram of the surface oxidation as a function of time of a polymer sample, determined in the ATR mode
with an IR microscope. The carbonyl index (CI) is the size of the carbonyl band between 1695 and 1760 cm–1,
normalized with respect to a band that is not affected by the oxidation (in the case of polyamide 6,6, the 1193 cm–1
peak). Drawn after Wei et al. (2018)

Fig. 3.11 Raman confocal microscopy in-depth profiling of a two-layer material. Drawn after Tomba et al. (2007)

gauche conformations show absorption at 695 cm–1. Spectral overlap is, however, a problem in this
kind of analysis.
The absorption peaks at 850, 975, 1120, 1370 and 1470 cm–1, all assigned to the ethylene group,
are stronger in semicrystalline PET than in amorphous PET. It is known that the preferred conforma-
tion (i.e. the crystalline state) of this group is trans and, hence, these absorption peaks are assigned to
the trans conformation. The amide I region is known to be sensitive to the molecular conformation
(secondary structure) in peptides and proteins and less sensitive to the actual amino acid composition.
By deconvolution and curve resolution of the total absorbance in this region (Fig. 3.12), it is possible
to estimate the relative contents of various secondary structures. The analysis is based on known
absorbance bands determined on peptide structures with known secondary structure.
3.2 Vibrational Spectroscopy 119

Fig. 3.12 IR absorbance in the amide I region of a larvae protein, and the resolved component fittings based on specific
secondary structures (ATR mode). Dashed and red curves show the fitted and experimental total absorbance. Drawn
after Alipour et al. (2019)

Fig. 3.13 IR absorbance

() and mass uptake of
water by thermogravimetry
() as functions of the
thickness(L )-normalized
square root of time for a
polyamide 6,6 specimen.
Drawn after data of
Camacho et al. (2002)

Hydrogen bonds are the strongest secondary bonds in polymers. They appear in several important
polymers, such as polyamides, and they also play a vital role in certain miscible polymer blends.
Their effect on the IR spectrum is well-known. In polyamides, the vibrational stretching frequencies
of hydrogen-bonded OH- and NH-groups are shifted to lower values than in their non-hydrogen-
bonded analogues. The vibrational frequency of the NH-group drops from 3450 cm–1 to 3300 cm–1
(hydrogen-bonded). A similar shift of the absorption peaks in polymer blends (with reference to the
frequencies of the pure components) may indicate a specific interaction between the different
polymers, which is an important feature of miscibility. The detection of hydroxyl bands can be
used to determine the kinetics of water uptake in polymers. The advantage of doing this in the near-IR
region is the low absorbance by the sample, which enables transmission measurements through
relatively thick specimens, and the results then scale to the total water content throughout the
thickness, despite the gradient in water content in the sample during the uptake period. This, in
turn, makes it possible to compare the uptake with, e.g. data obtained by thermogravimetry. This is
shown in Fig. 3.13, where the water uptake in a 3 mm thick polyamide 6,6 sample is determined in
transmission by measuring the increase in the size of the multiple scatter-corrected 5150 cm–1
absorbance peak, which is due to a combination of hydroxyl asymmetric stretch and bend vibrations.
120 3 Spectroscopy and Scattering of Radiation by Polymers

IR spectroscopy is suitable for monitoring chemical reactions and polymerization in real time.
Figure 3.14 shows the decrease in free carboxylic groups and the increase in ester groups during the
self-catalyzed self-condensation polymerization of a birch bark monomer (9,10-epoxy-18-
hydroxyoctadecanoic acid). The monomer was placed directly on the ATR crystal and covered
with a glass slide, and the ATR unit was maintained at constant temperature (150  C) during the
reaction. The increase in height of the 1730 cm–1 ester band and the decrease in height of the 1717
cm–1 carboxylic band are due to the ring-opening polymerization of the epoxy group yielding an ester
and an alcohol (faster process) and direct esterification of alcohol and carboxylic acid (slower
process). The polymerization was also verified using 1H and C13 NMR spectroscopy. The advantage
with IR spectroscopy rather than NMR spectroscopy is that the reaction can be followed with the
former method into the later stages of the reaction when a gel structure is present (Fig. 3.15).
The use of the ATR unit facilitates real-time reaction monitoring by IR spectroscopy. Figure 3.16
shows one such case: the curing by a thiol-ene reaction of a trimercapto-acetate-methyl oleate system,
using a UV lamp above a droplet of the monomers placed on the ATR crystal (Fig. 3.15). The IR data
shows that methyl oleate rapidly underwent an isomerization from a cis- to a trans-configuration and
that the UV-induced consumption of unsaturated groups was slower. It is worthwhile to note the

Fig. 3.14 Polymerization

of a birch bark monomer as
revealed by IR
spectroscopy. Solid, dashed
and dotted curves
correspond to 0, 30 and
180 min of reaction, and an
insoluble gel was formed
after 50 min. Drawn after
Nameer and Johansson
(2017)

Fig. 3.15 Left: UV-curing setup with an ATR crystal (left). Right: the continuous curve shows the relative size of the
absorption band at 3010 cm–1 of the cis-unsaturated groups in methyl oleate as a function of irradiation dose, and the
broken curve shows the relative size of the 970 cm–1 band of the trans-unsaturated groups in methyl oleate as a function
of irradiation dose. Drawn after Samuelsson et al. (2004)
3.2 Vibrational Spectroscopy 121

Fig. 3.16 The thiol peak as

a function of time using
FT-Raman spectroscopy
with a NIR laser on a bulk
liquid. Drawn after
Claudino et al. (2014)

simplicity of the sample preparation and the fact that the data obtained were comprehensive
confirming the efficiency of the ATR-IR spectroscopy method.
The thiol group (–SH) is a good example of a structure that is more readily monitored by Raman
spectroscopy than by IR spectroscopy, simply because of its low dipole moment. Figure 3.16 shows
Raman spectroscopy data confirming the gradual consumption of thiol as a function of time in a
reaction between R-(+)-limonene and a monothiol (iso-tridecyl 3-mercaptopropionate).
Specimens subjected to a mechanical load show a frequency shift of IR absorption and Raman
peaks, which is in many cases proportional to the applied stress. There is also a change in shape of the
frequency-shifted peaks from symmetric to asymmetric, which has been interpreted as indicating that
different chain segments are subjected to different stresses, i.e. that the stress is unevenly distributed
(Kausch 1978). Wool (1975) suggests other causes, however, such as a decrease in the force constant
due to bond weakening.
With a rheo-Raman scattering setup involving a tensile stage, Kida et al. (2016) measured the
shift of a number of Raman peaks during the tensile drawing of a high-density polyethylene
sample with a pre-existing neck (Fig. 3.17). The light from a 639.7 nm laser was led through a
monochromator (laser line filter) and focused on the sample with a spot diameter of 1 mm. The
excitation light was absorbed with a Raman long-pass filter, and the scattered light was detected by a
monochromator-equipped charge-coupled device (CCD) camera. The two grips on the tensile stage
were moved symmetrically away from each other to ensure that the Raman scatter was continuously
capturing the deformation (1 mm min–1) in the neck region. Figure 3.17 shows that during the mainly
elastic deformation up to the first yield point, the two bands assigned to the symmetric (1130 cm–1)
and anti-symmetric C–C (1063 cm–1) stretching modes perpendicular and parallel to the crystalline
chain axis remain unchanged but that the band at 1080 cm–1, ascribed to the C–C stretching of
amorphous chain segments, showed a Raman red shift. This is explained by the hypothesis that a
mechanical stress causes a tensile strain mainly on the amorphous chain segments. Beyond the first
yield point, the stress exerted a tensile strain on the crystalline chains (1063 cm–1 red shift). The
different (blue shift) behaviour of the 1130 cm–1 Raman peak between the first and second yield point
is ascribed to a densification/compression perpendicular to the chain axis accompanying the
stretching along the chain axis in the crystal. At the second yield point, a second neck is formed in
the initial neck region, which relieves, partly by plastic deformation, imposed stresses on the chains.
Immediately after the second yield point, the stress is therefore relaxed in the crystal, leading to a
short plateau in the 1063 cm–1 shift, whereas the 1130 cm–1 shows a red shift. At the same time,
the stress on the amorphous chains is also partly relieved shown again by looking at the trend in the
122 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.17 Peak shift in Raman scattering and stress as a function of strain (left). The solid, dashed and dash-dotted
curves correspond respectively to the peaks appearing at 1063, 1130 and 1080 cm–1. A positive shift is a blue shift. a and
b refer to the first and second yield points (sketched in the right-hand drawing). Drawn after data of Kida et al. (2016)

1080 cm–1 peak shift. In the strain hardening region, the two crystalline bands show a weakly varying
red shift (and the 1080 cm–1 shift remains red) indicating an almost uniform stress imposed on the
chain segments throughout this region. The trends in the shift of the CH2-bending modes also reveal
information on the constraining of the chain segments in the crystal-amorphous interphase during the
tensile test. For further details, cf. Kida et al. (2016).
Another useful application of IR spectroscopy is the assessment of chain (group) orientation
through infrared dichroism measurements. This method has been described by Gedde and Hedenqvist
(2019a). Raman spectroscopy can also be used to determine chain orientation, and both IR and
Raman spectroscopy are useful for characterization of the physical structure of semicrystalline
polymers (cf. Fig. 3.17). Both IR and Raman spectroscopy provide information about the crystallin-
ity, although the actual crystallinity values obtained by these methods often deviate from the values
obtained by the three preferred methods: WAXS, density measurement and DSC. The strength of
Raman spectroscopy is that it provides information about the crystal-amorphous interphase, which is
also possible with solid-state C13 NMR (cf. Sect. 3.3).
The major factor differentiating a chain in a crystal from a chain in the amorphous phase is that the
former is in its preferred conformational state. An additional requirement is the presence of lateral
order in the crystal. There are reported values of crystallinity obtained by IR or Raman spectroscopy
that are based on absorption peaks characteristic only of the preferred conformation. Zerbi, Ciampelli
and Zamboni (1964) divided what they called regularity peaks into three categories: (i) The first
group, being the ‘weakest’ indicator of crystallinity, includes absorption bands assigned to certain
single-bond conformations. The ratio of the 1450 cm–1 (trans) and 1435 cm–1 (cis) bands of trans-1,4,-
polybutadiene increases with increasing crystallinity, but chains with trans conformations are also
present in the amorphous phase. (ii) The second group is a group of adjacent chain segments with a
specific chain conformation, e.g. the 31 helix of isotactic polypropylene. Absorption peaks at
805, 840, 898, 972, 995 and 1100 cm–1 are characteristic of this conformation, but they are replaced
by a broad peak centring on 972 cm–1 in molten isotactic polypropylene. There are several arguments
to account for the presence of the 31 helix conformation in the amorphous phase but at a much lower
concentration than in the crystalline phase. Thus, the intensity of these regularity peaks may not be
proportional to the crystallinity. (iii) The third group shows what may be called true crystallinity
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 123

peaks. Their occurrence is a consequence of both intra- and intermolecular interactions typical of the
crystal phase. These peaks disappear completely when the crystals melt. Poly(oxy methylene) and
syndiotactic poly(vinyl chloride) are examples of polymers showing such regularity peaks.
One of the uses of Raman spectroscopy in morphological analysis is for the measurement of the
longitudinal acoustic mode (LAM) in polyethylene, which goes back to early work (Mizushima and
Simanouti 1949) on paraffins. The LAM frequency (υ (LAM-n)) can be converted to an all-trans
chain length or crystal thickness (Lc) according to the following formula:

2
n E
Lc ¼ ð3:12Þ
2υ ρ

where n is the order of the vibration (n ¼ 1 is practically the only value used), E is the Young’s
modulus in the chain axis direction and ρ is the density. The LAM (n ¼ 1) frequencies are extremely
low, 10–20 cm–1. There are however several complicating factors which may affect the results; one of
these is that the elastic modulus is not precisely known, although reasonably accurate estimates
suggest values close to 300 GPa for polyethylene (Holliday and White 1971). Secondly, there is often
a chain tilt in polyethylene crystals, i.e. the length of the all-trans chain is greater than the thickness of
the crystal. Thirdly, there is almost always a distribution of crystal thicknesses, and it is not clear
which type of average is assessed by Raman spectroscopy. Crystals contain defects, and the exact
effect of these on the LAM frequency is not known. LAM frequency measurements to assess crystal
thickness have also been performed on a number of other polymers, e.g. poly(oxy methylene), poly
(ethylene oxide) and poly(tetrafluoro ethylene), but the interpretation of these results is less straight-
forward than for polyethylene (Bower and Maddams 1989).

3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy

3.3.1 Fundamentals

NMR is based on the principle that nuclear spins under the influence of an external magnetic field are
split into two energy levels. NMR-active nuclear spins are those associated with atoms having an odd
number of either protons or neutrons. Hydrogen (1H) is NMR-active and is the subject of much work.
Other nuclei which have been subjected to studies are 13C, 2H and 15N (Ibbett 1993), but these nuclei
are present only in low concentrations in polymers, because of their low abundance in nature. The
nuclear spins are associated with an angular momentum, characterized by the angular momentum
vector (I):

h
j Ij ¼ ðI ðI þ 1ÞÞ0:5 ð3:13Þ
2π

where h is Planck’s constant and I is the nuclear magnetic spin quantum number. The latter is ½ for
1
H, 13C, 15N and 19F and 1 for 2H (nuclei with a quantum number larger than ½ are referred to as
quadrupolar). The magnetic moment (μ) is proportional to I:

μ ¼ γI ð3:14Þ

where γ is the magnetogyric ratio. The magnetic moment is given in scalar units by:
124 3 Spectroscopy and Scattering of Radiation by Polymers

μ ¼ gN β N I ð3:15Þ

where gN is a dimensionless constant and βN is the nuclear magneton. Both these take different values
for different nuclei. The magnetic quantum number (mI) can, according to quantum theory, only take
specific values: I, I – 1,... , –I. For 1H and 13C, mI can take the values +½ and ½. The magnetic
moment can thus only take discrete values according to:

μ ¼ gN β N m I ð3:16Þ

The potential energy (E) of the isolated nucleus placed in a magnetic field (B0) is given by:

E ¼ μβ0 ð3:17Þ

and the energy difference (ΔE) between a nucleus with m1 ¼ +½ and –½ is obtained by combining
Eqs. (3.16) and (3.17):

ΔE ¼ hυr ¼ E1=2  Eþ1=2 ¼ gN βN B0 ð3:18Þ

where υr is the resonance (precession/Larmor) frequency. Equation (3.18) states the condition for
resonance. The resonance frequencies for 1H and 13C, at B0 ¼ 1.0 T, are equal to 42.577 and
10.705 MHz, respectively. The absorption of the radio frequency radiation is detected only if there
is an excess of spins in the low-energy state. Equilibrium can only be obtained if a fraction of the
excited spins is permitted to return to the ground state. The excessive energy has to be transferred to
the surrounding lattice, by a relaxation process characterized by the spin-lattice relaxation time (T1).
There is second relaxation process involving the transfer of energy between spins characterized by the
spin–spin relaxation time (T2).
NMR spectroscopy may be carried out by searching for the resonance conditions either at constant
magnetic field strength or at constant frequency. Both methods are used in practice. NMR spectros-
copy would not be particularly useful if it were not for the occurrence of a chemical shift. The
extranuclear electrons surrounding a particular nucleus affect the effective magnetic field (Beff) felt
by the nucleus:
Beff ¼ B0 ð1  σ Þ ð3:19Þ

where σ is the screening constant, the magnitude of which depends on the chemical environment,
which in turn affects the electron density around the nucleus. This screening is also referred to as
shielding and the opposite of de-shielding. The resonance condition formula is then given by:

ΔE ¼ hυr ¼ gN βN B0 ð1  σ Þ ð3:20Þ

The resonance condition depends on the field strength B0, which is undesirable. The accepted
standard is to express the chemical shift with reference to tetramethylsilane (TMS), a substance which
shows a single 1H resonance peak, according to:


υs  υTMS
δðin ppmÞ ¼ 106 ð3:21Þ
υTMS

TMS is also used as the reference for expressing the chemical shift in 13C NMR spectroscopy.
Figures 3.18 and 3.19 show typical chemical shifts in 1H NMR and 13C NMR spectroscopy. The ppm
range for 13C is much wider than for 1H, but the signal is less intense for 13C than that for 1H which is
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 125

Fig. 3.18 1H NMR chemical shifts of some commonly occurring groups, relative to TMS. Drawn from data presented
by Ibbett (1993); some of the data were obtained from sources which were not confirmed to have been peer-reviewed

due to the fact that only 1.1% of all carbon atoms are 13C, whereas 1H constitutes 99.99% of all
hydrogen. The wide ppm range for 13C NMR (with reference to that of 1H NMR) is primarily due the
presence of a greater number of electrons (6) surrounding the 13C nucleus; 1H has only one electron.
Increasing de-shielding (moving electrons further away from the nucleus), decreasing the magnetic
field at constant frequency or increasing frequency at constant magnetic field leads to a ‘downfield’
shift (higher chemical shift), equivalent to moving towards the left in Figs. 3.18 and 3.19. For
example, the reason for the larger chemical shift of the hydrogen in the carboxylic group compared
to that of the hydrogen in methane is the stronger de-shielding effect in the former case (oxygen
withdrawing the electron from the proton). The larger chemical shift of an ester carbon compared to
that of a methane carbon is also due to the greater de-shielding in the former case. The exact chemical
shift of an aromatic depends on the type of substitution on the ‘benzene’-group. An electron-donor
group (e.g. – O – CH3) and an electron-withdrawing group (e.g. – CN) shift the aromatic resonance
towards lower and higher chemical shifts, respectively. As shown in Fig. 3.18, the chemical shift
associated with hydrogen bonding (shown here for hydroxyl groups) occurs over a broad ppm region.
The stronger the overall hydrogen bonding in the network, the more is the hydrogen de-shielded and
126 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.19 13C NMR chemical shifts of some commonly occurring groups, relative to TMS. Drawn from data presented
by Ibbett (1993)

the greater is the chemical shift. Compounds with π-electrons, like aromatic groups, show magnetic
anisotropy, which means that the magnetic field varies over the structure with both shielded and
de-shielded parts and, depending on the location of the protons, the net effect can be either an increase
or a decrease in the magnitude of the chemical shift.
Nearby nucleus spins affect the local magnetic field by an amount ΔB according to the equation:

3  
ΔB ¼  gN βN 1  3 cos 2 ðθÞ r 3 ð3:22Þ
4

where θ is the angle between the magnetic field vector and the vector between the interacting spins
and r is the distance between the nuclei. This so-called spin–spin interaction causes a broadening of
the resonance line. However, if the mobility of the molecular segments is high, as in a solution or a
liquid, ΔB is low, because hcos2θi approaches a value close to 1/3. Solid samples show considerable
broadening in their resonance lines due to spin–spin interaction. The problem can, however, be
eliminated by rotating the sample at an angle of 54 440 (cos2 (54 550 ) ¼ 1/3) to the magnetic field.
This angle is known as the magic angle.
The ‘coupling effect’ refers to an interaction between nearby nuclei. The so-called scalar through-
bond interaction occurs due to polarization of the bonds by the nearby spins. This causes a splitting of
the resonance lines into 2NI + 1 lines, where N is the number of magnetically equivalent nuclei with
spin I. The shifts are expressed with a coupling constant (J) given in Hertz, the size of which is
essentially independent of the magnetic field strength but dependent on chemical structure. The
relative intensities of the split lines are given by Pascal’s triangle shown in Fig. 3.20. Starting at the
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 127

Fig. 3.20 Pascal’s triangle including rows 1 to 7

Fig. 3.21 Splitting into a quartet of a CH2 group resonance signal when the CH2 group is attached to –CH3 group. It is
assumed that no other protons are nearby which may affect the CH2 group

top, N ¼ 0 (first row), there are no interacting nuclei and only one band appears, etc. This rule is valid
only when the chemical shift differences are much greater than the coupling constants, which is
normally the case.
An illustrative example of splitting is the resonance signal from the protons on a methylene unit
coupled to a methyl group (– CH2 – CH3) (www.chem.ucalgary.ca). The three methyl protons are
oriented magnetically either along or against the applied magnetic field. This gives rise to eight
possible combinations of orientation for the protons, where only four are uniquely different, as sensed
by the adjacent methylene hydrogen (Fig. 3.21). Note that nuclei with the same magnetic properties,
128 3 Spectroscopy and Scattering of Radiation by Polymers

e.g. the two methylene hydrogens, do not ‘split’ each other. When the three protons are all magneti-
cally oriented along the field, the de-shielding is at a maximum (highest chemical shift), whereas a
maximum shielding is observed when the three protons are oriented opposite to the field. The two
cases when one or two of the three protons are oriented along the field are associated with intermedi-
ate chemical shifts. Hence, and according to the Pascal triangle, the methyl group next to the
hydrogens in focus results in four split lines for the CH2 group, and the two in the middle have the
highest intensities, so that the relative intensities are 1:3:3:1. The size of the coupling constant is
different for different chemical groups, bonds and bond configurations, and, as such, it gives valuable
information regarding the local chemistry. Different couplings present at the same time give split line
groups with variations in J and in the distance between split lines. In 13C-NMR, because the 13C
nuclei are so few, the coupling for the carbon occurs only with nearby hydrogens. In the 1H-coupled
13
C NMR spectrum of polyisobutylene, the methyl groups therefore occur as a quartet because of the
three adjacent hydrogens, and the methylene occurs as a triplet because of the two hydrogens.

3.3.2 Instrumentation

An NMR instrument consists of a high-field magnet, a radio frequency source, an NMR probe, a
sweep system and an amplification and recording system. Stronger magnetic fields can be obtained by
using superconducting solenoids. The main advantage of using a stronger magnetic field is that both
the resolution and sensitivity are greater. The magnetic field used is from below 1 T in small portable
magnets to above 20 T in stationary systems (Moser et al. 2017).
Modern high-resolution instruments are Fourier transform (FT) spectrometers that record the
response of the nuclei to relatively strong radio frequency (RF) pulses, converting the resulting signal
time response by a mathematical procedure into a frequency spectrum. The NMR instrument is
usually referred to by the resonance frequency of the proton obtained with the spectrometer (e.g. a
400 MHz instrument). Today’s instruments go up to 900 MHz and even into the GHz region. The
advantage of the FT technique, compared to the older continuous wave technique, is that a spectrum is
obtained in seconds and several hundred spectra can be added together, which greatly increases the
signal-to-noise ratio. This is particularly important for 13C NMR due to the low concentration of 13C
nuclei in the samples. Also, the development of parallel acquisition with multiple receivers for
different nuclei makes it possible to carry out multidimensional NMR experiments to be made within
a reasonable period of time. A large increase in the sensitivity of the technique was achieved when
cryoprobes were introduced in the late 1990s (Spiess 2017a, b). Figure 3.22 illustrates a common
NMR experiment involving a radio frequency (RF) pulse. When an RF pulse is applied over the static
magnetic field, the magnetization vector tilts, and the degree of tilting depends on the pulse width
(Fig. 3.23). The tilting is due to the fact that the RF field is applied along the x-axis yielding a rotation
about the x-axis of the magnetic field. After the pulse, the excited nuclei will spin with a ‘Larmor’
frequency, which depends on the type of nuclei and the magnetic field strength. In this acquisition
time period, the sinusoidal signal emitted by the sample will decay with an exponential rate described
by the time/parameter T2*, and the event is referred to as a free-induction decay (FID). T2* depends
on the true transverse relaxation time T2 (Fig. 3.23) and on spatial inhomogeneities in the magnetic
field. The sinusoidal FID pattern is a consequence of the fact that the magnetic vector rotates and that
its signal component in the xy plane is captured by coils along the y-axis. A maximum NMR signal is
obtained when the pulse width generates a 90 pulse. The delay time needed between repetitive pulses
(to improve the signal-to-noise ratio) in order not to obtain unwanted FID overlap/mismatch depends
on the time needed for the magnetization vector to relax/orient back to the static magnetic field
direction. This time increases with increasing tilt angle, and, in order to save spectral collection time,
the pulse width is often shortened to yield a smaller tilt angle.
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 129

Fig. 3.22 A single (repetitive) pulse experiment illustrating the delay, pulse and acquisition time and the Fourier
transform of the FID time domain data to frequency domain

Fig. 3.23 Illustration of the tilting of the magnetic vector from the static magnetic field in the z-direction towards the
xy plane using an RF pulse. This is illustrated here in the rotating xy frame. In the experimental frame, the magnetic
vector rotates in one direction around the z-axis

As mentioned above, the relaxation back from the high-energy tilted magnetization to a magneti-
zation along the static field involves spin–matrix (T1) and spin–spin (T2) relaxation (Fig. 3.23).
Compounds with less mobile molecules generally show very different T1 and T2 values. For a solid
material, the release of energy through spin-lattice transfer is significantly slower than through spins
(T1 > T2), but for a non-viscous liquid, they are of the same magnitude (Martı́nez-Richa and Silvestri
2017). T2 also affects the NMR signal line shape; the faster the relaxation, the broader the peak. There
is another useful relaxation time, denoted T1ρ and referred to as the relaxation time in the rotating
frame, which means that it is the relaxation time T1 under a constant 90 (spinlock) RF field. It is
considered to be more sensitive to slow fluctuations, and it has a lower value than T1 in solid
materials.
The NMR technique is usually a non-destructive method, although liquid NMR requires a sample
in a dissolved form. The choice of solvent depends on the chemical characteristics of the sample, and
it is important that the NMR signal from the solvent does not overlap with that of the dissolved
polymer. For this reason, deuterated or partially deuterated solvents are regularly used, and this is also
beneficial in the case of 13C NMR. A commonly used solvent is deuterated chloroform (CDCl3). In
water-soluble systems, D2O can be used, and for wood-based materials, deuterated dimethyl
sulphoxide (DMSO-d6) is often used. For a detailed description of NMR solvents, Hakada and
Kitayama (2004) is recommended. In solid-state NMR, it is important to pack the sample so that it
is possible to spin the sample-containing rotor without any problem. In modern table-top mobile
NMR equipment, with a ‘measuring spot’ in front of the magnet, the sample can be in essentially any
geometry. As with FTIR and Raman spectroscopy, NMR can also be used as an imaging technique.
130 3 Spectroscopy and Scattering of Radiation by Polymers

Its use to obtain solute concentration gradients in polymers is, however, described in Chap. 7. This
has found important use especially in medical care, and it is therefore beyond the scope in this
chapter. The state of the art on this technique is described by Moser et al. (2017).
A range of techniques are employed to enhance spectral lines and method sensitivity:
• The nuclear Overhauser effect (NOE) enhances specified lines by changing the relaxation
behaviour of one nuclear transition by saturation with other nuclear transitions. The coupling
effect is only possible when the two nuclei are sufficiently close, in a through-space coupling.
• Spin decoupling is used to simplify the analysis of complex spectra and is performed by modula-
tion of the radio frequency radiation. The sample is irradiated with the resonance frequency of one
nucleus, while the radiation sweeps through the entire spectral region. An example is dipolar
decoupling (DD) of 1H to reduce the line broadening of 13C signals in solid-state NMR.
• Broadband decoupling is used when the resonance frequency of the decoupled nucleus covers a
wide range. It is used in 13C NMR work; the spectra are obtained with decoupling of the 1H nuclei.
• The cross-polarization (CP) technique is used to enhance the 13C signal using a transfer of
magnetization from the abundant 1H spins to the dilute 13C spins.
• Magic angle spinning (MAS) in solid 13C NMR reduces the effect of magnetic anisotropy on the
signals. High-resolution 13C NMR spectra are obtained usually by a combination of DD, MAS and
CP.
• Dynamic nuclear polarization (DNP), which involves spin transfer from electrons to nuclei, is a
relatively new technique to enhance the NMR signal in, e.g. liquid NMR (dissolution-DNP) and in
combination with MAP in solid-state NMR (Saalw€achter 2019; Martı́nez-Richa and Silvestri
2017; Ashbrook and Hodgkinson 2018).

3.3.3 Polymer Applications of NMR Spectroscopy

NMR spectroscopy is the primary method used in the assessment of the chemical structure of
polymers, including tacticity, and chain branching. Assessment of meso and racemic dyads in tactic
polymers is routine, and even longer steric sequences such as triads can be determined. Head–tail
configuration and the sequence distribution of comonomers in copolymers are two good examples of
problems solved by NMR.
An example is given below showing how NMR spectroscopy can be used to determine the
products of a polyester polymerization reaction and subsequently those of a copolymerization with
the same monomers (Kainulainen et al. 2020). At a time when the interest in using bio-based
polymers is greater than ever, an example is given here with two biomass-derived monomers and a
dialcohol (Fig. 3.24). 1H NMR (400 MHz) of the furan (DF) monomer shows a singlet spectral peak at
a chemical shift of 3.93 ppm and a singlet peak at 7.23 ppm. Integration of the areas beneath the two
peaks gives a relative size of 2:6. Since in proton NMR the size of the peak relates to the number of
hydrogens of the same type, this ratio tells us that the large ppm peak corresponds to the protons in the
furan-ring (H(a)) and that the small ppm peak corresponds to the methyl protons. This is also in line
with the larger shielding of the methyl protons leading to a lower shift, and it is in line with the
‘group’ positions in Fig. 3.18. The reason that there is no splitting is that the two types of hydrogens
are too far apart to couple and that the protons with the same shift do not ‘split’ each other. For the
bifuran (DB) monomer dissolved in the same solvent, apart from the methyl singlet, two doublet
peaks appeared. This is because the two protons (H(b) and H(c)) in the bifuran system now have an
environment (shielding) different from that in the symmetric furan case in DF. They appear at 7.26
and 6.90 ppm, where the former peak is due to (H(b)) which is at a shift similar to that of the H(a) in
DF. H(c) experiences a somewhat larger shielding than H(b). The coupling constant is the same in
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 131

Fig. 3.24 From top to

bottom: dimethyl
2,5-furandicarboxylate
(DF), dimethyl
2,20 -bifuran-
5,50 -dicarboxylate
(DB) and 1,4-butane diol

Fig. 3.25 The two homopolymers poly(butylene furanoate) (PBF, top) and poly(butylene bifuranoate) (PBBF, bottom)

both doublets (J ¼ 3.7 Hz), which tells us that the splits come from vicinal hydrogens (H(b)–C–C–H
(c)), i.e. from the same coupling system, since Jbc ¼ Jcb. The ratio of the integrated peaks for the three
hydrogens are as expected 6:2:2 for methyl H:H(b):H(c).
The two homopolymers PBF and PBBf (Fig. 3.25) were investigated with 1H NMR and 13C NMR.
The latter technique, which was broadband proton-decoupled, spans a larger chemical shift range and
is thus able to reveal more details. The proton spectrum of PBF shows a singlet peak for H(a) at
7.33 ppm (close to that in the monomer). The H(d) experiences more de-shielding than H(e) due to the
adjacent electron-withdrawing oxygen, and it appears, consequently, at a greater shift (4.51 ppm) than
H(e) (1.98 ppm) (cf. Fig. 3.18). The two peaks are also broader than the H(a) peak since these are
actually triplets with a small coupling constant (two pairwise H(d) and H(e) protons). The relative
sizes of the H(a), H(d) and H(e) peaks are 2:4:4, which is in accordance with the polymer structure
(Fig. 3.25). The bifuran polymer PBBf has four proton spectral groups. The shifts of the doublets for
H(b) and H(c) are close to those in the monomer. There are also two broad peaks for H(f) and H
132 3 Spectroscopy and Scattering of Radiation by Polymers

(g) very close to those of H(d) and H(e) in PBF. The relative sizes of the four proton peaks H(b):H(c):
H(f):H(g) are 2:2:4:4, again in agreement with the polymer structure.
The chemical shift of the 13C NMR peak of the methylene carbon next to the oxygen depends on
whether it is close to a furan (68.97 ppm) or to a bifuran (68.64 ppm) group, because of different
degrees of shielding (Fig. 3.25). The overall position of this methylene group is that of the methylene
in Fig. 3.19. The peaks appear as singlets without splitting due to the proton decoupling and because
the natural content of the 13C isotope is, as mentioned earlier, so low that two 13C are unlikely to be
close enough to cause any noticeable coupling in the spectrum. The fact that the chemical shift of the
methyl group depends on the presence of furan and bifuran can be used to investigate the structure
when the two polymers are instead copolymerized. The chemical shift for the methylene group next to
the oxygen is 69.06 ppm when it is close to the furan group and 68.54 ppm when it is close to the
bifuran group in a BF-co-BBf sequence/unit. The size of an integrated carbon peak is normally not
directly proportional to the content of the carbon, due to long T1 relaxation times, unless long
acquisition times are used. In addition, the nuclear Overhauser effect (NOE) in broadband proton-
decoupled 13C NMR means that the size of the carbon signal depends on dipolar through-space
carbon-proton coupling and that different carbons will have different signal enhancements. Never-
theless, when PBF and PBBf were mixed in a molar ratio of 1:1 in TFA-d, it was found that the
relative sizes of the methylene carbon peaks (68.97 and 68.64 ppm) were also 1:1. The relative peak
sizes of the four methylene groups (69.06, 68.97, 68.64 and 68.54 ppm) were therefore used to
determine the type of copolymerization here. For a random copolymer, the randomness index Ri is
close to unity:

1 1
Ri ¼ þ ð3:23Þ
LFF LBB

where

AFF þ 0:5  ðAFB þ ABF Þ

LFF ¼ ð3:24Þ
0:5  ðAFB þ ABF Þ

and

ABB þ 0:5  ðAFB þ ABF Þ

LBB ¼ ð3:25Þ
0:5  ðAFB þ ABF Þ

where A is the area of the methylene peak, L is the number average sequence length and the indexes F
and B refer to furan and bifuran units, respectively. For example, for a 50/50 molar copolymer
(PBF50Bf50), LFF ¼ 2.08 and LBB ¼ 1.89, yielding Ri ¼ 1.01, and for a PBF90Bf10 copolymer,
LFF ¼ 8.84 and LBB ¼ 1.16, leading to Ri ¼ 0.98. For the other copolymers (75/25, 25/75/10/90), Ri
was always close to unity. This shows that 13C NMR can be used to assess the degree of randomness
in a copolymer, which was high in this example (a more block-shaped copolymer has a Ri-value
closer to 0 and a copolymer with alternating repeating units has a Ri value closer to 2). Furthermore,
the relative sizes of the different 1H NMR peaks along the copolymer series were the same as the
feeding ratios, indicating a high degree of purity in the copolymer obtained and similar reactivities of
the furan and bifuran monomers.
Molar mass averages (M n and M w) and the molar mass dispersity (D ¼ M w/M n) can be obtained
from NMR spectra (further about molar mass assessment, cf. Chap. 4). The number average molar
mass is best obtained from 1H NMR measurements through the analysis of repeating units and end
groups and proton quantification (Izunobi and Higginbotham 2011). It can also be obtained from 13C
NMR measurements if the experimental conditions are such that the resonance peaks are quantitative.
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 133

The molar mass and molar mass distribution can also be obtained from proton-pulsed field gradient
spin echo NMR diffusion measurements (cf. Chap. 7; Guo et al. 2017). The measurements require
strong gradients, and it is recommended that a high-power gradient diffusion probe be used. The
weight average molar mass is obtained from the polymer diffusivity and diffusivity probability
distribution using the Stokes–Einstein relation, where the molar mass is inversely proportional to
the diffusivity.
In multiscale NMR, such as 2D NMR, it is possible to obtain further information on how nuclei are
linked and on how atoms are connected in a molecule. This information is obtained by manipulating
the pulse sequence. Figure 3.26 illustrates how this can be done. By systematically changing the
evolution time (t1) and recording the acquisition signal during time t2, a 2D NMR chart/spectrum is
obtained via a double Fourier transform step. The time axes data are mathematically transformed into
the frequency axes f1 and f2. The resolution is higher, but the sensitivity is lower in the 2D NMR data
than in the data obtained by separate 1D NMR experiments (Koenig 1992).
If the 2D-NMR experiment is performed as a homonuclear experiment (one type of nucleus,
e.g. protons) with a 90 pulse, a 2D COSY (correlational spectroscopy) spectrum is obtained, which
typically correlates nuclei over two to three bonds (Hatada and Kitayama 2004). The following
example illustrates the method (Abdul-Karim et al. 2017). In the ring-opening polymerization of
ethylene carbonate, CO2 is lost, and two different species (repeating units) appear in the polymerized
structure, ethylene carbonate and ethylene oxide (Fig. 3.27). The proton–proton 2D COSY spectrum
is shown in Fig. 3.28. The signals appearing along the diagonal correspond to those from the same
group, but those outside the diagonal provide information about nuclei that are spin-coupled (cross-
peaks). The cross-peaks in the spectrum tell us that the protons labelled 3 sit on a carbon next to the
carbon with the protons labelled 4, i.e. three bonds away.
In addition, if a distortionless enhancement by polarization transfer (DEPT) NMR experiment is
carried out, a C13 NMR spectrum is obtained in which the resonance peaks become positive or

Fig. 3.26 The pulse sequence methodology generating a 2D-NMR spectrum. The first grey column represents a set of
pulses (the preparation phase). Signals evolve during the evolution time (t1). The second grey column represents a
second set of pulses (the mixing phase) leading to a specific FID pattern during the acquisition time (t2)

Fig. 3.27 Ethylene carbonate (left) ring-open polymerized into the most typical ‘copolymer’ repeating unit (right)
134 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.28 A 2D COSY spectrum of the system described in Fig. 3.27. (Observe that not all peaks are shown). The
numbers refer to the protons labelled in Fig. 3.27. Drawn after Abdul-Karim et al. (2017)

Fig. 3.29 The upper

schematic spectrum shows
a 13C NMR spectrum with,
from low to high chemical
shift (right to left), peaks
associated with carbons
3, 4, 5 and 1 in Fig. 3.24.
The lower schematic
spectrum is a DEPT-NMR
spectrum with a 135 final
proton tip angle in the pulse
program

negative depending on the number of hydrogens connected to the carbon. For an even number of
hydrogens on the carbon, the signal is negative (Fig. 3.29). This gives the additional information that
all carbons are bonded to two hydrogens except for the carbon labelled 2 (Fig. 3.27).
Whereas the 2D COSY shows spin couplings through bonds, a 2D NOESY (nuclear Overhauser
effect spectroscopy) spectrum yields signals from nuclear spin couplings through space. The distance
between the nuclei should be typically less than 5 Å to yield a signal. Figure 3.30 shows an example
where there is a coupling between the methyl group of an alanine amino acid unit and the amine in a
valine amino acid unit in a peptide, information that can be used to assess the secondary structure, in
this case involving a β (II)-turn (Dybal et al. 2002).
In ‘heteronuclear single-quantum correlation’ (HSQC) spectroscopy, the coupling is recorded over
one bond. This is illustrated in Fig. 3.31 for the coupling of carbon and hydrogen, in an example
where the method was used to verify the presence of starch in a water-soluble fraction from an
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 135

Fig. 3.30 2D (1H-1H) NOESY spectrum of a polypentapeptide. Note that only one specific coupling is shown. Drawn
after Dybal et al. (2002)

Fig. 3.31 An HSQC 2D NMR spectrum showing only the peaks from carbon and hydrogen pairs on a starch repeating
unit. Drawn after B€
orjesson et al. (2018)

extraction of barley husk (B€orjesson et al. 2018). Further examples of 2D NMR methods are
presented by Hatada and Kitayama (2004).
Proton NMR is suitable for determining the degree of substitution in a grafted polymer. This is
illustrated in Fig. 3.32 for butyl glycidyl ether grafted onto hemicellulose in order to increase the
136 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.32 The grafting of

butyl glycidyl ether onto a
galactoglucomannan chain.
Drawn after H€ardelin et al.
(2020)

hydrophobicity, lower the glass transition temperature and increase the processability. The degree of
molar substitution is calculated from the ratio of the integral of the methyl proton signal to those of
the protons on the C1 carbon (two resonances occur due to the presence of both α and β anomers)
(Eq. 3.26). Note the division by three because there are three methyl protons. Sometimes, as in this
case, there may be some overlap between the signals associated with the solvent and with the sample,
and it is then advisable to complement the measurement with another NMR measurement. With a
long delay/repetition time (allowing for the relaxation of the carbon T1), it is also possible to estimate
the molar substitution from the 13C NMR spectrum with the protons decoupled during acquisition. In
the example shown in Fig. 3.32, the molar substitution was estimated from the size of the C1 peak on
the hemicellulose and the size of the C11 peak on the graft molecule.
Ð
ðCH 3 Þ
MSH‐NMR ¼ Ð Ð  ð3:26Þ
3 αH1 þ βH1

Thermal transitions and relaxation processes can be revealed by broadline solid-state NMR.
Detailed information about the mobility of individual carbon atoms obtained by 13C NMR is indeed
a valuable complement method to the data obtained by classical thermoanalytical techniques
(cf. Chap. 1). NMR spectroscopy is useful in many other applications, e.g. for the assessment of
miscibility (cf. Gedde and Hedenqvist 2019d) and chain orientation (cf. Gedde and Hedenqvist
2019a).
Solid-state NMR is used to assess the presence of components with different molecular mobilities
and different physical structures, for instance, in semicrystalline polymers (Kitamaru et al.
1986). Data from a study of branched polyethylene by Mattozzi et al. (2010) are shown in
Fig. 3.33. In a polyethylene with 1.6 mol.% 1-octene (hexyl branches), the spectrum was resolved
into an orthorhombic crystal core (OC) component, an amorphous (A) component and a crystal-
amorphous interphase (I) component. In a copolymer with 7.2 mol% 1-octene, the crystallinity was
significantly reduced, which is revealed by the large amorphous (A) peak. In this polymer, an
additional peak was resolved for a monoclinic crystal (MC) phase. Peaks due to side-chain
3.3 Nuclear Magnetic Resonance (NMR) Spectroscopy 137

Fig. 3.33 13C NMR CP/MAS spectra of two polyethylene grades with different comonomer contents (degree of chain
branching). The upper continuous spectrum refers to a polyethylene with 7.2 mol.% 1-octene, and the lower dotted
curve refers to a polyethylene with 1.6 mol.% 1-octene. Drawn after Mattozzi et al. (2010)

(SC) and main-chain (MC) units are also apparent in the 7.2 mol.% sample. Crystallinity from these
13
C NMR data did not agree with crystallinity estimates based on WAXS and density data (Mattozzi
et al. 2010). There was, however, a systematic decrease in the A peak relative to the OC peak with
increasing crystallinity. The IC component as revealed by NMR spectroscopy was much larger than
that shown by WAXS data (Mattozzi et al. 2010), which illustrates the fact that care should be taken
when comparing peak sizes in 13C NMR measurements. Apart from T1 relaxation issues,
complications may also arise from spin diffusion between the different phases and variations in
signal-to-noise ratio between samples with different copolymer compositions. In this specific case,
the signal-to-noise ratio decreased strongly with increasing 1-octene content.
T2-analysis with solid-state proton NMR is very useful for detecting physical changes in a polymer
material, which affects the molecular mobility. Modern portable NMR devices allow flexible and
rapid measurements in a non-destructive way. This is exemplified in Figs. 3.34 and 3.35 for a
thermally aged EPDM rubber material, which was analyzed with a portable permanent 0.31 T magnet
(NMR-mouse) and NMR console (Pourmand et al. 2016). The measurement spot, 17 mm in diameter
(along the sensor surface) and ca. 200 μm thick, was located 10 mm above the sensor surface. This
made it possible to detect profiles of T2 relaxation times within the sample simply by moving the
sample relative to the sensor surface. In order to eliminate effects of a spatially inhomogeneous
magnetic field, a Carr–Purcell–Meiboom–Gill (CPMG) pulse sequence was used, and T2 was
determined by fitting the decaying echo amplitude to the echo time with a single exponential
equation. The EPDM rubber aged at 170  C in air lost extractives, was oxidized close to the surface
and was anaerobically crosslinked, and these processes lead to a lowering of the molecular mobility,
which was assessed by the NMR setup. The T2 profiles in Fig. 3.34 show that the unaged material had
a uniform high T2 along the sample thickness. With ageing, T2 decreased and showed a pronounced
gradient towards the outer surface due to oxidation of the material closer to the surface. The decrease
in T2 in the interior was due to anaerobic cross-linking. The sample also lost the extractable low molar
mass component, and this also had an impact on T2. By combining the NMR data with data obtained
138 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.34 NMR relaxation time (T2) as a function of distance from the surface of an EPDM sheet aged at 170  C in dry
air. The four curves shows data from different ageing times as shown in the graph. Drawn after data of Pourmand et al.
(2016)

Fig. 3.35 NMR relaxation time (T2) as a function of indenter modulus (Ei) for EPDM samples aged for different times
at 170  C in dry air. The ageing times are displayed in the graph. Drawn after data of Pourmand et al. (2016)
3.4 Other Spectroscopic Methods 139

Fig. 3.36 The principle and a sketch of the instrumentation of X-ray photoelectron spectroscopy

by IR spectroscopy and thermal analysis, it was possible to establish the kinetics of the three
processes (Pourmand et al. 2016). Figure 3.35 shows the linear relationship between indenter
modulus (Ei) and T2. One of the advantages of the NMR method is that it is non-invasive.

3.4 Other Spectroscopic Methods

3.4.1 X-ray Photoelectron Spectroscopy (XPS)

This method has two different names: X-ray photoelectron spectroscopy (XPS) and electron spec-
troscopy for chemical analysis (ESCA). Figure 3.36 shows a sketch of the experimental configuration
and of the basic principle of XPS. The sample is irradiated by monochromatic X-rays, and commonly
used sources are Mg Kα (1254 eV), Al Kα (1487 eV) and Ti Kα (4510 eV). The sample is in vacuum
(10–5 Pa). The high-energy X-rays are able to detach electron from the different elements, since the
X-ray energies are greater than the usual ionization energies of the electrons. The kinetic energy
spectrum (Ukin) of the emitted electrons is recorded by a detector, and the detector keeps also track of
the number of electrons as a function of Ukin.
The energy required to liberate an electron from an atom, the ionization energy (Ub), is obtained
according to (cf. Fig. 3.36):

U b ¼ hν  Ukin ð3:27Þ

where hν is the energy of the X-ray radiation. Equation (3.27) is an idealized equation; each
spectrometer has a certain work function (ϕ), and this needs to be included in order properly to
calculate Ub:

U b ¼ hν  U kin  ϕ ð3:28Þ

Polymers which have a low electrical conductivity become positively charged when electrons are
emitted and this field cause an attractive force which reduces Ukin. This effect has to be quantified by
calibration, and the result is expressed in terms of the work function (ϕ).
Figure 3.37 shows a survey spectrum (a sketch) which presents an overview of the elements in the
sample. Each peak is labelled with the responsible element and the particular electron configuration.
For instance, oxygen is present in the example spectrum in two different peaks: an intense oxygen
peak is labelled O 1s, which has a fairly high ionization energy (~550 eV), and a less intense peak is
140 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.37 Sketch of a survey spectrum showing the peaks from electrons of a selection of elements

Fig. 3.38 XPS spectrum

(C 1S) of trifluorobenzene
with peak assignments.
Drawn after Clark (1976)

labelled O 2p with a low ionization energy. Carbon and silicone also appear with two different
electron configurations, where the intensity of each peak is proportional to the concentration of the
element in the sample volume, the XPS samples being from an extremely thin surface layer, only a
few nm thick. All elements except hydrogen can be detected, and this is a drawback for polymer
studies because organic polymers contain substantial amounts of bonded hydrogen. In order to obtain
the correct atomic percentage values for a given sample, each raw signal value is divided by a
so-called relative sensitivity factor, and finally, all the data are brought together to achieve normali-
zation. It should be remembered that hydrogen is not on the element list (with atomic percentages)
although a lot of hydrogen is present in such sample.
XPS is capable of providing detailed information about the chemical structure, as is illustrated by
the three examples presented in Figs. 3.38, 3.39 and 3.40. The electronic structure is affected by the
surrounding elements, and these studies require another adjustment of the pass energy level of the
spectrometer. Figure 3.39 shows the carbon 1s spectrum for a low molar mass compound (trifluor-
obenzene); note the narrow Ub band; the two carbon peaks differ only by 10 eV. The aromatic carbon
atoms are six in number, and the carbon bonded to fluorine is only one. This number ratio appears also
3.4 Other Spectroscopic Methods 141

Fig. 3.39 XPS spectra of

poly(ethylene
terephthalate) showing O
1S (left) and C 1s (right)
spectra with peak
assignments. Drawn after
Clark (1976)

Fig. 3.40 XPS data

showing Si 2p spectra of
silicone rubber samples: (a)
a fresh, unaged sample; (b)
a sample exposed to MW
oxygen plasma. This
spectrum were resolved
into three peaks: 102.1,
102.8 and 103.4 eV. Drawn
after data of Hillborg et al.
(2000)

in the areas (A) under the two peaks: A(286 eV)/A (294 eV)  6. The two peaks are also well resolved
and symmetrical.
The O 1s and C 1s regions in the high-resolution spectra of poly(ethylene terephthalate) are even
more compact than that of trifluorobenzene (Fig. 3.39). Three different carbon atoms are present: in
the phenylene group, in the ester unit and adjacent to oxygen in an ether link; and the number ratios of
them are 6:2:2 per repeating unit, and thus the areas should be 3:1:1. The assessment of these areas
requires curve resolution, however, because two of them show spectral overlapping. The oxygen
spectrum shows overlapping for the two different oxygens. However, it seems without careful
analysis that the peaks are of the same size.
Figure 3.40 shows high-resolution spectra (Si 2p) of silicone rubber samples. Unaged rubber
consists only of poly(dimethyl siloxane) (PDMS) where the silicon is bonded to two methyl groups
and two oxygen atoms. Notice that the peak is symmetrical about 102.1 eV, whereas a sample which
has been exposed to oxygen plasma has been oxidized and the Si 2p peak has a pronounced
asymmetric character. The two additional peaks at 102.8 and 103.4 eV correspond to silicon bonded
to three or four oxygen atoms. Overall this structure is referred to as SiOx. One problem in this
analysis is that low molar silicone oil (some of which is naturally present and some of which is formed
by degradation) migrates to the surface of the specimen, and it must be removed by vacuum treatment
before the XPS analysis.
In conclusion, XPS reveals the elemental composition and also the structural surroundings of the
detectable elements. XPS also provide knowledge about the electronic structure. XPS is a high-
vacuum technique, and samples containing volatiles are difficult to study. Recommended texts on
XPS are those by Briggs (1998) and Watts and Wolsenholme (2020).
142 3 Spectroscopy and Scattering of Radiation by Polymers

3.4.2 Electron Spin Resonance Spectroscopy

Electron spin resonance (ESR) spectroscopy provides information about the concentrations and types
of radicals. A radical has a spin, which is closely related to a magnetic moment. In the absence of an
external magnetic field, all radicals (unpaired electrons) have the same energy, but in an external
magnetic field (magnetic field strength ¼ B), the two possible spins, parallel and anti-parallel, have
different energies (Fig. 3.41), and the energy difference (ΔU ) is proportional to the strength of the
applied magnetic field according to:

ΔU ¼ ge μB B ð3:29Þ

where ge is the free electron g-factor (2.0023) and μB is Bohr’s magneton. The effect of the magnetic
field and the condition for resonance are shown in Fig. 3.41.
Figure 3.42 shows how the ESR spectrum is obtained given that the frequency of the microwave
radiation is kept constant and the magnetic field strength is gradually increased; the change in
absorbance (A) is measured as a function of the change in B, i.e. dA/dB is assessed.
The structure of the surrounding atoms affect the ESR spectrum, which enables the structure of the
radicalized molecule to be assessed. Figure 3.43 shows how the number of adjacent hydrogen atoms
affects the carbon-centred radical; the effect is referred to as isotropic, independent of the direction of
the applied magnetic field, and it originates from the magnetic moments of the protons (Rånby and
Rabek 1977).

Fig. 3.41 Energy of

unpaired electrons with
parallel and antiparallel
spins in a magnetic field as
a function of field strength
(B). The condition for
resonance (energy
absorption of the
electromagnetic radiation)
is sketched: hν ¼ geμBB

Fig. 3.42 The resonance

region shown as the
absorbance (A) as a
function of the magnetic
field (B) and the first
derivative (dA/dB) as a
function of B
3.4 Other Spectroscopic Methods 143

Fig. 3.43 Splitting of spectrum by interaction between radical and nearby protons. Drawn after Rånby and Rabek
(1977)

ESR spectroscopy is used in studies where radicals play an important role such as radical
polymerization, polymer degradation, radiation of polymers and fracture of oriented polymers,
where a sufficient number of molecular fractures occurs. Nitroxide radicals are unusually stable
and are used in studies of molecular mobility. Recommended texts about ESR spectroscopy are those
of Rånby and Rabek (1977) presenting the early contributions, Kausch (1978) presenting a critical
review of polymer fracture including the ESR findings and Bertrand (2020), which is a modern
comprehensive text.

3.4.3 UV-VIS Spectroscopy

Ultraviolet-visible (UV-VIS) spectroscopy is primarily used to solve chemical problems. Absorption

of visible light and UV radiation is controlled by the chemical bonds.
Sigma bonds absorb in the far UV range (λ < 200 nm). Polyethylene with only sigma bonds shows
absorption peaks at 146 nm and 180 nm, without any difference between HDPE, LDPE and long-
chain paraffins (Kl€opffer 1989). Hence, chain length and chain branching have no significant impact
on the spectrum in the far UV range. Far UV spectroscopy is not to be confused with UV-VIS
spectroscopy, and it is a much more difficult technique because it can only be conducted in vacuum
and almost no solvents are transparent to this radiation. UV-VIS spectroscopy covers the spectral
range of near UV radiation (λ ¼ 200–400 nm) and visible light (λ ¼ 400–700 nm). In this spectral
range, the work can be done in air, and π electrons have absorption. The groups which are of primary
interest are units with conjugated double bonds, phenyl groups, vinyl and vinylene groups and
carbonyl groups. UV-VIS spectroscopy provides information about remaining unsaturation of
thermosets or in some thermoplastic polymers. The concentration of photopolymerization initiators,
144 3 Spectroscopy and Scattering of Radiation by Polymers

photo stabilizers and antioxidants can be determined based on the absorption of the radiation by
applying Lambert–Beer’s law (I ¼ I0  10–εcd, where I is intensity of the radiation at depth d, I0 is the
incoming intensity of radiation, ε is the molar absorption coefficient and c is the molar concentration,
cf. Sect. 3.2.3). The underlying theory for UV-VIS spectroscopy is given by Murrell (1963) and
Perkampus (1992); the latter also provides details about the spectrometer instruments.

3.5 Scattering Methods

The scattering methods usually applied to characterize polymers is light scattering primarily for the
assessment of molar mass and the coil dimension of polymers in solutions (Sect. 3.5.1); wide-angle
X-ray scattering (WAXS) for the determinations of crystal unit cell, crystallinity and crystal size and
perfection (Sect. 3.5.2); small-angle X-ray scattering for the assessment of long period, crystal
thickness and crystallinity (Sect. 3.5.3); electron diffraction which is only covered brief in Sect.
3.5.4; and finally neutron scattering, which is a method to characterize the conformation of polymers
in condensed states on different length scales (Sect. 3.5.5). Some of these methods are also useful to
assess chain orientation of whole samples or part of the structure such as the crystalline phase in
semicrystalline polymers. This topic is treated in a chapter of the companion volume, cf. Gedde and
Hedenqvist (2019a). A theoretical background, description of instrumentation, useful methods,
application cases and useful references are provided in the sections.

3.5.1 Light Scattering

First briefly about light scattering, the cause of it. Light is an electromagnetic wave motion. The
electric and the magnetic fields are perpendicular to the propagation vector of the light. The electric
field strength is given by:
 
2πct
EðtÞ ¼ E0  cos ð3:30Þ
λ

where E0 is the amplitude (i.e. maximum electric field strength), c is the velocity of light and λ is the
wavelength. Note that f ¼ c/λ, f is the frequency (given in s–1). The variable electric field induces
polarization (P) of the scattering molecule according to:
 
2πct
P ¼ α  E ¼ α  E0  cos ð3:31Þ
λ

where α is the polarizability. The time-dependent-induced dipole moment causes a time-dependent

electromagnetic field, a secondary field, which propagates in a spherical fashion from the molecule. It
is important to note that the secondary field has the same frequency as the incident light, i.e. the
scattered electromagnetic wave has the same energy as the incident wave; this is referred to as elastic
light scattering. The same principle obeys for both visible light and X-rays. The same scattering
phenomenon occurs simultaneously in other nearby molecules which gives rise to interference
between the different rays. The different electromagnetic waves are unaffected by each other, and
the resulting beam intensity is simply the sum of the nearby waves. For instance, if the phase shift is
an integer of the wavelength, the intensity is maximized. If, on the other hand, the phase shift is half
3.5 Scattering Methods 145

Fig. 3.44 (a) The

scattering angle (denoted θ)
is defined as is done in this
section; (b) this
representation of θ is
commonly done to
illustrate Bragg’s law. The
angle θ is a glancing angle;
the scattering angle is here
2θ

the wavelength, then the scattered intensity is zero. The magnitude (‘length’) of the scattering vector
Q is:



4πn θ 4π θ
Q¼  sin ¼  sin ð3:32Þ
λ0 2 λ 2

where n is the refractive index of the medium, λ0 is the wavelength of the light in vacuum and θ is the
angle between the incident beam and the scattering direction. Figure 3.44 shows two different ways to
represent θ.
The scattering vector has the unit m–1. The wavelength in the medium (λ ¼ λ0/n) is inserted in
Eq. (3.32). Equation (3.32) is applicable to both the scattering of visible light and X-rays. A common
way of expressing scattering data is through the function denoted differential scattering cross-section
per unit volume of sample (Σ (Q or θ)):

X I ðQ or θÞ  r 2
ðQ or θÞ ¼ ð3:33Þ
I0  V

where I (Q or θ) is the scattered intensity at angle θ (or at Q); the two are related according to
Eq. (3.33) at a certain radial distance (r) from the sample. The magnitude of scattering is dependent
on different characteristics of the matter: for light scattering, the refractive index, and for X-ray
scattering, the electron density.
The important equations relating the scattered light intensity as a function of the scattering angle
for polymer molecules in solution emerge from a very lengthy derivation (recommended reading is
Burchard (1982) and Boyd and Phillips (1993)), and in this section, only the main results are
displayed.
The early work by Strutt (Lord Rayleigh) formed a foundation of the work carried out which
establish the relation between the scattered intensity (I (θ) as a function of the scattering angle (θ) and
the molar mass (M ) and the radius of gyration (s). Thus, a quantity used is the Rayleigh ratio (Rθ)
which is defined according to:


2
I ðθ Þ r
Rθ ¼   Fð θ Þ ð3:34Þ
I0 Vs

where I0 is the intensity of the incident beam, r is the distance between the scattering object and the
place where the scattered intensity is measured, Vs is the scattering volume and F(θ) is experimentally
related geometry factor. It should be noted that a requirement for the elastic Rayleigh scattering is that
146 3 Spectroscopy and Scattering of Radiation by Polymers

the scattering centres should be much smaller than the wavelength of the light, which is the case for
dilute polymer solutions; the radius of gyration is less than 10 % of the wavelength of the visible light.
Another requirement is that polarizability of the polymers segments is different from that of the
solvent molecules. This is often expressed by the change in the refractive index with increasing
polymer concentration. Another requirement is that each photon is elastically scattered only once.
The lengthy derivation of the scattered intensity as a function of polymer structure (cf. Boyd and
Phillips 1993; Burchard 1982) yields the following expression:

Rθ ¼ K s cðM  PðQÞ þ V s  N A  c  SðQÞÞ ð3:35Þ

where Ks is a factor describing the change in the refractive index of the solution (ns) with increasing
polymer weight fraction (c) and P(Q) is a scattering function describing the effects from intramolec-
ular interaction; thus function approaches value 1 as Q approaches zero (small scattering angle).
S(Q) describes the effect from intermolecular interaction on the scattered intensity. The Ks factor is
given by the following equation:
!

2
1 ∂ns
Ks ¼  2πn0  ð3:36Þ
N A λ0
4 ∂c

where NA is the Avogadro’s number. It may be noticed that if the refractive index of the solution
remains constant with increasing polymer concentration, then Ks would be zero and so would Rθ and
I(θ) be. Hence, the requirement for light scattering is that the polymer segments have another
polarizability than the solvent. Low Q conditions and the dilute solutions with minimum intermolec-
ular interaction imply that P ¼ 1 and S ¼ 0, which inserted in Eq. (3.35) yields:

Rθ
¼ cM ð3:37Þ
Ks

Polymers are never single molar mass, and each molecule contributes to Rθ (i.e. to the scattered
intensity) according to its mass fraction, which then yields to following equation:
Rθ
¼ cMw ð3:38Þ
Ks

This is why light scattering yields the weight average molar mass. The low scattering angle region
(Q ! 0) provides direct information about molar mass and the radius of gyration (s). The perfect
simplicity is when P ¼ 1 (Q ! 0) and S ¼ 0 (c ! 0); the low Q region requires further information
how P varies with Q, which was solved for random coils by Debye (1947):
   2   1=2
2
PðvÞ ¼ 4  ev þ v2  1 ; v ¼ Q  s2 ð3:39Þ
v

P is thus a direct function of v which is turn is a function of Q and s. It may be noticed that when
Q ! 0, then v ! 0. The function P (v) is expressed in a series form near v ¼ 0 (i.e. for Q ¼ 0):

v2 
Pð v Þ ¼ 1  þ . . . ¼ 1  Q2 s2 =3 þ . . . ð3:40Þ
3

which after insertion in Eq. (3.35) yields:

3.5 Scattering Methods 147

Fig. 3.45 Schematic Zimm plot showing extrapolation to reveal weight average molar mass, the radius of gyration and
the second virial coefficient

 !
  s2 Q2
Ksc 1
¼  1þ þ ... ð3:41Þ
Rθ M 3

where Q2 can be expressed in accordance with Eq. (3.32) as:

 

4πn0 2 2 θ
Q ¼
2
 sin ð3:42Þ
λ 2

In a plot of Ksc/Rθ as a function of x ¼ (4πn0/λ)2sin2(θ/2), the intercept is 1/M w (i.e. M w ¼1/

intercept), and the slope is hs2i/(3Mw) (hs2i ¼ slope  3Mw). The concentration dependence of Ksc/Rθ
is described according to Burchard (1982):
Ksc 1
¼ þ A2 c þ . . . ð3:43Þ
Rθ M

where A2 is a coefficient (referred to as the second virial coefficient). In a plot of Ksc/Rθ versus c, the
intercept is 1/M w and the slope is A2. The two different extrapolations are combined in the well-
known Zimm plot (Zimm (1948; cf. Fig. 3.45).
This so-called static light scattering is thus a powerful technique to reveal molar mass and
relationship between molar mass and the radius of gyration for dilute polymer solutions (cf. Gedde
and Hedenqvist 2019b). Another relate technique is dynamic light scattering which can assess
particles down to very small size, 1 nm. Berne and Pecora (2000) is a very nice textbook on this topic.
148 3 Spectroscopy and Scattering of Radiation by Polymers

3.5.2 Wide-Angle X-ray Scattering

Wide-angle X-ray scattering (usually abbreviated WAXS) has been used to study polymers since the
birth of the polymer concept. The regular packing of atoms in polymer crystallites is the major topic.
Deviation from the perfect order, i.e. the limited crystal size and the internal crystal disorder, has also
been an important topic (Baltá-Calleja and Vonk 1989). Semicrystalline polymers are not 100 %
crystalline, and essentially all properties of interest depend strongly on the degree of crystallinity
(cf. Gedde and Hedenqvist 2019c), which is most directly assessed by WAXS. One of the early
driving forces for polymer science concerned the possible synthesis of polymers that could be
transformed into a highly oriented state, i.e. the grand synthetic fibre programs. Indeed, the assess-
ment of the orientation of polymer crystals is perfectly done by WAXS. This is a topic comprehen-
sively dealt with in the companion volume (Gedde and Hedenqvist 2019a). The detailed structure of
amorphous polymers, i.e. the local packing of atoms, is also studied by WAXS and electron
diffraction (Fitzspatrick and Ellis 1973; Lowell et al. 1979; Mitchell 1984). The scattering angular
range covered by WAXS is from 5 and higher, which for Cu Kα1 radiation (λ ¼ 1.544 Å)
corresponds to repeating distances from 18 Å and shorter; for scattering angles (2θ): 5 – 18 Å,
10 – 9 Å; 20 – 4.4 Å, 40 – 2.3 Å and 60 – 1.5 Å.
X-rays are high-energy electromagnetic radiation with a wavelength range between 0.1 and 100 Å.
A useful equation relating photon energy (E in keV) and wavelength (λ in Å) is:

hc 12:4
E ¼ ) E ð3:44Þ
λ λ

where h is Planck’s constant and c is the velocity of light. A typically wavelength of X-rays used in
scattering studies is 1.5 Å, which corresponds to a photon energy of 8 keV. X-rays are traditionally
generated in high-vacuum sealed X-ray tubes in which a high voltage is applied between a cathode
and an anode target made of a specific metal (Cu in Fig. 3.46). The high-energy electrons cause the
anode to emit X-rays consisting a broad component and a few distinct X-ray peaks associated with
transition between particular electronic states. For Cu, it is Kβ1 (K ! M3), λ ¼ 1.392 Å and Kα1 (K !
L3), λ ¼ 1.540 Å and Kα2 (K ! L2), λ ¼ 1.544 Å. The latter two forms marge into one peak in the
diagram (Fig. 3.46)
A drawback of the sealed X-ray tubes is the dissipated heat, typically 1 kW on a few (mm)2, which
has to be transferred to a flowing cooling water system. X-ray tubes equipped with a rotational anode
are an elegant solution to this problem, and they allow much higher power to be used. Synchrotron
radiation, which include visible light, UV radiation and X-rays, is available since almost 50 years.
The intensity of an X-ray beam from a synchrotron source is much higher than that obtained from the
conventional X-ray tubes. Other advantages with the synchrotron X-rays are the high collimation and
the high polarization compared to the X-rays obtained from conventional sources.
The X-rays used for diffraction work has to be monochromatic, and since all the sources provide
radiation with a wavelength dispersity, a monochromator must be used. Single crystal
monochromators (e.g. LiF, MgO, NaCl, calcite and quartz) use a reflection, according to Bragg’s
law, to emit only one wavelength at a certain angle. Absorption filters such as a combination of Ni and
Co (absorption edges) are also used to obtain monochromatic CuKα radiation.
The X-ray diffraction patterns have been historically revealed by cameras using a photographic
film. The developed film shows dark and bright parts based in the X-ray irradiation which are
converted into an intensity scale using a densitometer. Further information about photographic
films and densitometers has been presented by Baltá-Calleja and Vonk (1989) and Vonk and
Pijpers (1981).
3.5 Scattering Methods 149

Fig. 3.46 Emitted X-rays from a Cu-target

Fig. 3.47 Different X-ray cameras used for WAXS

The flat film camera (Laue camera) uses a flat photographic film, which in Fig. 3.47a is placed
according to the transmission arrangement. The photographic film can be placed on the other side of
the sample and thus recording according to the back reflection mode. The X-ray source includes a
monochromator. The samples can be isotropic samples which yield concentric rings revealing the
different diffracting (hkl) planes. Anisotropic specimens like fibres show only equatorial and polar
sectors of the rings from which the c-axis orientation can be determined (cf. Gedde and Hedenqvist
2019a). The Debye–Scherrer camera, a cylindrical camera, shown in Fig. 3.47b, mimics a camera
developed by Buerger (1945). The diffracted beams with low scattering angles are recorded by the
photographic film positioned near exit tube intersection with the cylinder, whereas the diffracted rays
150 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.48 Sketches of diffractometer designs: (a) principal sketch of a diffractometer; (b) Johann type of powder
diffractometer. Drawn after Baltá-Calleja and Vonk (1989)

with high scattering angles are recorded near the entrance of the receiving tube. Two different
focusing cameras are displayed in Fig. 3.47c and d. The fact that the two displayed beams shown
in Fig. 3.47c have the same angle between the two different beams entering and exiting the curved
sample was proven by Euclid already in 300 BC. Details about the Guinier–H€agg camera are
provided by H€agg (1947). The Guinier camera (for full details, cf. Guinier 1952) shown in
Fig. 3.47d uses a curved single crystal monochromator with the sample positioned at the periphery
of the cylinder.
X-ray diffraction patterns are also revealed in diffractometers (Fig. 3.48). The different scattering
angles are reached by both rotating the sample and the detector with respect to the X-ray source but
with different angular rates. These equipment are using detectors often referred to as counters
(gas-filled counters, scintillation counters or solid-state detectors), which convert each individual
X-ray photon into a voltage. Thus, these detectors yield a precise measure of the scattered intensity.
An excellent text providing details about the X-ray intensity detectors is the textbook of Baltá-Calleja
and Vonk (1989). Figure 3.48b shows the Johann type of powder diffractometer (cf. Johann 1931).
The sample is irradiated by the collimated raw X-rays, and the scattered beam is reflected through
slits and onto a bent crystal monochromator, and only the monochromatized beam reaches the X-ray
counting detector. The 2θ space is covered by coordinated rotations of the monochromator crystal and
the detector.
The development of position-sensitive detectors that not only provide the intensity associated with
a single 2θ-value but also provide the value of the scattered intensity as a function of 2θ has been a
major force in the development of X-ray scattering. Both one-dimensional and two-dimensional
counters are available, and the latter replace essentially photographic film. Baltá-Calleja and Vonk
(1989) and He (2018) provide further details on this topic.
The focus of this chapter is how to use WAXS for solving problems in polymer science and
engineering. This how-to approach is brief and leaves out lengthy derivations which can be found
elsewhere. However, references are provided which help the reader to find such vital information.
Some basic concepts need to be introduced before dealing with the problems. WAXS is directly (not
3.5 Scattering Methods 151

Fig. 3.49 The 14 Bravais lattices, from Gedde and Hedenqvist (2019c)

exclusively however) related to crystal structure. The crystal structures can be divided into 14 Bravais
lattices (Fig. 3.49), and they include 7 crystal systems which are expressed by the unit cell vectors
a, b and c and by the angles between these vectors: α ¼ ∠bc; β ¼ ∠ac; and γ ¼ ∠ab. A unit cell
defined by the vectors is a repeating unit structure of the three-dimensional crystal. Some of crystal
systems are Cartesian with only 90 angles. The crystal systems are (i) cubic, a ¼ b ¼ c, α ¼ β ¼ γ ¼
90 ; (ii) tetragonal, a ¼ b 6¼ c, α ¼ β ¼ γ ¼ 90 ; orthorhombic, a 6¼ b 6¼ c, α ¼ β ¼ γ ¼ 90 ;
hexagonal, a ¼ b 6¼ c, α ¼ β ¼ 90 ; γ ¼ 120 ; trigonal, a ¼ b ¼ c, α ¼ β ¼ 90 ; γ ¼ 120 ; monoclinic,
a 6¼ b 6¼ c, α ¼ γ ¼ β 6¼ 90 ; and triclinic, a 6¼ b 6¼ c, α 6¼ β 6¼ γ 6¼ 90 . The convention for polymer
crystals is that the chain axis is the c-axis with one exception; the monoclinic cell where the chain axis
is the unique axis (b-axis). Unit cells with only one unique motif are called primitive. It is always
possible to generate a primitive cell from a given lattice, but in many cases, end-, face- or body-
centred representations are better because they show greater symmetry than the primitive cell.
Crystals exhibit symmetry which are expressed in different symmetry operations: rotation about an
axis, inversion about a point, combination of rotation and inversion about an axis, reflection across a
152 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.50 Lattice plane index system explained by a generic instruction (i) and three examples (ii–iv). From Gedde and
Hedenqvist (2019c)

plane, combined translation and rotation along an axis (screw axis), combination of translation within
a plane and reflection across the same plane (glide plane). These symmetry operations add up to
230 space groups for the 14 Bravais lattices. The point groups (totally 32 possible types) are another
set of symmetry operations. This compact information can be further enjoyed by the wonderful text of
Hammond (1997). Don’t miss this text!
The labelling of planes and directions in the crystals are obligatorily used in conjunction with the
interpretation and analysis of X-ray diffraction results. A direction with reference to a given unit cell
is given as the digits originating from the vector addition of the unit cell vectors a, b and c:
r¼uaþvbþwc ð3:45Þ

where u, v and w are the smallest integer numbers describing the direction of the vector. Example 1:
u ¼ 4, v ¼ 2, w ¼ 0. These number are written with square parenthesis like [420]. There is one mistake
here; the number should be integers and as small as possible. The number can be divided by 2 to yield
[210]. Example 2: label the three vectors a, b and c. They are respectively [100], [010] and [001].
Example 3: label all direction in the ab-plane. [uv0]. Example 4: label all directions possible;
[uvw]. The planes are termed according to Fig. 3.50. Figure 3.50 (i) illustrates the principle. The
first plane (pl. 1) is intersecting the origin (0,0,0). The next plane (pl. 2) is intersecting the three axes
at xi ¼ a/h (x or a-axis), yi ¼ b/k (y or b-axis) and zi ¼ c/l (z or c-axis). The solutions to the three
equations have to be integer numbers, and they are expressed as (hkl).
The first two examples shown in Fig 3.50, ii and iii, are simple: (110) and (010), whereas in the last
example, the second plane intersects the b-axis at b/2, which then yields the notation (020). Thus, the
group of planes denoted (010) is a subset of (020). The lattice plane system thus provides the
orientation of the planes (given by the normal) and the distance between two adjacent planes of the
group. One more illustrating example: suppose plane  2 intersects
 the b-axis at yi ¼ –1 and that xi ¼ zi
¼ 1. This plane is not labelled (0–10) but instead 0 1 0 . A minus sign symbolized with a bar above
3.5 Scattering Methods 153

Fig. 3.51 Constructive interference of diffracted X-beams from two adjacent (hkl) planes

the number. All surfaces of a cubic unit cell are denoted {100}. This group notation includes the
following six planes.
     
ð1 0 0Þ, 1 0 0 , ð0 1 0Þ, 0 1 0 , ð0 0 1Þ, 0 0 1 ð3:46Þ

The well-known equation relating the distance (dhkl) between adjacent (hkl) planes and the
glancing angle (θ) for constructive interference between the scattered beams from the two adjacent
layers was derived by Bragg and Bragg (1913); the difference in length for the two beams is for a
given value of θ equal to 2dhkl sin θ ¼ L21 + L13 (Fig. 3.51). If this length is equal to λ or actually nλ,
where n is an integer, the interference between the two beams is fully constructive. This is the basis of
Bragg’s law:

2 dhkl sin θ ¼ nλ ð3:47Þ

The drawing shown in Fig. 3.51 assumes that the two layers of atoms are matched along the layers.
However, if one of the layers is shifted horizontally with respect to the other layer, it has no effect on
the path difference between two beams diffracted by the two atoms; it remains 2dhkl sin θ.
The calculation of dhkl is carried out provided that the unit cell type and the dimensions are known
(Hammond 1997):

1 h2 k 2 l 2
2
¼ 2þ 2þ 2 ð3:48Þ
d hkl a b c

which is valid for orthorhombic (a 6¼ b 6¼ c), tetragonal (a ¼ b 6¼ c) and cubic (a ¼ b ¼ c) unit cells,


1 4 h2 þ hk þ k2 l2
2
¼  2
þ 2 ð3:49Þ
dhkl 3 a c

which is valid for the hexagonal unit cell (a ¼ b 6¼ c),

 2 
1 h þ k2 þ l2  sin 2 α þ 2  ðhk þ kl þ hlÞ  ð cos 2 α  cos αÞ
¼ ð3:50Þ
d2hkl a2  ð1  3 cos 2 α þ 2 cos 3 αÞ

which is valid for the rhombohedral unit cell and

154 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.52 Unit cell of polyethylene (in blue) and two sets of (hkl) planes indicated, (200) in grey and (110) in red

Fig. 3.53 WAXS patterns (CuKα radiation) of two polyethylene copolymers with (a) a hexyl-branch content of 0.4 per
100 main chain carbon atoms and (b) a hexyl-branch content of 3.6 per 100 main chain carbon atoms. Peak
assignments: A, amorphous component; B, interfacial component linked to (110); C, orthorhombic (110); D, interfacial
component linked to (200); E, orthorhombic (200); and F, monoclinic phase. From Mattozzi et al. (2010) with
permission from Elsevier, UK

2 
1 1 h k2 sin 2 β l2 2hl cos β
¼  þ þ 2 ð3:51Þ
d2hkl sin 2 β a2 b2 c ac

which is valid for the monoclinic cell (a 6¼ b 6¼ c and α ¼ γ ¼ 90 ; β 6¼ 90 ).

The use of WAXS in a study of a series of polyethylene copolymers before and after saturation of
the polymers in n-hexane at room temperature is used as an illustrating example to demonstrate
3.5 Scattering Methods 155

changes of the unit cell dimensions, to reveal an interfacial component (neither crystal nor amor-
phous) and also to note polymorphism. The crystal unit cell with the precise dimensions of linear
polyethylene is known, and the effect of chain branches has also been reported in great many studies.
Figures 3.52 and 3.53a, b shows the orthorhombic unit cell of polyethylene together with the WAXS
scattering patterns for two polyethylene materials with different branch contents; the branches are
hexyl groups. There are two (hkl) planes that give rise to strong reflections in the 2θ-range between
10 and 30 , (110) and (200) (Fig. 3.52). The interplanar distances are calculated according to
Eq. (3.48) using the a and b values for linear polyethylene shown in Fig. 3.52: d110 ¼ 0.4115 nm
and d200 ¼ 0.3703 nm, which corresponds respectively to a scattering angle (2θ) of 21.63 and 24.07 .
The scattering displayed in Fig. 3.53 consists of four symmetrical components, amorphous
component, orthorhombic (110) and (200) crystal planes and monoclinic phase (assigned according
to Pezzutti and Porter 1985), which were fitted with Pearson VII functions. This function, which is
commonly used for curve fitting of X-ray data, provides symmetric peaks about the maximum that
can be adjusted from Lorentzian to Gaussian depending on the value of one of the adjustable
parameters (Gupta 1998). The fittings left additional reflections at the low 2θ sides of the symmetrical
orthorhombic (110) and (200) reflections, which according to Baker and Windle (2001) can be
assigned to an interfacial component. The (110) and (200) peaks are shifted in the branched polymers
with respect to those of linear polyethylene (LPE): (110) – 21.6 (LPE), 21.5 (0.4 % branches) and
21.2 (3.6 % branches); (200): 24.1 (LPE), 23.9 (0.4 % branches) and 23.5 (3.6 % branches). The
crystallinity of the copolymer shown in Fig. 3.53b was only 20 wt.% of which half is in a monoclinic
phase. The presence of an interfacial component in this polymer was not revealed (Fig. 3.53b). The
peak position of the amorphous halo (peak labelled A) is at 20 (2θ) in the low-crystallinity polymer
(Fig. 3.53b) and at 21 (2θ) in the high-crystallinity polymer.
The steady expansion of the orthorhombic unit cell accompanying the addition of hexyl branches
is evident for both (110) and (200). Note that the c-dimension remains constant; all expansions of the
polymer unit cells are along a- and b-axes. The addition of 4 branches per 100 backbone carbon atoms
expands the a from 0.741 nm (LPE) to 0.758 nm, which places them in the same category as the
longer branches (butyl and longer) according to data collected by Baltá-Calleja and Vonk (1989), but
the cell expansion in methyl-branched PE is much stronger (Swan 1962). The saturation of these
polymers in n-hexane causes a densification of the crystal unit cell, which is surprising in view of the
fact that no detectable crystal melting is induced by this solvent at room temperature (cf. Fig. 3.54).
The shrinkage in the a-dimension is from maximum 0.758 nm in pristine polymer to 0.750 nm in
corresponding n-hexane-saturated polymer. It is possible that the increase in free volume resulting
from the uptake of the solvent makes possible release of stresses in the interfacial region caused by
the chain branches, which in turn allows a better packing of the crystalline stems. This assumption is
also supported by the data shown in Fig. 3.55, which show that the n-hexane-saturated samples
contains less interfacial component than the pristine samples.
The width of the crystalline peaks is an indication of the size of the crystal and the concentration of
crystal defects. The broadening of diffraction peak (e.g. (200) in the orthorhombic crystal phase of
polyethylene) is due to the finite size of the crystal orthogonal to diffracting planes according to the
Scherrer equation (Scherrer 1918):

Kλ
Dhkl ¼ ð3:52Þ
β cos θ

where K is the Scherrer shape factor, which is close to unity, λ is the wavelength of the X-rays and β is
the width at the half height of the diffraction peak in radians. A more advance theory, the
paracrystalline theory, which include peak broadening due to both finite crystal size and the presence
of internal crystalline defects was developed by Hosemann (1950a, b, 1963, 1982):
156 3 Spectroscopy and Scattering of Radiation by Polymers

(a) (b)
22.0 25.0

21.8
24.5

21.6

2 θ (°)
2 θ (°)

24.0

200
110

21.4 b b
a
23.5
21.2 a

21.0 23.0
0 2 4 6 8 10 0 2 4 6 8 10
Comonomer content (mol.%) Comonomer content (mol.%)

Fig. 3.54 Scattering angle at the centre of diffraction peaks as a function of the 1-octene comonomer content in
polyethylene of the following orthorhombic crystal planes: (a) (110); (b) 200. Open symbols represent data taken for
the pristine polymer (curve a); filled symbols represent data taken from samples saturated in n-hexane. Drawn after data
of Mattozzi et al. (2010)

Fig. 3.55 X-ray peaks

associated with the
interfacial component
(linked respectively to
(110) and (200)); (a)
denotes pristine
polyethylene copolymer
with 0.8 hexyl branches per
100 backbone carbon
atoms and (b) the same
polymer after being
saturated in n-hexane.
Drawn after data of
Mattozzi et al. (2010)

1 π 4 g4 4
β2s ¼ 2
þ n ð3:53Þ
Dhkl d 20

where βs is the breadth of the diffraction peak in s units (s ¼ (2 sin θ)/λ) after subtraction of the
instrumental broadening, d0 is the interplanar distance, n is the order of reflection and g is the degree
of statistical fluctuation of the paracrystalline distortions relative to the separation distance of the
adjacent lattice cell. The parameters of the equation can be obtained by fitting a plot of βs2 as a
function of n4. Further recommended reading about the paracrystalline theory and its use in practice is
the textbook of Baltá-Calleja and Vonk (1989).
Two important uses of WAXS for the study of semicrystalline polymers are only mentioned
briefly. One is the assessment of crystallinity (i.e. the mass or volume fraction of crystalline phase in a
3.5 Scattering Methods 157

Fig. 3.56 Sketch

illustrating the concept
radial distribution function
by four oligomer molecules
in the view field

polymer), which is a major topic in the companion volume Fundamental Polymer Science (Gedde
and Hedenqvist 2019c), and for that reason, it is not further treated here. We are impressed by the
careful text in Baltá-Calleja and Vonk (1989), especially on the Ruland theory and the personal
engagement by these authors. The paper by Vonk (1973) on the computerization of the Ruland X-ray
method is also a recommended reading. Another field of great importance is the structure assessment
(expressed in a unit cell) of new, emerging crystalline polymers. This is an advanced form of study
and requires careful X-ray work and modelling. It is known that polymers adapt to a low-energy
conformation in the crystalline phase. One task is thus to find this ‘golden’ structure by molecular
modelling from which one or several crystalline structures can be built. The packing of the crystalline
stems can differ depending on structural variations and crystallization conditions, as in case of
isotactic polypropylene (cf. Gedde and Hedenqvist 2019c). A recommended text on this topic is
Ladd and Palmer (2013).
Amorphous polymers either in the liquid state or as frozen-in structure in glasses have a structure
on a local scale which can be described by the radial distribution function. This structure can be
obtained by different methods of which WAXS is one such method and molecular dynamics
simulation is another method. Figure 3.56 shows a sketch illustrating a method to describe this
local structure using a quantity referred to as the radial distribution function.
The local structure is given a functional form by selecting a central unit (bead), which is marked in
the figure. In the first layer, no other bead is found; a bead is counted only if the centre of the bead is
located within the layer. The next layer position at an average radial distance (r) from the centre
contains two beads, which means that average density of beads at the radial distance r is 2/(4πr2Δr).
The next layer contains five beads, and the bead density is consequently 2/(4πr2Δr), not that the
r-value is larger than for the foregoing layer. This has to be repeated to cover a larger volume. The
radial distribution function gi(r) for each centre is given by:
ni ðr Þ
gi ð r Þ ¼ =ρ ð3:54Þ
4πr 2  Δr beads

where ρbeads is the average bead density. This calculation has to be repeated for great many different
centres (assume the total number being k) to obtain a statistically relevant radial distribution function:
158 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.57 Example of typical radial distribution function

P
k
gi ð r Þ
i¼1
gð r Þ ¼ ð3:55Þ
k

The following sketch shows some typical features of the radial distribution function. The first two
peaks are intramolecular, controlled by bond length, bond angle and torsional freedom, and the outer
waves are controlled by the secondary interactions. The local memory is the length until the average
density of beads is established, and it seems to be approximately five times the bead period. One of the
controls of the molecular dynamics simulation code and input force field parameters is to calculate the
radial distribution function (readily done by a subroutine) and to compare with data of the radial
distribution function obtained from WAXS data. (Fig. 3.57)
The procedure to obtain the radial distribution function (g(r)) from WAXS data starts with
obtaining the scattered intensity as a function of Q:
4π sin θ
I ðQÞ ¼ f ðQÞ; Q ¼ ð3:56Þ
λ

The total scattering structure function S(Q) is given by:


I ðQÞ  f 2 ðQ Þ comp
Sð Q Þ ¼ 1 þ 2
ð3:57Þ
h f ðQÞi comp

where f is atomic structure function, which is a function of Q. The average f function thus depends on
the elemental composition. The reduced total scattering structure function F(Q) is:
FðQÞ ¼ QðSðQÞ  1Þ ð3:58Þ

Finally the Fourier transform yields the radial distribution function:

3.5 Scattering Methods 159

Fig. 3.58 Normalized

differential radial
distribution function
(DRDF) for amorphous
poly(vinyl alcohol) at
260  C. Experimental data
by WAXS (shown in black;
Gupta and Yeh 1978) and
data by molecular
dynamics (MD) simulation.
Drawn from Karlsson et al.
(2002)

Qð
max
2
gð r Þ ¼  FðQÞ sin ðQr ÞdQ ð3:59Þ
π
Qmin

where in practice the Q-range considered must be limited; for synchrotron X-ray radiation, the
maximum value should be in the range 30 to 50 Å–1.
Figure 3.58 compares differential radial distribution data obtained by WAXS and molecular
dynamics simulation for amorphous poly(vinyl alcohol), and most of the peaks are found in both
sets of data. The region of less good agreement is between 3 and 4 Å.

3.5.3 Small-Angle X-ray Scattering

Small-angle X-ray scattering (SAXS) uses monochromatic X-rays, and the details about X-ray
sources and monochromators are found in Section 3.5.2. SAXS assesses scattering at very small
angles, 0.1 to 5 , which originates from motives with a repeating distance between 1 and 100 nm.
WAXS is primarily used for the study of the arrangement of atoms within the crystal unit cell,
whereas SAXS is used for the study of larger objects, such as crystal lamellae and amorphous
interlayers in crystal lamellae stacks, both are sized around 10 nm (cf. Gedde and Hedenqvist
2019c). Other objects relevant to polymer science and engineering are two-phase structures in general
such as polymer blends, particles in composites and nanocomposites, voids and crazes in polymers
(cf. Kausch 1978) and structures in biomacromolecules (cf. Wu and Xing 2018).
The most important for SAXS is to have a highly collimated X-ray beam, because the scattering at
very small angles is to be studied. One method to achieve this goal is to use pinhole collimation,
i.e. the use of several circular openings with a diameter between 100 and 500 μm. The scattering
pattern is in the case of oriented samples undistorted, but the drawback is the long exposure times due
to the large percentage of rejected X-ray photons. Figure 3.59 shows the geometry associated with a
two-pinhole collimator. Although the diameter of the two pinholes are small, perhaps only a few
hundred μm, the direct X-ray radiation is occupying a much larger area at the registration plane. The
beam stop at the registration plane should be of the order of 1 mm. Examples of pinhole-collimated
camera are the Kiessig camera (Kiessig 1942) and the Statton camera (Statton 1962), and there are
several more recent Japanese cameras commercially available. Slit collimation cameras use slits
160 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.59 The effect of two pinholes located ‘before’ the sample on the beam illuminating the registration plane.
Drawn inspired by Baltá-Calleja and Vonk (1989)

Fig. 3.60 Schematic representation of the Kratky camera

instead of pinholes and either photographic film or position sensitive counters for recording of the
SAXS pattern.
The slit collimation cameras are more efficient than the pinhole cameras; fewer X-ray photons are
discriminated, and also the use of modern position sensitive counters makes the data harvesting
efficient. There are excellent Japanese cameras (by Rigaku) and also the well-known Kratky camera
(Kratky 1954, 1956, 1958) with an outstandingly high resolution (Fig. 3.60); the Kratky camera can
measure scattering at extremely small angles (0.04 ) corresponding to periodic distances of a few
thousand nm (Kratky and Laggner 1987). This makes this camera useful for a wide range of
applications such as long period assessment in semicrystalline polymers and crazing and cavitation
in polymers.
Semicrystalline polymers crystallize usually in spherulites or axialites with internal stacks of
almost parallel lamellae. In between adjacent crystal lamellae, there is always a layer of amorphous
material (cf. Gedde and Hedenqvist 2019c). Several methods indicate also the presence of an
interfacial component (neither crystalline nor ‘perfectly’ amorphous): WAXS (Baker and Windle
2001; Mattozzi et al. 2010) and SAXS (Baltá-Calleja and Vonk 1989), Raman spectroscopy (Mutter
et al. 1993) and NMR (Kitamaru et al. 1977, 1986). SAXS is a standard method used for assessing
crystal thickness or more precisely the long period. The ideal structure is sketched in Fig. 3.61. The
long period is the sum of the crystal and amorphous layer thicknesses, LP ¼ Lc + La. If both the long
3.5 Scattering Methods 161

Fig. 3.61 Sketch of a stack of crystal lamellae. Definition of the long period (LP). The long period (LP) is calculated
from the scattering angle (2θ) according to Bragg’s law

period and the volume crystallinity (ϕc) are known, it is possible to calculate the thicknesses of the
individual components: Lc ¼ ϕcLP and La ¼ (1– ϕc)LP.
The requirement for a SAXS peak associated with the long period is that a periodic structure of this
kind is present and that is that the LP is near constant in the crystal lamellae stack. Figure 3.62 shows
four polyethylene blends based on a low molar mass linear polyethylene (LPE) and a high molar mass
branched PE. The sample with 80% LPE shows stacks of at least 30 crystals in the stacks and with
uniform LP (Fig. 3.62a). The sample shown in Fig. 3.62b (60% LPE) shows also stacks with uniform
LP; the stacks contain five to ten lamellae. The last two samples, containing 20% LPE and a sample
with only the branched PE, show much less developed stacks. The branched PE showed also crystals
with a more pronounced variation in thickness. It is questionable if these samples would show a peak
in the SAXS pattern.
Polymers with a broad distribution in molar mass, which are cooled from the melt to a fairly high
temperature and allowed to crystallize to some extent (say to 20%) at this particular temperature and
then rapidly cooled so that the remaining 80 % of the crystallization occur. Show crystals formed at
the high temperature which are thick (large Lc), whereas the crystals formed at the lower temperatures
are much thinner. Figure 3.63 shows a transmission electron micrograph of such a sample with a
bimodal crystal thickness distribution, and this sample is expected to show two SAXS peaks.
The procedure to obtain the long period and the crystal thickness from SAXS data is as follows:
(i) The raw data is first corrected from the recorded data in low-angle region by running an empty cell.
(ii) The effects from other density fluctuation such as particles, etc. need to be eliminated. This can be
done by melting the sample and take a diffractogram which is subtracted from the real SAXS.
Baltá-Calleja and Vonk (1989) have proposed empirical equations like B ¼ a + b(2θ)n, where a, b and
n are adjustable parameters to be determined obtained by fitting the equation to experimental WAXS
and SAXS data (low-angle region). (iii) An important next step is to perform desmearing of the SAXS
data, which is only required for data obtained from a slit-collimated instrument; details about
procedure are presented by Baltá-Calleja and Vonk (1989). Software for desmearing is available
together with the modern SAXS instruments. (iv) If the intensity (I) vs. 2θ shows a maximum, then
the Lorentz correction can be applied: Is2 is plotted as a function of s; s ¼ 2sinθ/λ. The multiplica-
tion of the intensity with s2 sets the y-axis to zero at s ¼ 0 (i.e. θ ¼ 0), and the maximum is shifted to
slightly larger s (θ) values by the Lorentz correction. The s-value associated the maximum yields after
application of Bragg’s law, the SAXS long period. Cser (2001) presents convincing arguments that
this correction produces artificial Bragg periodicity in systems that lacks a pronounced small-angle
peak in the uncorrected I – s plot. Returning to micrographs d in Fig. 3.62, it is indeed difficult to find
any long period regularity in this system. Not even the neighbouring crystal is correlated with the
reference crystal. In this micrograph, there are correlated dominant crystals, although not perfectly
162 3 Spectroscopy and Scattering of Radiation by Polymers

Fig. 3.62 Transmission electron micrographs of chlorosulphonated samples of blends of a low molar mass linear PE
fraction (Mw  2500 g mol–1) and a high molar mass branched PE (1.5 mol.% ethyl branches): (a) 80 wt.% linear PE;
(b) 60 wt.% linear PE; (c) 20 wt.% linear PE; (d) pristine branched PE. From Conde Braña and Gedde (1992) with
permission of Elsevier, UK

parallel, but in between them, segregated material is present, which at the later state crystallize in a
complex morphology, truly different from the regular structures shown in Fig. 3.62 a and b.
SAXS is used in many other application areas which can be found in several excellent textbooks:
Guinier and Fournet (1955), Glatter and Kratky (1982), Fergin and Svergun (1987), Tolan (1999) and
Stribeck (2007).

3.5.4 Electron Diffraction

Electron diffraction, which is carried out in an electron microscope, provides information about
crystal structures similar to that obtained by WAXS, but with the difference that size of the object
studied (the beam size) can be made very small so that the diffraction pattern from isolated lamella
can be studied. This is a great advantage of electron diffraction compared to WAXS. Typically great
many crystal lamellae are assessed by WAXS, and the diffraction patter is in the form of sharp rings
(powder diffractograms). Polymers are sensitive to electron radiation, and this is a real problem; the
electron diffraction pattern may rapidly become fuzzy due to degradation.
3.5 Scattering Methods 163

Fig. 3.63 Transmission electron micrograph of linear PE samples after crystallization at 130.6  C for 24 h and then
rapid cooling to room temperature. The beautiful structure had been revealed by permanganic acid etching. Courtesy of
D. C. Bassett. From Bassett et al. (1981) with permission of the Royal Society, UK

The wavelength of the X-rays used in WAXS is typically 1 Å (10–10 m), whereas the electrons
accelerated in a 100 kV electron microscopy has a wavelength of 4 pm (410–12 m), which is much
small than the lattice parameters (a, b and c) in crystals. This means that the angles for constructive
interference are very small for the electron. According to Bragg’s law, sin θ ¼ nλ/(2dhkl), which
means that θ is very small, sin θ  θ and Bragg’s law can be reformulated to dhkl ¼ nλ/(2θ). The
analysis of an electron diffractogram is greatly facilitated by having the access of reciprocal space
diagrams for different crystal structures (cf. Hammond 1997; this is a lovely text!). Another
consequence of then small diffraction angles is that lattice plans have to be almost parallel to the
electron beam. Electron diffraction pattern in some instances is very useful to determine unit cells of
polymer crystals (Claffey et al. 1974; Dorset et al. 1984), although in many real cases the crystals are
small and the pattern is in fact obtained from oriented textures of several crystals.

3.5.5 Neutron Scattering

Neutron scattering utilizes thermal neutrons which have a wavelength of approximately 0.5
nm. X-rays interact with the electrons which causes their scattering and diffraction in polymers.
Neutrons are different because they are electrical neutral, and this gives them high penetration power,
and they interact with the atom nuclei, and their scattering is based on the characteristics of the atom
nuclei of the different elements and isotopes. Neutrons are sensitive to light elements, and they react
differently for different isotopes of the same element; such variation provides possibility for coherent
scattering. Hydrogen is present in almost all polymers; polytetrafluoroethylene is one exception. A
typical polymer sample to be studied by neutron scattering is a binary solution of polymers based on
polymers containing different isotopes of hydrogen: ‘normal’ hydrogen (1H) and deuterium (2H).
These isotopes have different coherent scattering length, –3.75 fm (10-15 m) (1H) and 6.67 fm (2H).
164 3 Spectroscopy and Scattering of Radiation by Polymers

Hydrogen (1H) scatters neutrons mainly incoherently, whereas deuterium (2H) scatters in a coherent
fashion. This means that a polymer chain ‘labelled’ with deuterium in a solution of the same polymer
with only hydrogen 1H scatters neutrons in the same way as a polymer solution with a low molar mass
solvent scatter light, i.e. by Rayleigh scattering. This means that coherent neutron scattering provides
the same type of information as light scattering from a polymer solution: average molar mass (Mw)
and radius of gyration (s) are revealed. The chain conformation of polymers in the condensed states
such as polymer melts, rubbers, semicrystalline polymers and glassy polymers is disclosed. The
wavelength of the neutrons is much shorter than that of visible light, and therefore neutron scattering
provides information about the global dimensions of the polymer coils at much lower angles than that
used in light scattering. Small-angle neutron scattering (SANS) is the name of this method. It is
possible to make similar studies at higher scattering angles, by wide-angle neutron scattering
(WANS), and it provides information about the local chain conformations (about the geometry of
shorter sequences of the polymer chain).
A few more things need to be made clear. It is important that the deuterated chains are present in a
relatively low concentration and that the molecules are uniformly distributed in the host polymer.
Clustering (that two or more deuterated molecules are together) destroys the possibility to disclose the
conformation. Studies by SANS of polyethylene crystallization and the resulting structure have a
problem; the deuterated chains segregate during slow crystallization, and therefore information about
the chain conformation in this polymer can only be obtained for rapidly cooled systems. Another
problem concerns the generation of the thermal neutrons, which is not available in many places,
because their formation requires reactions such as nuclear fission or nuclear fusion. SANS
instruments are enormously large and are only available at a few places worldwide. Examples of
such facilities are the Laue-Langevin, Grenoble, with an 80 m instrument, and Oak Ridge National
Laboratory with a 30 m instrument. Smaller instruments (less than 5 m) are available elsewhere. All
the instruments are equipped with two-dimensional positron-sensitive arrays. In our country
(Sweden), a very large facility (European Spallation Source, ESS) is under construction.
Neutron scattering (Rayleigh type) has provided data which have shed new light on several areas
in polymer science. The Flory theorem (cf. Gedde and Hedenqvist 2019b) was confirmed when SANS
showed that the molar mass dependence of the radius of gyration (s) in the bulk (molten) state was the
same as in a theta solvent: s / M1/2. The effect of deformation and swelling of rubbers was also
studied by SANS, and the affine deformation model and the phantom network model were critically
tested (Gedde and Hedenqvist 2019f). Crystallization of polyethylene (mixtures of hydrogenated and
deuterated chains) was studied by both SANS and WANS. The radius of gyration showed only a weak
molar mass dependence for the solution crystals. This finding, together with WANS data, led to the
proposal of the super-folding model (cf. Gedde and Hedenqvist 2019c). The radius of gyration of
relatively rapidly cooled melt-crystallized samples (to avoid segregation of the deuterated molecules)
is the same as that in the molten state prior to crystallization. It is not clear if this scheme also holds
for samples crystallized at a lower degree of supercooling (cf. Gedde and Hedenqvist 2019c).
Recommended general texts about neutron scattering are Wignall (1993); Higgins and Benoit
(1994); Furrer et al. (2009); and Squires (2012).
Another class of method is the dynamic neutron scattering, which is based on incoherent scatter-
ing. This group of methods which include neutron time-of-flight spectroscopy and neutron spin echo
spectroscopy is spatial sensitive and has the possibility to assess the dynamics also down to very short
periods of time. The time scales that are probed with these techniques have been extended and cover
times scales from picoseconds to several hundreds of nanoseconds. A recommended text providing
details is by Boyd and Smith (2007).
3.7 Exercises 165

3.6 Summary

The spectroscopy and scattering methods have been and still are major tools in polymer science and
engineering. It is a good investment to learn the basics about these methods. At the same time, it is
almost an overwhelmingly difficult task to present this field in less than 100 pages. The attempt has
been to provide the basics without too much derivation of equations. In parts, this chapter is more of a
how-to-do text without 100 % of complete derivation and full explanation. We have therefore
provided many excellent references, in which derivations, further explanations and applications are
available. When you feel that something missing, please visit these references. The main spectro-
scopic methods described in more detail are vibrational spectroscopy (infrared and Raman spectros-
copy), nuclear magnetic resonance (NMR) spectroscopy and X-ray photoelectron (XPS)
spectroscopy. Electron spin resonance (ESR) spectroscopy, ultraviolet (UV) spectroscopy and visible
light (VIS) spectroscopy are presented in a brief form. The scattering methods use various types of
radiation, which include visible light, UV radiation, X-rays, electrons and neutrons. Light scattering
focusing on the Rayleigh scattering from polymer solutions revealing molar mass and radius of
gyration is dealt with in more detail, and the coherent neutron scattering (also of Rayleigh type) is
also presented but in a brief form. The focus parts of the scattering section are wide-angle X-ray
scattering and small-angle X-ray scattering, presenting details of instrumentation, methods useful for
the analysis of data and also some application examples. Electron diffraction is presented more
briefly.

3.7 Exercises

3.1 Suggest two different methods for the assessment of chain branching in polyethylene.
3.2 Suggest spectroscopy and diffraction methods by which the crystal thickness in a polyethylene
sample can be determined.
3.3 Can the miscibility of two amorphous polymers be judged by IR spectroscopy or by NMR
spectroscopy? Describe how and what are the limitations?
3.4 The following crystallinity values were obtained for poly(propylene-stat-ethylene) by X-ray
diffraction using the Ruland method and by IR spectroscopy using the method of Trotignon:

Ethylene content X-ray IR

(mol.%) crystallinity crystallinity
0 49 67
3.6 36 49
6.6 33 48
9.3 37 47
Data by S. Laihonen, KTH Royal Institute of Technology, Stockholm
Suggest a possible explanation of the above data

3.5 A fibre of a semicrystalline polymer is highly stressed. How can the deformation be recorded by
spectroscopy and scattering techniques?
3.6 Polyolefins are readily oxidized at high temperatures, particularly when no antioxidant is
present. IR spectroscopy is commonly used to determine the concentration of the oxidation
products. What is the main problem in the IR analysis?
3.7 Chain orientation can be assessed by both spectroscopy and scattering methods. Write a short
description of a few of the methods.
166 3 Spectroscopy and Scattering of Radiation by Polymers

3.8 In an undecoupled 13C NMR spectrum, the only couplings occurring are those between carbon
and hydrogen. Explain this.
3.9 If you run a proton-coupled 13C NMR experiment on poly(isobutylene), how many different
carbons will you see? How are they located relative to each other on the chemical-shift axis?
How are the signals split and what are the relative intensities of the split peaks?
3.10 If the resonance frequency is 50 MHz for TMS, and a resonance signal from a proton occurs at
500 Hz downfield from TMS, what is the chemical shift of the proton resonance, and what
chemical group is probably responsible for this resonance?
3.11 The precession (Larmor) frequency is found at 500 MHz for a proton under a certain magnetic
field. What is the static magnetic field? What is the corresponding precession frequency of 13C in
the same magnetic field?

References

Abbe, E. (1873). Archiv für Mikroskopische Anatomie, 9, 413.

Abdul-Karim, R., Hameed, A., & Imran Malik, M. (2017). RSC Advances, 7, 11786.
Alipour, N., Vinnerås, B., Gouanve, F., Espuche, E., & Hedenqvist, M. S. (2019). Polymers, 11, 287.
Ashbrook, S.E., & Hodginson, P. (2018). The Journal of Chemical Physics, 149, 040901.
Baker, A. M. E., & Windle, A. H. (2001). Polymer, 42, 667.
Baltá-Calleja, F. J., & Vonk, C. G. (1989). X-ray scattering of synthetic polymers. Amsterdam: Elsevier Science
Publishers.
Bassett, D. C., Hodge, A. M., & Olley, R. H. (1981). Proceedings of the Royal Society of London, A377, 39.
Bec, K. B., & Huck, C. W. (2019). Frontiers in Chemistry, 7, 48.
Berne, B. J., & Pecora, R. (2000). Dynamic light scattering with applications to chemistry, biology and physics.
Mineola: Dover Publications.
Bertrand, P. (2020). Electron Paramagnetic Resonance Spectroscopy. Cham: Springer Nature Switzerland AG.
Block, F., Hansen, W. W., & Packard, M. (1946a). Physical Review, 69, 127.
Block, F., Hansen, W. W., & Packard, M. (1946b). Physical Review, 70, 474.
Bloembergen, N., Purcell, E. M., & Pound, R. V. (1948). Physical Review, 73, 679.
B€orjesson, M., H€ardelin, L., Nylander, F., Karlsson, K., Larsson, A., & Westman, G. (2018). Bioresources, 13, 6201.
Bower, D. I., & Maddams, W. F. (1989). The vibrational spectroscopy of polymers. Cambridge: Cambridge University
Press.
Boyd, R. H., & Phillips, P. J. (1993). The science of polymer molecules. Cambridge: Cambridge University Press.
Boyd, R. H., & Smith, G. D. (2007). Dynamic neutron scattering. In Polymer dynamics and relaxation. Cambridge:
Cambridge University Press.
Bradbury, E. R., Bradbury, A., & Eades, R. G. (1958). Nature, 182, 1659.
Bragg, W. H., & Bragg, W. L. (1913). Proceedings of the Royal Society of London A, 88, 428.
Bravais, A. (1850). Journal de l’Ecole Polytechnique, 19, 1.
Briggs, D. (1998). Surface analysis of polymers by XPS and static SIMS. Cambridge: Cambridge University Press.
Buerger, M. J. (1945). Journal of Applied Physics, 16, 501.
Burchard, W. (1982). Advances in Polymer Science, 44, 1.
Butler, H. J., Ashton, L., Bird, B., Cinque, G., Curtis, K., Dorney, J., Esmonde-White, K., Fullwood, N. J., Gardner, B.,
Martin-Hirsch, P. L., Walsh, M. J., McAinsh, M. R., Stone, N., & Martin, F. L. (2016). Nature Protocols, 11, 664.
Camacho, W., Hedenqvist, M. S., & Karlsson, S. (2002). Polymer International, 51, 1366.
Chantry, G. W., Gebbie, H. A., & Hilsum, C. (1964). Nature, 203, 1052.
Claffey, W., Gardner, K., Blackwell, J., Lando, J., & Geil, P. H. (1974). Philosophical Magazine, 30, 1223.
Clark, D. T. (1976). In. In K. J. Jvin (Ed.), Structural studies of macromolecules by spectroscopic methods. London:
John Wiley & Sons.
Claudino, M., Jonsson, M., & Johansson, M. (2014). RSC Advances, 4, 10317.
Coblentz, W. W. (1905–1908). Investigation of Infrared Spectra (Publications Nos 35, 65 and 97). Washington, DC:
Carnegie Institution.
Conde Braña, M. T., & Gedde, U. W. (1992). Polymer, 33, 3123.
Cooley, J. W., & Turkey, J. W. (1965). Mathematics of Computation, 19, 297.
Cser, F. (2001). Journal of Applied Polymer Science, 80, 2300.
References 167

Damadian, R. (1971). Science, 171, 1151.

Debye, P. (1947). The Journal of Physical Chemistry, 51, 18.
Dorset, D. L., Moss, B., Wittmann, J. S., & Lotz, B. (1984). Proceedings of the National Academy of Science of USA,
81, 1913.
Dybal, J., Schmidt, P., Kurková, D., Kriz, J., Rodriquez-Cabello, J.C., & Alonso, M. (2002). Spectroscopy, 16, 251.
Einstein, A. (1905). Annalen der Physik, 17, 132.
El Khoury, Y., & Hellwig, P. (2017). Chemical Communications, 53, 8389.
Ernst, R. R. (1966). Advances in Magnetic. Resonance, 2, 1.
Ernst, R. R., & Anderson, W. A. (1966). Review of Scientific Instruments, 37, 93.
Ewald, P. P. (1912). Dispersion und Doppelbrechung von Electronengittern. Ph. D. Thesis, Ludwig Maximilian
University of Munich, Germany.
Fahrenfort, J. (1961). Spectrochimica Acta, 17, 679.
Fellgett, P. B. (1951). Thesis, University of Cambridge, UK.
Fergin, L. A., & Svergun, D. T. (1987). Structure analysis by small-angle X-ray and neutron scattering. New York:
Plenum Press.
Fitzspatrick, J. R., & Ellis, B. (1973). X-ray diffraction studies of the structure of amorphous polymers. In R. N. Haward
(Ed.), The physics of glassy polymers. London: Applied Science Publishers.
Ford, M. (2002). The Internet Journal of Vibrational Spectroscopy, 6, 4.
Fresnel, A. (1818). Memoire sur la Diffraction de la Lumière. In Oeuvres Complètes (Vol. 1, pp. 247–363). Paris:
Imprimerie Imperiale, 1866–70.
Furrer, A., Mesot, J., & Str€assle, T. (2009). Neutron scattering in condensed matter physics. Singapore: World
Scientific.
Gedde, U. W., & Hedenqvist, M. S. (2019a). Chain orientation (Chapter 9). In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Conformations in polymers (Chapter 2). In Fundamental polymer science.
Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019c). Morphology in semicrystalline polymers (Chapter 7). In Fundamental
polymer science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019d). Polymer solutions (Chapter 4). In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019e). Introduction to polymer science (chapter 1). In Fundamental polymer
science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019f). Rubber elasticity (Chapter 3). In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Glatter, O., & Kratky, O. (1982). Small-angle X-ray scattering. London/New York: Academic Press.
Guinier, A. (1952). X-ray crystallographic technology. London: Hilger and Watts.
Guinier, A., & Fournet, G. (1955). Small Angle Scattering of X-rays. New York: John Wiley & Sons.
Guo, X., Laryea, E., Wilhelm, M., Luy, B., Nirschl, H., & Guthausen, G. (2017). Macromolecular Chemistry and
Physics, 218, 1600440.
Gupta, S. K. (1998). Journal of Applied Crystallography, 31, 474.
Gupta, M. R., & Yeh, G. S. Y. (1978). Journal of Macromolecular Science. Part B. Physics, B15, 119.
H€agg, G. (1947). Review of Scientific Instruments, 18, 371.
Hahn, E. L. (1950). Physical Review, 80, 580.
Hammond, C. (1997). The basics of crystallography and diffraction. Oxford: Oxford University Press.
H€ardelin, L., Bernin, D., B€orjesson, M., Str€om, A., & Larsson, A. (2020). Biomacromolecules, 21, 1832.
Hartmann, S. R., & Hahn, E. L. (1962). Physical Review, 128, 2042.
Hatada, K., & Kitayama, T. (2004). NMR spectroscopy of polymers. Berlin: Springer.
He, B. B. (2018). Two-dimensional X-ray diffraction (2nd ed.). Hoboken: John Wiley & Sons.
Hermans, J. J., Hermans, P. H., Vermaas, D., & Weidinger, A. (1946). Recueil des Travaux Chimques des Pays-Bas, 65,
427.
Herschel, W. (1800a). Philosophical Transactions of the Royal Society, 90, 255.
Herschel, W. (1800b). Philosophical Transactions of the Royal Society, 90, 284.
Herschel, W. (1800c). Philosophical Transactions of the Royal Society, 90, 293.
Herschel, W. (1800d). Philosophical Transactions of the Royal Society, 90, 437.
Higgins, J. S., & Benoit, H. C. (1994). Polymers and neutron scattering. New York: Oxford University Press.
Hillborg, H., Ankner, J. F., Gedde, U. W., Smith, G. D., Yasuda, H. K., & Wikstr€ om, K. (2000). Polymer, 41, 6851.
Holliday, L., & White, J. W. (1971). Pure and Applied Chemistry, 26, 245.
Hosemann, R. (1950a). Zeitschrift für Physik, 128, 1.
Hosemann, R. (1950b). Zeitschrift für Physik, 128, 465.
168 3 Spectroscopy and Scattering of Radiation by Polymers

Hosemann, R. (1963). Journal of Applied Physics, 34, 25.

Hosemann, R. (1982). Physica Scripta, T1, 142.
Hosemann, R., Hentschel, M. P., Balta-Calleja, F. J., Lopez Cabarcos, E., & Hindeleh, A. M. (2001). Journal of physics
C: Solid state. Physics, 18, 249.
Huygens, C. (1678). Traité de la Lumière (drafted 1678; printed 1690). Leyden: van der Aa.
Ibbett, R. N. (1993). NMR spectroscopy of polymers. Berlin: Springer.
Innes, P. D. (1907). Proceedings of the Royal Society of London, Series A, 79, 442.
Izunobi, J. U., & Higginbotham, C. L. (2011). Journal of Chemical Education, 88, 1098.
Jacquinot, J. (1954). 17e Congres du GAMS, Paris, p. 25.
Jeener, J. (1971). Ampere international summer school. Jugoslavia: Basko Polje.
Johann, H. H. (1931). Zeitschrift für Physik, 69, 185.
Kainulainen, T. P., Hukka, T. I., Özeren, H. D., Sirvi€o, J. A., Hedenqvist, M. S., & Heiskanen, J. P. (2020).
Biomacromolecules, 21, 743.
Karlsson, G. E., Johansson, T. S., Gedde, U. W., & Hedenqvist, M. S. (2002). Journal of Macromolecular Science, Part
B. Physics, B41, 185.
Kastler, A. (1950). Journal de Physique, 11, 255.
Kausch, H. H. (1978). Polymer fracture. Berlin/Heidelberg/New York: Springer.
Kida, T., Hiejima, Y., & Nitta, K. (2016). International Journal of Experimental Spectroscopic Techniques, 1, 001.
Kiessig, H. (1942). Kolloid Zeitschrift & Zeitschrift fur Polymere, 98, 213.
Kirchhoff, G. (1883). Annalen der Physik, 254, 663.
Kitamaru, R., Horii, F., & Hyon, S.-H. (1977). Journal of Polymer Science: Polymer Physics Edition, 15, 821.
Kitamaru, R., Horii, F., & Murayama, K. (1986). Macromolecules, 19, 636.
Kl€opffer, W. (1989). Introduction to polymer spectroscopy. Berlin: Springer Verlag.
Koenig, J. L. (1992). Spectroscopy of Polymers. Washington, DC: American Chemical Society.
Kratky, O. (1954). Zeitschrift für Elektrochemie, 58, 49.
Kratky, O. (1956). Zeitschrift für Elektrochemie, 60, 245.
Kratky, O. (1958). Zeitschrift für Elektrochemie, 62, 66.
Kratky, O., & Laggner, P. (1987). X-ray small-angle scattering. In R. Meyers (Ed.), Encyclopedia of physical science
and technology (Vol. 14, pp. 693–742). New York: Academic Press.
Ladd, M., & Palmer, R. (2013). Structure determination by X-ray crystallography. New York: Springer Science and
Business Media.
Lauterbur, P. C. (1973). Nature, 242, 190.
Lowell, R., Mitchell, G. R., & Windle, A. H. (1979). Faraday Discussions of the Chemical Society, 68, 46.
Martı́nez-Richa, A., & Silvestri, R., L. (2017). Developments in solid-state NMR spectroscopy of polymer systems. In
E. Sharmin and F. Zafar (Eds.), Spectroscopic analyses. London: IntechOpen.
Mattozzi, A., Larsson, P. T., Hedenqvist, M. S., & Gedde, U. W. (2010). European Polymer Journal, 46, 381.
Michelson, A., & Morley, E. (1887). American Journal of Science, 34, 333.
Miller, W. H. (1839). A treatise on crystallography. Cambridge/London: J. J. Deighton and John W. Parker.
Mitchell, G. R. (1984). Polymer, 25, 1562.
Mitsutake, H., Poppia, R. J., & Breitkreitz, M. C. (2019). Journal of the Brazilian Chemical Society, 30, 2243.
Mizushima, S., & Simanouti, T. (1949). Journal of the American Chemical Society, 71, 1320.
Moser, E., Laistler, E., Schmitt, F., & Kontaxis, G. (2017). Frontiers in Physics, 5, 33.
Murrell, J. N. (1963). The theory of electronic spectra of organic molecules. London: Methuen.
Mutter, R., Stille, W., & Strobl, G. (1993). Journal of Polymer Science: Polymer Physics Edition, 31, 99.
Nameer, S., & Johansson, M. (2017). Journal of Coatings Technology and Research, 14, 757.
Nordling, C., Siegbahn, K., Sokolowski, E., & Wapstra, A. H. (1956). Nuclear Physics, 1, 326.
Perkampus, H.-H. (1992). UV-VIS spectroscopy and its applications. Berlin, Heidelberg: Springer.
Pezzutti, J. L., & Porter, R. S. (1985). Journal of Applied Polymer Science, 30, 4251.
Pines, A., Gibby, M. G., & Waugh, J. D. (1973). Journal of Chemical Physics, 59, 569.
Porto, S. P. S., & Wood, D. C. (1962). Journal of the Optical Society of America, 52, 251.
Pourmand, P., Linde, E., Hedenqvist, M. S., Furó, I., Dvinskikh, S. V., & Gedde, U. W. (2016). Polymer Testing, 53, 77.
Purcell, E. M., Torrey, H. C., & Pound, R. V. (1946). Physical Review, 69, 37.
Rabi, I. I. (1937). Physics Reviews, 51, 652.
Rabi, I. I., Zacharias, J. R., Millman, S., & Kusch, P. (1938). Physical Review, 53, 318.
Raman, C. V., & Krishna, K. S. (1928). Nature, 121, 501.
Ramsey, N. (1950a). Physical Review, 77, 567.
Ramsey, N. (1950b). Physical Review, 78, 699.
Ramsey, N. (1951). Physical Review, 83, 540.
Ramsey, N. (1952). Physical Review, 86, 243.
References 169

Ramsey, N. (1953a). Physical Review, 89, 527.

Ramsey, N. (1953b). Physical Review, 90, 232.
Ramsey, N. (1953c). Physical Review, 91, 303.
Rånby, B., & Rabek, J. F. (1977). ESR spectroscopy in polymer research. Berlin/Heidelberg/New York: Springer.
R€ontgen, W. C. (1898). Annalen der Physik, 300, 12.
Ruland, W. (1961). Acta Crystallographica, 14, 1180.
Saalw€achter, K. (2019). Applications of NMR in polymer characterization – an introduction. In R. Zhang, T. Miyoshi,
& P. Sun (Eds.), NMR methods of characterization of synthetic and natural polymers. London: Royal Society of
Chemistry.
Samuelsson, J., Jonsson, M., Brinck, T., & Johansson, M. (2004). Journal of Polymer Science: Part A: Polymer
Chemistry, 42, 6346.
Scherrer, P. (1918). Nachrichten von der Gesellschaft der Wissenschaften zu G€ ottingen, Matematisch-Physikalische
Klasse, 2, 98.
Siegbahn, K., & Edvardson, K. (1956). Nuclear Physics, 1, 137.
Siesler, H. W. (2011). International Journal of Polymer Analysis and Characterization, 16, 519.
Smekal, A. (1923). Naturwissenschaften, 11, 873.
Sokolowski, E., Edvardson, K., & Siegbahn, K. (1956). Nuclear Physics, 1, 160.
Spiess, H. W. (2017a). Macromolecules, 50, 1761.
Spiess, H. W. (2017b). Macromolecules, 50, 2017.
Squires, G. L. (2012). Thermal Neutron Scattering (3rd ed.). Cambridge: Cambridge University Press.
Statton, W. O. (1962). Journal of Polymer Science, 58, 205.
Stein, R. S. (1958). Journal of Polymer Science, 31, 327–335.
Stribeck, N. (2007). X-ray scattering of soft matter. Berlin, Heidelberg: Springer.
Strobl, G. (1997). The physics of polymers: Concepts for understanding their structures and behavior. Berlin/
Heidelberg/New York: Springer.
Strutt, J. W. (1871). The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 41. 107, 274
and 447.
Strutt, J. W. (1899). The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 47, 375.
Swan, P. R. (1962). Journal of Polymer Science, 56, 409.
Tolam, M. (1999). X-ray scattering from soft-matter thin films. Berlin, Heidelberg: Springer.
Tomba, J. P., Arzondo, L. M., & Pastor, J. M. (2007). Applied Spectroscopy, 61, 177.
T€orm€al€a, P., Weber, G., & Lindberg, J. J. (1980). In R. F. Boyer & S. E. Keinath (Eds.), Molecular motion in polymers
by ESR. New York: Harwood Academic Publishers.
Von Laue, M. (1912). Eine quantitative Prüfung der Theorie für die Interferenzerscheinungen bei R€ ontgenstrahlen.
Sitsungberichte der Bayerische Akademi der Wissenschaften, p. 363.
Von Laue, M., Friedrich, W., & Knipping, P. (1912). Interferenzer-scheinungen bei R€ ontgenstrahlen. Sitsungberichte
der Bayerische Akademi der Wissenschaften, p. 303.
Vonk, C. G. (1973). Journal of Applied Crystallography, 6, 148.
Vonk, C. G., & Pijpers, A. P. (1981). Journal of Applied Crystallography, 14.
Watts, J., & Wolsenholme. (2020). An introduction to surface analysis by XPS and AES (2nd ed.). Hoboken: John Wiley
& Sons.
Wei, X.-F., Kallio, K. J., Bruder, S., Bellander, M., Kausch, H.-H., Gedde, U. W., & Hedenqvist, M. S. (2018). Polymer
Degradation and Stability, 154, 73.
White, J. L., & Spruiell, J. E. (1981). Polymer Engineering and Science, 21, 859.
Wignall, G. D. (1993). Scattering techniques. In J. E. Mark (Ed.), Physical properties of polymers (2nd ed.).
Washington, DC: American Chemical Society.
Wilchinsky, Z. W. (1963). Journal of Applied Polymer Science, 7, 923.
Wool, R. P. (1975). Journal of Polymer Science, Polymer Physics Edition, 13, 1795.
Wrobel, T. P., & Bhargava, R. (2018). Analytical Chemistry, 90, 1444.
Wu, Z., & Xing, X. (2018). Small angle X-ray scattering and its applications (Chap. 7). In X. Zhang (Ed.), Synchrotron
radiation applications. Singapore: World Scientific.
Zavoisky, E. (1944). Paramagnetic Absorption in Perpendicular and Parallel Fields for Salts, Solutions and Metals,
Ph. D. Thesis, University of Moscow.
Zavoisky, E. (1945). Fizicheskii Zhurnal, 9, 211.
Zerbi, G., Ciampelli, F., & Zamboni, V. (1964). Journal of Polymer Science. Polymer Symposia, 7, 141.
Zhu, X., Xu, T., Lin, Q., & Duan, Y. (2014). Applied Spectroscopy Reviews, 49, 64.
Zimm, B. H. (1948). The Journal of Chemical Physics, 16, 1099.
Chapter 4
Chromatographic Analysis of Polymers

4.1 Introduction

Chromatography is a powerful technique for the separation of mixtures for both preparative and
analytical purposes, to identify and quantify compounds in the mixture. Chromatography is based on
two mutually immiscible phases, a stationary phase and a mobile phase. The mobile phase carries the
sample through the system, and during its passage, different compounds are retained for different
periods of time depending on their interactions with the stationary phase. The partition of components
between the mobile and stationary phases is based on the physical and chemical properties of the
components. By the correct choice of chromatographic system, different types of mixture can be
separated into their individual components. The two main types of chromatography are liquid
chromatography (LC) and gas chromatography (GC), depending on whether the mobile phase is a
liquid or a gas. Chromatography is further divided into several subclasses based on the type of
stationary phase (Fig. 4.1). In polymer science, a special type of liquid chromatography, size
exclusion chromatography (SEC), is an important tool for assessing molar mass distribution. This
technique is also called gel permeation chromatography (GPC). In this chapter, the principles of SEC,
LC and GC, are briefly described, and their applications in the polymer field are reviewed. SEC is
basically a liquid chromatographic method, but due to its special nature and importance for polymers,
it is here presented under its own subtitle. Recommended introductory texts about the various
chromatographic techniques are given by Poole (2003) and Vitha (2017).
The chromatographic methods provide vital information about the composition of polymeric
materials. Commercial plastics and rubbers usually contain a number of additives that are included
to provide particular physical and chemical properties. These additives include plasticizers, inorganic
fillers, antioxidants, heat and light stabilizers, processing aids, cross-linking agents, pigments and
flame retardants. Monomeric or oligomeric residues sometimes remain after the synthesis step, and
new low molar mass compounds are formed due to the degradation of polymer and additives during
the synthesis, processing, service life and recycling. Knowledge of these compounds is crucial for
ensuring the performance and safety of the materials, as will be discussed further in this chapter
(Hakkarainen and Karlsson 2000). SEC also provides detailed information about the molar mass
distribution and the molecular architecture of the polymer component in polymeric materials, which
have a major impact on some rheological and mechanical properties (Mori and Barth 1999). The
choice of chromatographic method depends on the type of compound to be analysed and on the
purpose of the analysis (Albertsson and Hakkarainen 2008; Cazes 2010). One of the most important
parameters determining the choice of method is the molar mass (Fig. 4.2); others include polarity and

# Springer Nature Switzerland AG 2021 171

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_4
172 4 Chromatographic Analysis of Polymers

Fig. 4.1 Classification of gas and liquid chromatographic techniques

Fig. 4.2 Typical molar

mass ranges for GC, LC
and SEC

solubility. Due to the complex nature and the molecular heterogeneity of polymers, multidimensional
techniques combining, for example, SEC and high-performance liquid chromatography (HPLC) have
also been developed (Pasch and Trathnigg 2013).

4.2 General Concepts in Chromatography

A chromatographic system can utilize either gas or liquid as the mobile phase and either a solid or
liquid as the stationary phase. Liquid chromatography usually utilizes a solid stationary phase
(liquid–solid chromatography), the simplest liquid–solid chromatographic methods being paper
chromatography and thin-layer chromatography (TLC), whereas HPLC and SEC require more
sophisticated instrumentation. Gas chromatography can also utilize a solid stationary phase
(gas–solid chromatography), where the typical separation or retention mechanism is adsorption, a
process where an analyte adheres to the solid surface of the stationary phase (SEC is excepted). The
adsorption is controlled by secondary intermolecular bonds established between the analytes in the
mobile phase and the stationary phase. If these interactions are strong, the analyte spends a long time
trapped by the stationary phase, and the time for the analyte to pass through the column is long,
because the analyte moves forward only during the time it spends in the mobile phase. In
liquid–liquid and gas–liquid chromatography the stationary phase is liquid, generally applied on a
solid support. This means that it either covers solid particles for LC or is located inside the walls of
open tubular capillary columns for GC. When the stationary phase is a liquid, the interaction between
the analyte and the stationary phase is due to dissolution instead of adsorption. The analyte molecules
then dissolve in the liquid phase and move forward through partition between the two phases. The
analyte molecules move forward only during the time they spend in the mobile phase and an analyte
which is highly soluble in the liquid phase (stationary phase) spends a long time in the column.
Some of the important terms that describe chromatographic separation are related to those used for
liquid–liquid extraction. The distribution coefficient (KC) is the ratio of the concentrations of the
compound of interest in the stationary phase (CS) and in the mobile phase (CM):
4.2 General Concepts in Chromatography 173

CS n =V
KC ¼ ¼ S S ð4:1Þ
CM nM =V M

where VS and VM are, respectively, the volumes of the stationary and mobile phases. In the case of a
solid stationary phase, VS is related to the surface area of the particles. A high KC value corresponds to
a long time for the compound to spend in the stationary phase before it is eluted, i.e. a long retention
time (TR). This is sometimes expressed on a volume basis as the retention volume (VR), where VR is
the volume of the mobile phase required to carry the solute through the column to elution. If the
stationary phase is liquid, the partition coefficient (KX) is given by the ratio of the number of moles of
analyte in each of the two phases:

nS
KX ¼ ð4:2Þ
nM

By combining Eqs. (4.1) and (4.2), the following relationship between KC and KX is obtained:


VM
KC ¼ KX  ð4:3Þ
VS

A mixture of analytes can be separated in a chromatographic column provided that the KC or KX

values for the different analytes are sufficiently different. The ratio of the KC or KX values for two
analytes in a mixture is referred to as the selectivity factor (α):

K2
α¼ ð4:4Þ
K1

where K1 and K2 are the KC or the KX values of the first and second eluting analytes, respectively. A
large value of α ensures an adequate separation of the analytes (Fig. 4.3). For short analysis time,
however, the optimum α-value is not the maximum value, but rather a value that is sufficiently high to
ensure separation of the chromatographic peaks while still leading to their elution as close to each
other as possible.
Another important parameter in chromatographic separation is the retention factor (k) (or capacity
factor), which is in turn related to the retention time of the analyte, which is directly related to its
interactions with the stationary phase. k is equal to the partition coefficient, but it can also be
determined directly from the chromatogram (Fig. 4.3c) by the equation:

tR  t0
k¼ ð4:5Þ
t0

where tR is the retention time of the compound of interest and t0 is the retention time of an unretained
compound with no interaction with the stationary phase (usually assumed to be the solvent used to
dissolve the compounds to be analysed). Both t0 and tR are also dependent in the column used.
The resolution (RS) of two compounds shows by how much the two compounds or analytes of
interest are separated from each other. It is calculated as:

tRB  tRA
RS ¼ 1 ð4:6Þ
2  ðW A þ W B Þ
174 4 Chromatographic Analysis of Polymers

Fig 4.3 Separation of two

compounds with (a) large
and (b) small α value. (c)
Sketch of a chromatogram
showing the parameters
used to calculate the
retention factor

Fig. 4.4 Separation of two

substances (A and B) with
the resolution values: (a)
0.75, (b) 1.0 and (c) 1.5.
The bottom chromatogram
also illustrates the retention
times (tri) and peak width
(Wi) parameters

where tRA and tRB are the retention times of the two compounds and WA and WB are the baseline
widths of the two peaks (Fig. 4.4). This graph demonstrates the resolution concept by showing
chromatograms for a mixture of two substances with the Rs values 0.75, 1.0 and 1.5. At a resolution
value of 1.0, the peaks are separated just enough to almost reach the baseline between the peaks. The
separation of two peaks increases with increasing RS value, but a higher resolution also means a longer
analysis time.
4.3 Size Exclusion Chromatography 175

The separation power of the chromatographic column can be expressed as the number of theoreti-
cal plates (N ), where the chromatographic column is divided into a large number of imaginary
theoretical layers or theoretical plates. Assuming that the compound of interest moves through the
column by a series of steps (from one theoretical plate to next) and that the time the compound spends
within each step is controlled by the strength of its interaction with the stationary phase, a column
with a larger number of steps (i.e. theoretical plates) is more likely to separate two components with
good resolution than a column with only a small number of steps. The number of plates for a certain
compound can be calculated as follows:
 2
tR
N ¼ 16  ð4:7Þ
W

where tR is the retention time of the compound and W is the baseline width of the peak. Although the
value N is calculated for a specific compound, the N value obtained for another compound of the
mixture does not usually differ greatly, because the peaks for rapidly eluting compounds are usually
sharp, whereas late elution peaks are broad, leading to a greater peak width. To calculate the number
of plates for a certain column, the column length (L ) is divided by the number of theoretical plates.
This gives the height equivalent of the theoretical plate (H ). The smaller the value of H, the greater is
the separation power of the column.

4.3 Size Exclusion Chromatography

Size exclusion chromatography (SEC) is a technique for separating polymer molecules based on their
hydrodynamic volume, and it is capable of providing a complete picture of the molar mass distribu-
tion. The separation in the chromatographic column takes place without the occurrence of a specific
interaction such as adsorption between the polymer molecules and the column packing material. The
hydrodynamic volume of a given polymer molecule is related to its molar mass, but it is also
influenced by factors such as the solvent power, temperature and molecular architecture (Gedde
and Hedenqvist 2019a). In 1959, Porath and Flodin (1959) utilized cross-linked dextran to separate
proteins based on their molar mass, and a few years later this approach was applied to synthetic
polymers (Moore 1964). Application to synthetic polymers advanced in the 1970s, and SEC has since
then made a significant contribution to polymer characterization and has become the method of
choice for the determination of molar mass distributions. The principles of SEC analysis and its
different applications have been thoroughly described in several dedicated books and review articles
(Mori and Barth 1999; Striegel et al. 2009).
SEC is an indispensable tool for the analysis of synthetic polymers and biopolymers during their
whole life cycle from their synthesis, the influence of processing, the effect of long-term service and
subsequent recycling (Berek 2010). SEC can be used for comparative purposes or, for example, for
assessing the polymerization kinetics and polymer structure such as long-chain branching. The
analysis can be a simple quality control during polymer production, the establishment of reaction
kinetics during the synthesis of new polymers or the evaluation of new catalysts or reaction
conditions. SEC is also a highly useful tool for evaluating the state of service-aged or laboratory-
aged polymeric materials. In addition to the actual molar mass distribution, advanced SEC provides
information about molecular architecture, and it is valuable for the analysis of biopolymers such as
polysaccharides, proteins and lignin, providing information on sample purity, solubility and aggre-
gation (Churms 1996).
176 4 Chromatographic Analysis of Polymers

4.3.1 Molar Mass and Molar Mass Distribution

The molar mass of a polymer has a significant impact on both rheological properties and fracture
toughness. The basic definition is that ‘a polymer is a substance composed of molecules characterized
by the multiple repetition of one or more species of atoms or groups of atoms (i.e. constitutional
repeating units) linked to each other in amounts sufficient to provide set of properties that do not vary
markedly with the addition of one or a few of the constitutional repeating units’ (Jones et al. 2009).
This definition excludes oligomers and materials that do not have chain entanglements. The melt
viscosity of materials which have a molar mass greater than the entanglement molar mass, i.e. greater
than 4000–40,000 g mol–1, increases by a factor of 2000 for each increase by a factor of ten of the
molar mass (cf. review by Graessley 1983). This pronounced increase in melt viscosity with
increasing molar mass makes it very difficult to melt process high molar mass polymers. Another
important rheological property, the melt strength (i.e. the strength of the melt subjected to extensional
flow), increases with increasing molar mass (e.g. La Mantia and Acierno 1985). Recommended texts
about the molar mass dependence of the rheological properties are Graessley (1983, 2004),
Rubinstein and Colby (2003) and Gedde and Hedenqvist (2019b). The fracture toughness of solid
polymers increases with increasing molar mass, particularly in semi-crystalline polymers, because the
fracture toughness requires molecular connections between adjacent crystal lamellae which are
created by tie chains and trapped entanglements. Recommended texts on this topic are Kausch
(1978), Gedde and Hedenqvist (2019c) and Moyassari et al. (2019a, b). The impact of molar mass
on the glass transition temperature (Tg) for a polymer is small. In atactic polystyrene, Tg increases
from 96 to only 100  C when the Mn is increased from 30,000 to 106 g mol–1 (Lin 1990). The structure
and properties of semi-crystalline polymers show a molar mass dependence due to a special reason:
the crystallization of polymers involves diffusion of the crystallizable units, and this diffusion is
dependent on molar mass. A high molar mass polymer thus crystallizes much more slowly than a low
molar mass polymer. If the sample is given only a short period of time for crystallization, as in the
case of a high cooling rate, the crystallinity is expected to show a pronounced decrease with
increasing molar mass (cf. Gedde and Hedenqvist 2019c).
Small molecule compounds have an exact molar mass. Some natural polymers such as many
proteins have a precise structure and molar mass. Both these groups of compounds are referred to as
monodisperse (‘single molar mass’). Commercially available polymers and most natural polymers
(e.g. cellulose and natural rubber) show a distribution in molar mass and are referred to as polydis-
perse (‘multi molar mass’). Depending on the polymerization method, the molar mass distribution
varies from narrow (living polymerization) to very broad (cf. Fig. 4.5); one example of the latter is
polyethylene produced by radical polymerization at high pressure (low-density polyethylene
(LDPE)). The fact that the molecules of polydisperse polymers have different molar masses is the
reason why the molar mass of a polymer is generally given as an average or in fact as several different
averages, the number average (Mn) and the weight average (Mw) being those most frequently used.
The total mass of the polymer sample, m, is the sum of the masses of all the polymer chains in the
sample, i.e.:
X X
m¼ mi ¼ ni M i ð4:8Þ
i i

where ni is the number of moles of molecules with a certain molar mass Mi.
The number average molar mass (Mn) is defined as the total mass of the sample divided by the total
number of molecules:
4.3 Size Exclusion Chromatography 177

Fig. 4.5 Different polymerization processes yield polymers with different degrees of molar mass distribution ranging
from narrow to broad. Drawn after Mori and Barth (1999)

P P
mi ni M i
i i
Mn ¼ P ¼ P ð4:9Þ
ni ni
i i

and it is sensitive to the presence of even small mass fractions of low molar mass species.
The weight average molar mass (Mw) is defined as:
P P
mi M i ni M2i
i i
Mw ¼ P ¼P ð4:10Þ
mi ni M i
i i

and is insensitive to small mass fractions of low molar mass species. It represents the ‘centre of
gravity’ of the molar mass distribution.
The Z-average molar mass (Mz) is more sensitive to the presence of high molar mass species than
Mn and Mw. Hence, a variation in Mz can indicate chain transfer yielding long-chain branching or the
beginning of cross-linking. It is defined as:
P
ni M3i
i
Mz ¼ P ð4:11Þ
ni M2i
i

The molar mass dispersity (Ð) is a quantity used to describe the homogeneity of a polymer sample
with regard to molar mass, and it is defined as:

Mw
D¼ ð4:12Þ
Mn
178 4 Chromatographic Analysis of Polymers

Fig. 4.6 Molar mass distribution and the relationship between different molar mass averages. Drawn after Mori and
Barth (1999)

The molar mass dispersity is closely related to the standard deviation (σ) of the molar mass
distribution: σ/Mn ¼ (Ð  1)1/2. Note that for Ð ¼ 1, σ ¼ 0 and such a polymer is monodisperse with
Mn ¼ Mw ¼ Mz. A polydisperse polymer has a Ð-value greater than 1, and in this case, Mn < Mw < Mz.
Recommended reading about molar mass distribution functions and the more mathematical aspects of
these parameters are the texts by Boyd and Phillips (1993) and Gedde and Hedenqvist (2019d). For a
polydisperse material, a distribution curve can be obtained by plotting the molar mass versus the
fraction of molecules with that molar mass in the sample (Fig. 4.6). A sample’s molar mass dispersity
(Ð) provides a simple, digital representation of the breadth of the molar mass distribution. A broad
molar mass distribution is indicated by a high Ð value. Note that the x-axis has an unusual orientation
with higher values to the left.
The molar mass can be determined by several methods. Mn can be obtained by measurement of the
colligative properties by ebulliometry, cryoscopy or osmometry or by end-group analysis using
titration or spectroscopy (IR or NMR). These methods are absolute and require no calibration using
molar mass standards. The drawback is that the molar mass range is limited in some of these methods:
ebulliometry and cryoscopy cannot deal with polymers with a Mn greater than 104 g mol–1, and
end-group analysis has also a similar limitation. Mw can be determined by light scattering and Mz by
ultracentrifugation. A recommended text about these methods is that of Boyd and Phillips (1993).
SEC is a powerful technique for simultaneous and relatively easy determination of all these averages.
In addition, it is a fast and convenient method in practice.

4.3.2 Principles of SEC Systems

The vital components of a SEC system include a solvent reservoir, a solvent pump, an injection valve,
a column (or series of columns) packed with porous beads or gel, a detector and a computer for data
collection and processing (Fig. 4.7). A high-quality pump is essential for the accurate measurement of
retention time and to force the solvent through the tightly packed column. A dilute filtered polymer
solution is introduced by syringe into the injection valve and carried towards the column by the
mobile phase. The SEC columns are often placed inside an oven to keep the temperature constant or
to provide an elevated temperature in cases where the polymers to be analysed are too viscous or are
insoluble at room temperature. After the molecules are eluted from the SEC column, they pass
through a detector which gives a signal related to the concentration of analyte.
4.3 Size Exclusion Chromatography 179

Fig. 4.7 Schematic representation of a SEC system

Fig. 4.8 Illustration of the porous spherical particles in an SEC column and the passage of polymer molecules through
the column

The number of polymers that can be analysed is limited to those soluble in the particular solvent
used for the available SEC equipment. Typical mobile phases for SEC include chloroform,
dichloromethane, tetrahydrofuran, dimethylformamide, dimethyl sulphoxide and water or aqueous
alkali. High-temperature SEC (HTSEC) for the analysis of polyolefins is typically run with trichlor-
obenzene as the mobile phase at a temperature above 100  C. An additive such as 0.05 M lithium
bromide is needed for polar mobile phases such as dimethylformamide. Otherwise, dipole
interactions between a polar solvent and polar polymers may cause shoulders on the high molar
mass end of the distribution.

4.3.3 Separation in the SEC Column

Once the polymer solution is injected, it will flow through the column filled with finely divided
porous spherical particles (Fig. 4.8). These particles can be either solid particles such as microporous
glass or, more commonly, a solvent-swollen gel of cross-linked polymer particles such as cross-
linked polystyrene. Water-soluble polymers can be analysed by aqueous SEC with a hydrophilic
polymer gel such as poly(glyceryl methacrylate) or poly(vinyl alcohol) with pores or openings where
180 4 Chromatographic Analysis of Polymers

Fig. 4.9 Separation of molecules in an SEC column

the degree of cross-linking is used to control the pore size. Large molecules are unable to enter most
of the pores, and they will pass rapidly through the column, leading to the rule of thumb for SEC:
large molecules come out first (Fig. 4.9). Smaller molecules can access a larger pore volume, and they
take a longer time to pass through the column.
Separation thus occurs based on the hydrodynamic volume (Vhd) of the molecules in the particular
solvent used as a mobile phase: Vhd / r3 (r ¼ average end-to-end distance). The solvents used are
either theta solvents or good solvents (these concepts are explained in detail in Gedde and Hedenqvist
2019a). For a linear polymer in a theta solvent: r / M1/2 and thus Vhd / M3/2, where the
proportionality coefficient depends on the segmental flexibility (the characteristic ratio (C)) at the
prevailing temperature. For a linear polymer in a good solvent: r / M3/5 and Vhd / M9/5, which is a
stronger molar mass dependence than that in a theta solvent. The hydrodynamic volume increases
with increasing solvent power. A perfectly suitable solvent causes a maximum expansion of the
4.3 Size Exclusion Chromatography 181

polymer coil with respect to the theta state. It is important to note that the molecular architecture, the
long-chain branching, has a pronounced effect on the relationship between Vhd and molar mass; the
hydrodynamic volume of a long-chain branched polymer molecules is lower than that of a linear
polymer with the same molar mass. Further information about the effect of long-chain branching on
Vhd and SEC data is presented in Sect. 4.3.6. The fundamental aspects of chain conformation and
hydrodynamic volume are discussed by Gedde and Hedenqvist (2019a). The best separation is
achieved in the molecular range where the diameter of the polymer coils or spheres is close to the
pore size of the column. The separation process can be made more effective by using a series of
columns with different pore sizes. Another option is to use ‘mixed bed’ columns, which contain a
blend of particles with different pore sizes. This may, however, lead to lower resolution, with a pore
volume less than that of a system consisting of several columns coupled in series.

4.3.4 Detection of the Eluting Molecules

The amount of polymer sample injected is small, and a sensitive detector is required to detect the
polymer molecules when they elute from the column. The detectors are commonly divided into
concentration-based, structure-selective and molar mass-sensitive detectors (Barth et al. 1998).
Concentration-based detectors are universal, and they measure a change in a physical property of
the eluting mobile phase. The most common detector used for SEC is a refractometer (RI), which
measures the difference between the refractive index of the pure solvent and that of the elute
containing polymer molecules, and the polymer has a refractive index different from that of the
solvent used, and the refractive index of a polymer is constant when the molar mass is higher than
1000 g mol–1. The detector response is therefore proportional to the concentration of the polymer. The
structure-sensitive class includes UV and IR detectors measuring the absorption of radiation in,
respectively, UV and IR spectral ranges. These detectors are sensitive to structure in a precise
manner. A UV detector is very sensitive, but its applicability is restricted to polymers that contain
groups which absorb UV radiation at the wavelengths of the detector. An IR detector is more
universal, but it is also more expensive, although it provides additional information about the
composition. UV and IR detectors respond in proportion to the concentration of polymer in the
elute. An SEC instrument equipped with an RI, UV or IR detector gives only relative molar mass
values, and the instrument thus requires calibration with standard compounds with known molar
masses.
There are also molar mass-sensitive detectors such as viscometers and light-scattering detectors
(cf. Chapter 3). In addition to being molar mass-sensitive, these detectors also provide information
about the concentration of the eluted polymer. The viscometer detector assesses the intrinsic viscosity
[η] of the eluted polymer solution, and if the viscometer is combined with a RI detector, it is possible
to obtain information about the absolute molar mass by applying universal calibration (cf. Sect. 4.3.6)
and also to predict the concentration of long-chain branches (Gaborieau and Castignolles 2011).
Triple-detector SEC combines RI, an online capillary bridge differential viscometer and a low-angle
laser light-scattering (LALLS) detector, where the viscometer gives the specific viscosity and the
LALLS provides the limiting excess Rayleigh ratio (Wyatt 1993). This combination of detectors thus
makes possible the continuous measurement of concentration, intrinsic viscosity [η] and molar mass
and allows the direct application of universal calibration. Multi-angle laser light-scattering (MALLS)
detectors are available for the simultaneous measurement of light scattering at several angles, giving
further information on molecular conformation, molecular architecture and copolymer structure
(Buchard 1999). As an example, SEC–MALLS has been utilized for assessing the long-chain
branching in low-density polyethylene (Tackx and Tackx 1998) and in polyethylene copolymers
(Stadler et al. 2006). The theoretical basis for the analysis of long-chain branching in polyolefins by
182 4 Chromatographic Analysis of Polymers

triple detector SEC has been reviewed by Liu et al. (2017), and the effect of both the degree and the
local distribution of long-chain branching on the analysis has been reviewed by Gaborieau and
Castignolles (2011). The recent study by Orski et al. (2020), which applied multidetector SEC to
model linear low-density polyethylene-like compounds with precisely controlled alkyl branch fre-
quency and length, is recommended reading.

4.3.5 Relative Narrow Standard Calibration

The separation in an SEC column is related to the hydrodynamic volumes of the molecules to be
separated. The separation achieved leads to elution of the molecules from the SEC column at different
times, i.e. at different retention times or different retention volumes. The retention time is equal to the
time it takes for a molecule with a given hydrodynamic volume to travel through the column, and the
retention volume is the amount of solvent that is pumped through the system during the passage of the
molecule through the column. There is, therefore, a direct correlation between the two, but to relate
the retention time to the actual molar mass requires calibration (Konstanski et al. 2004). This means
that SEC is usually a relative rather than an absolute technique for molar mass determination.
The simplest and most common calibration is relative calibration, also called relative narrow
standard calibration. This involves running a set of well-characterized standard compounds, each
with a different known molar mass and narrow molar mass distribution through the system. The
logarithm of the molar mass is then plotted against the retention time or retention volume so that a
specific molar mass is related to a specific retention time or retention volume (Fig. 4.10). The problem
is that different polymers with the same molar mass can have different hydrodynamic volumes in the
SEC solvent used. The hydrodynamic volume is influenced not only by the molar mass but also by the
solubility of the molecule and by its molecular architecture (e.g. linear, long-chain branched,
hyperbranched, dendritic) and the flexibility of the polymer chain. Branching usually decreases the
hydrodynamic volume (Mourey et al. 1992). If a branched and a linear polymer have the same
repeating unit and the same molar mass, the branched polymer will elute later due to its smaller
hydrodynamic volume (Gaborieau et al. 2007). Ideally, the standards used for calibration should
therefore consist of the same polymer as that requiring analysis. In practice, this is not usually
possible. It is challenging to produce SEC standards as they need to be monodisperse, and this means
that we are usually limited to a few different commercially available standard compounds. The most

Fig. 4.10 SEC

chromatogram with four
narrow standards and a
relative narrowly standard
calibration curve. Drawn
after Mori and Barth (1999)
4.3 Size Exclusion Chromatography 183

common standards are polystyrene standards, since PS can be polymerized by living anionic
polymerization leading to narrow dispersity polymers. In other cases, expensive fractionation is
utilized to prepare standards with a narrow molar mass distribution. The dispersity should be less
than 1.10 for a calibration standard to be considered suitable for SEC calibration.
This calibration technique is called relative calibration because the molar mass determined is
relative to that of the calibration standard. For example, if the column is calibrated with narrow
polystyrene standards, the molar mass obtained for an unknown polymer is based on the retention
times for the PS standards, and there will be a relative error. This is acceptable in most cases and for
comparative purposes. PS standards are the most common, but other standards are available such as
poly(methyl methacrylate), polybutadienes, polylactide, poly(ethylene glycol) and pullulans
(polysaccharides).

4.3.6 Universal Calibration

Viscometry is often performed on dilute polymer solutions with an Ubbelohde viscometer, which
measures the time (t) for a given volume of solution to flow through a capillary. The flow-through
time (t0) for the pure solvent is also measured in the same viscometer. Measurements are made on a
series of solutions with different polymer concentrations, and the intrinsic viscosity ([η]) is obtained
by extrapolation of the data to zero polymer concentration (C) according to:

lim η  η0 lim t  t0
½η  ¼ ¼ ð4:13Þ
C ! 0 C  η0 C ! 0 C  t0

where η and η0 are the viscosities of, respectively, the solution and the pristine solvent. The change in
viscosity when a small impenetrable rotating spheres house the polymer molecule in a suspension was
dealt with by famous Albert Einstein (1906) resulting in the simple equation:


5ϕ2
η ¼ η0  1 þ ð4:14Þ
2

where ϕ2 is the volume fraction of the suspended spheres, which is given by:

3 3
N 4πr eq 5 4π N r eq
ϕ2 ¼  ) ½η  ¼   A  ð4:15Þ
V 3 2 3 100 M

where V is the volume of the solution, NA is the Avogadro number, M is the molar mass and N is the
number of spheres, each with a volume of 4πreq3/3, and req is the effective radius of the spheres. It
should be noted that the increase in viscosity is caused by rotation of the spheres in the shear field and
that req is related to the end-to-end distance (r) and also to the hydrodynamic volume of the spheres.
Equation (4.15) can be further rearranged to reveal another way of expressing the volume of each
suspended sphere (Veq):

 ! !
5 NA 4πr 3eq 4πr 3eq
½ηM ¼  ) ½ηM / ¼ V eq ð4:16Þ
2 100 3 3
184 4 Chromatographic Analysis of Polymers

Fig. 4.11 Calibration curves prepared with different narrow dispersity standard compounds for (left) relative
(traditional) calibration versus (right) universal calibration

It can be shown that the different radii associated with the molecular coils are proportional to each
other (Gedde and Hedenqvist 2019a):

r eq / r / s / r ob ð4:17Þ

where r is the end-to-end distance, s is the radius of gyration and rob is the radius of the outer border of
the enclosing sphere defining the hydrodynamic volume (Vhd) of the polymer sphere. It can thus be
concluded that:

3 
4πr ob
½ηM / ¼ V hd ð4:18Þ
3

This expression forms the basis of the universal calibration method introduced by Grubisic et al.
(1967). If the logarithm of the molar mass of the calibration standards is plotted against the retention
time, the lines for different polymers in a given solvent do not coincide (Fig. 4.11), but if instead the
logarithm of the product of the intrinsic viscosity [η] and molar mass is plotted vs. retention time, a
single calibration curve is obtained, independent of polymer type. The points of all polymers fall on a
single line. This calibration utilizes the fact that the product [η]M is proportional to the hydrodynamic
volume (Eq. (4.18)). This means that, for two molecules with the same hydrodynamic volume, the
product [η]M is the same, which allows the assessment of the molar mass of polymer 2 provided that
the molar mass of the polymer standard (polymer 1) and the ratio [η1]/[η2] are known:

½η1 
½η1 M1 ¼ ½η2 M2 ) M2 ¼  M1 ð4:19Þ
½η2 

For both polymers, a calibration curve based on standards has to be established. Plotting the
intrinsic viscosity against the logarithm of the molar mass gives the Mark–Houwink–Sakurada plot,
from which two important constants, a from the slope and log K from the intercept, can be
determined. This gives the Mark–Houwink–Sakurada equation:
4.3 Size Exclusion Chromatography 185

½η ¼ KMa ð4:20Þ

By combining Eqs. (4.19) and (4.20) – the latter for both polymers 1 and 2 – the following
expression is obtained:




1 K1 1 þ a1
log M2 ¼  log þ  log M1 ð4:21Þ
1 þ a2 K2 1 þ a2

Values for K and a for common polymers can be found in Polymer Handbook (Brandrup et al.
1999). If these values are included in the SEC calculations, a more accurate molar mass for the
unknown polymers can be determined. Universal calibration gives good values for normal polymers
forming random coil structures, but it still does not work for polymers with more special molecular
architectures such as sphere-shaped or long-chain branched molecules. However, universal calibra-
tion can also be used to determine the degree of branching, especially long-chain branching. Long-
chain branching has a significant effect on the hydrodynamic volume and is often quantified by the g-
factor according to g ¼ hs2iLCB/hs2iL, where s is the radius of gyration and indices LCB and L refer to
a long-chain branched polymer and a linear polymer, respectively, both polymers having the same
molar mass. The factorial change in hydrodynamic volume due to long-chain branching is thus g3/2;
note that g < 1. Data for the g-factor for different long-chain branched polymer structures are
available (Kramers 1946; Zimm and Stockmayer 1949). Recommended reading about long-chain
branching is Scholte (1983).

4.3.7 Sample Preparation for SEC

SEC analysis is a separation method based on the hydrodynamic volume of the polymer in the selected
solvent, which is also influenced by the concentration of the polymer. This means that the choice of
solvent and the concentration are important for obtaining correct results. In practice, the option is
limited to the solvent used in the available SEC system. Some polymers are soluble in many solvents
at room temperature, but others are difficult to dissolve and require elevated temperatures with a
specific solvent, limiting the possibilities of SEC analysis. The ideal sample concentration depends on
the type of sample, its molar mass and the solvent used. The concentration should be high enough to
get a good detector signal, but not so high that the column becomes overloaded. The ideal sample
concentration depends on the molar mass, but a good rule of thumb is 1.0 mg (mL)–1 for the
preparation of the calibration standard and unknown sample solutions. SEC columns are sensitive
to particles, so the samples should always be filtered before injection. A 0.45 μm membrane filter is
often an adequate choice. If viscometry or a light-scattering detector is used, the exact mass injected
must be determined.

4.3.8 Molar Mass Averages from SEC Chromatograms

Modern software is capable of rapidly calculating the molar mass averages for the analysed samples.
Manual calculation requires several steps: first, the baseline is drawn, and the area under the
chromatogram is divided into equally thick slices (Fig. 4.12), and the height of each slice from the
baseline to the chromatogram intersect is measured. Note that slices correspond to mass fractions per
log molar mass and that the slices are uniformly distributed over a log molar mass scale. For the
186 4 Chromatographic Analysis of Polymers

Fig. 4.12 Calculation of

molar mass averages with
the help of a SEC
chromatogram and a
narrow standard calibration
curve. Drawn after Mori
and Barth (1999)

calculations, each slice is considered to be monodisperse. Thinner slices, i.e. a larger number of
heights measured, lead to more accurate molar mass averages. The height (Hi) of the slice i is
proportional to the mass fraction with molar mass (Mi), and Hi/Mi is thus proportional to the number
fraction of molecules with molar mass Mi. The molar mass of each individual slice can be obtained
from a calibration curve with the help of the retention volume for that slice (Vi) (Fig. 4.12). The molar
mass averages are then calculated as follows:
P
Hi
i
Mn ¼ P   ð4:22Þ
H i =M∗i
i

P 
H i M∗
i
i
Mw ¼ P ð4:23Þ
Hi
i

P 
H i M2∗
i
i
Mz ¼ P  ð4:24Þ
H i M∗i
i

It is important to note that the y-axis in the SEC chromatogram (Fig. 4.12) is dw/d(log M ) and that
the Mi values are distributed evenly in log M space.

4.3.9 Development of New Coupled SEC Systems

In order better to characterize more complex polymers with multidimensional heterogeneity

(e.g. through molar mass, molecular architecture and repeating unit structure), several other
4.4 High-Performance Liquid Chromatography 187

hyphenated SEC systems coupled to other procedures have been developed (Pasch 2013). One
example is SEC coupled with electrospray ionization mass spectrometry (SEC–ESI–MS) to obtain
more detailed structural information, including the analysis of end groups, copolymer composition,
molar mass distribution and even the analysis of single-chain nanoparticles (Jovic et al. 2019;
Winkler et al. 2012; Steinkoenig et al. 2017). Coupling with another mass spectrometric technique,
matrix-assisted laser desorption ionization mass spectrometry (MALDI–MS) makes it possible to
fractionate a sample into narrow fractions before the MALDI–MS analysis (Montaudo et al. 2006;
Murgasova and Hercules 2003), in order to remove the mass discrimination caused by easier
ionization of low molar mass fractions. Several types of interface have been evaluated to couple
SEC with nuclear magnetic resonance (NMR) spectroscopy (Hiller et al. 2014; Uliyanchenko 2017).
SEC–NMR coupling can determine the chemical composition for each SEC fraction and provide, for
example, the molar mass distribution and a structural analysis (focussing on the sequence of repeating
units) of copolymers (Hiller et al. 2010).
The concept of two-dimensional liquid chromatography (LCLC), where two liquid chromatog-
raphy techniques with different separation mode are combined, has also been expanded to the
analysis of polymers (LCSEC). This analysis includes, in addition to the molar mass distribution,
an analysis of other features such as the copolymer composition (Pirok et al. 2018; Raust et al. 2008;
Uliyanchenko et al. 2012). Interaction chromatography (IC) coupled with SEC has, for example, been
utilized to characterize lignosulphonates with respect to their chemical compositions (degree of
sulphonation) and molar mass distribution (Musl et al. 2020) and for the analysis of molar mass,
conformation and long branching in a series of polyethylenes (Plüschke et al. 2020). In the case of
polyolefins, multidimensional high-temperature chromatography has also been applied to improve
the separation of mixtures and copolymers with different molecular architectures (linear chains or
chains with both short-chain and long-chain branching (Ginzburg et al. 2011; Pasch and Trathnigg
2013). Thermal field-flow fractionation (ThFFF) has recently been online-coupled with SEC to
provide a ThFFF separation based on chemical composition followed by a SEC analysis of each
fraction (Viktor and Pasch 2020).

4.4 High-Performance Liquid Chromatography

Liquid chromatography can be divided into planar (e.g. thin-layer chromatography) and column
chromatography. This section deals only with column liquid chromatography and especially high-
performance liquid chromatography (HPLC). The acronym HPLC originally indicated high-pressure
liquid chromatography since a high pressure was used to achieve the flow through a packed column.
Fast development during the 1970s, when new injectors, pumps and columns were developed, led to
the creation of modern HPLC with an improved performance and a change of name to high-
performance liquid chromatography. The acronym, however, remained the same (Ettre and Horvath
1975). HPLC is among the most powerful analytical tools. It is suitable for separation and analysis of
organic compounds in the molecular range up to 2000 g mol–1 and down to extremely low analyte
concentrations, e.g. ppt levels. It is used for routine and research purposes in numerous fields,
including the analysis of all kinds of low molar mass compounds present in or migrating from
polymeric materials, such as additives, oligomers and degradation products or, for example, drugs,
antibacterial agents, fertilizers and pesticides from various slow-release formulations (Block et al.
2006, Simal-Gandara et al. 2002). The main requirement is that the compounds to be analysed must
be soluble in the mobile phase. If a UV detector is used, they must also have UV-absorbing groups in
the spectral range of the instrument. An advantage of HPLC compared to GC is that it is possible to
directly inject an aqueous solution and it also offers larger molar mass range and better possibilities to
analyse highly polar and thermally labile compounds. The HPLC systems can be divided into reverse
188 4 Chromatographic Analysis of Polymers

and normal phase, ion exchange and size exclusion (cf. Sect. 4.3) depending on the type of interaction
between the analyte and the column leading to the separation and retention of compounds. Analytical
scale instruments are the most commonly used, but preparatory scale systems are also available
(Wellings 2005), and these can be used for fractionation and purification of samples for further
analysis or use. Liquid chromatography and its applications are described in detail in several
dedicated books: Fanali et al. (2017) and Horváth (1988).

4.4.1 Principles of an HPLC System

An HPLC system is a chromatographic system that greatly resembles that of an SEC system
(Fig. 4.7). The main differences are in the type of column and the type of detector used. The main
components of the system are solvent and waste reservoirs for the mobile phase, high-pressure pump,
injector, HPLC column and detector and a computer to collect and process data.
During the analysis, the high-pressure pump generates a specified flow rate for the mobile phase,
and an injector introduces the sample to be analysed into the moving mobile phase which transports
the sample into the column where the different analytes are separated. The column contains a
stationary phase and a packing material that interacts with the analytes through different types of
interaction depending on the column material and on the mobile phase. A concentration-sensitive
detector detects the eluting compounds, and the retention times of the individual peaks are used to
identify compounds, while the peak areas are used to assess the concentrations of the compounds.
This is discussed in more detail in Sect. 4.6.
There are different detectors, but a UV-absorbance detector is most commonly used for
compounds with UV absorption at the wavelength(s) of the detector. If the compound fluoresces, a
fluorescence detector can be used. As a result, a chromatogram is obtained that presents the separation
that took place in the HPLC system. A chromatogram of a mixture of compounds contains a series of
peaks, each of which represents one of the components in the mixture. Ideally, these peaks are well-
separated in as short time as possible. The early eluting peaks are usually narrow, while the later
eluting peaks are wider and lower in height, because the detector signal is related to the concentration
and later eluting peaks are due to substances that require more time to be completely eluted from the
column. HPLC can also be coupled to a mass spectrometer (MS), and the system is then denoted
LC-MS (McMaster 2005). The MS coupling gives significantly more information as mass spectra can
be obtained for each eluting peak. The identification and quantitation of compounds is discussed in
more detail in Sect. 4.6.
Two elution modes, isocratic and gradient elution, are generally used in HPLC. In the isocratic
mode, the mobile phase is a single solvent or mixture of solvents that is kept constant during the
whole analysis. In gradient mode, the composition of the mobile phase is changed during the analysis,
in order to increase the elution strength of the mobile phase during the analysis. This makes it possible
to separate the early eluting compounds, and it accelerates the elution of late eluting compounds,
making it possible to analyse a wider range of compounds with a greater difference in polarity and
molar mass. The overall resolution of an HPLC column depends on its mechanical separation power,
determined by the column length and particle size (Fig. 4.13), and on its chemical separation power
controlled by the interactions of the compounds to be separated with the stationary phase. The
stronger the interaction with the stationary phase, the longer is the compound retained in the column.
A smaller particle size leads to a higher backpressure and a higher efficiency, i.e. the delivery of more
mechanical separation power in the same time. Efficiency can also be increased by increasing the
column length, although this also increases the analysis time and the consumption of solvent.
The most common applications of HPLC are for the identification and quantitative analysis of
compounds (analytical HPLC), but it is also possible to use HPLC to purify and collect compounds
using a fraction collector. This kind of preparative chromatography can be performed with a normal
4.5 Gas Chromatography 189

Fig. 4.13 Illustration of the effects of column length and particle size on the mechanical separation power. The column
properties are (a) column length 50 mm, particle size 5 μm; (b) column length 50 mm, particle size 1.7 μm; and (c)
column length 100 mm, particle size 5 μm

HPLC system, but to obtain large amounts of purified analytes, a special system with a large
preparative column and a large pump are used. Typical analytical columns have a length of
20–500 mm and an inner diameter of 1–100 mm, whereas columns with a length of 50–100 mm
and an internal diameter of 10–40 mm are commonly used for preparative work.

4.4.2 Separation Modes in HPLC

Polarity or electrical charge can be utilized to achieve separation in HPLC. Two primary separation
modes, normal-phase and reverse-phase chromatography, utilize polarity (Fig. 4.14), whereas ion
exchange utilizes either positive or negative charges. The type of functional groups in the molecule
generally determines its polarity or the presence of an electrical charge. For chromatographic
separation in the normal- or reverse-phase mode, mobile and stationary phases with opposite
polarities are chosen. In normal-phase chromatography, the mobile phase is non-polar, and the
stationary phase has polar functional groups, whereas in a reverse-phase system, a polar mobile
phase (often water, an aqueous buffer, methanol, acetonitrile or a mixture of them) and a non-polar
column material are utilized. During the chromatographic analysis, the sample components that have
a polarity similar to that of the stationary phase are retained due to the stronger attractions between the
analytes and the column material (Fig. 4.14). Compounds with a polarity similar to the mobile phase
remain in the mobile phase and move faster through the column. The column material (stationary
phase) in HPLC is often silica with a surface functionality suitable for the preferred separation mode.
For example, non-polar n-octylsilyl- [C8] and n-octadecylsilyl- [C18] moieties are common surface
modifications in reverse-phase columns. These are good general columns for a wide variety of
compounds. Around 75% of all HPLC analyses are performed in the reverse-phase mode, due to
the wide range of compounds that are suitable for this mode in combination with better reproducibil-
ity and column stability.
For separations based on polarity, like is attracted to like, and for example, polar compounds are
retained longer in a polar column. In separation by ion-exchange chromatography, opposites attract
each other. A cation-exchange column retains positively charged ions by a stationary phase that is
functionalized by surface groups with a negative charge. An anion-exchange column retains nega-
tively charged ions by a stationary phase functionalized by surface groups with a positive charge.

4.5 Gas Chromatography

Gas chromatography is the method of choice for the physical separation and analysis of volatile
thermally stable organic and inorganic compounds. The upper temperature limit is usually not greater
than 350  C, which means that the compounds to be analysed must be at least partly volatile below this
190 4 Chromatographic Analysis of Polymers

Fig. 4.14 Normal-phase and reverse-phase liquid chromatographic separation

temperature. In addition, the compounds or analytes must be thermally stable in the temperature range of
the analysis. Two of the major advantages of gas chromatograph are its ability to rapidly detect and
analyse extremely small quantities of the analytes and its high separation power. Modern GC was
invented by James and Martin (1952). The history and development of GC is summarized in an
interesting review by Bartle and Myers (2002). Many books have been devoted to the principles and
applications of GC (Jennings et al. 1997; Poole 2012) and GC–MS (McMaster 2011).

4.5.1 Principles of a GC System

The most important components of a gas chromatographic system include an inert carrier gas, a flow
controller to ensure a constant flow of carrier gas, a sample injector to introduce the sample, a column
in a temperature-programmable oven and a detector providing a signal proportional to the amount of
analyte (Fig. 4.15). The inert gas, such as nitrogen or helium, is used as the carrier gas. The purpose of
the carrier gas is to carry the analytes through the column and further to the detector. The common
mode of sample injection is direct injection often with the help of a split/splitless injector. The
injector is usually operated at elevated temperature, the liquid sample introduced being immediately
vapourized and moved partly (split injection) or, in some cases with trace analysis, fully (splitless
injection) into the column. A typical injection volume is 1 μL. Although the injection volume is small,
only ca. 1% of the injected sample is usually moved forward to the column, the rest being vented to
avoid column overload. Other injectors include headspace injection, programmed temperature
vapourizing injection, on-column injection and purge and trap. These injectors can be selected for
various reasons, such as the analysis of thermally labile compounds or gaseous samples.
The injected sample mixture moves through the column, and the compounds are separated as a
result of their partitioning between the mobile (gas phase) and the stationary phase (Fig. 4.16). A
compound with a boiling point higher than the temperature in the column condenses at the start of the
column and moves into the column as the column temperature increases. Some compounds are more
soluble in the stationary phase than others. The more soluble compounds spend a longer time
absorbed in the stationary phase, whereas the less soluble compounds spend a longer time in the
mobile phase. The analytes move forward only when they are present in the mobile phase. In practice,
the separation of the compounds is based largely on their boiling points, and it is further influenced by
the polarity of the analytes. Less volatile compounds spend a longer time in the stationary phase and
are retained in the column for a longer period of time. Thus, compounds with a high boiling point
have a longer retention time.
4.5 Gas Chromatography 191

Fig. 4.15 The principal components of a gas chromatography system

Fig. 4.16 Travel of the

analyte through a GC
column by partitioning
between stationary and
mobile phases in the
column

A higher column temperature facilitates the transfer of molecules into the gas phase and shortens
the retention times of all the analytes. A low column temperature leads to better separation, but,
negatively, it may take a long time before all the analytes pass through the column to complete the
analysis. The normal solution is to use a temperature program. At the start of the analysis, when the
sample is injected, the column is at a relatively low temperature to obtain a better separation of the
more volatile compounds. The temperature is then gradually increased to elute even the less volatile
compounds. Typical oven temperatures are in the range of 40–300  C.
In most cases, open tubular capillary columns are used for gas chromatographic analysis. These
are typically 5–30 m long with an inner diameter of 1 mm or less (typically 0.25–0.32 mm). The inner
surface of the open tubular capillary column is coated with the stationary phase, a thin liquid film or
192 4 Chromatographic Analysis of Polymers

cross-linked material. The thickness of the stationary phase is commonly 0.25 μm, but it can vary
between 0.1 and 5.0 μm. Many different stationary phases are available, ranging from low polarity
(e.g. 100% polydimethylsiloxane) to mid-polarity (polydimethylsiloxane with 35–50% diphenyl
substitution) and highly polar phases (e.g. polyethylene glycol and nitroterephthalic acid modified
polyethylene glycol). The selection of the column depends on the type of sample to be analysed.
Usually, a polar column is preferred for polar analytes and a non-polar column for non-polar analytes.
Mid-polar phases can be good ‘universal columns’ for mixtures. Non-polar columns are generally
more stable, tolerate higher temperatures and have a higher efficiency. The column temperature is
usually kept 10–15  C below the maximum temperature to avoid column bleed and to increase the
column life.
A detector is located at the end of the GC column in order to sense the compounds exiting the
column. The detector response created should be proportional to the amount of eluted analyte. The
most common and most universal detector for gas chromatography is the flame ionization detector
(FID), which is a general detector for all organic compounds. It has a wide linear range (important for
quantitative analysis) and a high sensitivity, and it is relatively inexpensive. However, an FID
detector is destructive as the sample is burnt in a flame.
Other GC detectors include element-specific detectors like the nitrogen-phosphorus detector
(NPD) (for analysis of N, P) and flame photometric detectors (FPD) (for analysis of S, P, N, Sn).
A thermal conductivity detector (TCD) measures the thermal conductivity of the effluent. A mass
spectrometer (MS) is commonly coupled to GC due to its high sensitivity and selectivity and the
possibility to collect mass spectra for each eluted analyte, which enables a more reliable identification
of unknown compounds. The three main goals of optimized GC analysis are a good resolution, a short
analysis time and a large sample capacity. Unfortunately, full optimization of one of these parameters
usually leads to a negative influence on the other two. Optimized GC analysis is therefore a
compromise of these three parameters. A fast analysis time, for example, can be achieved with a
shorter column, a higher column temperature and a faster flow of carrier gas. All these changes at the
same time affect resolution, and the influence can be complex. Increasing the temperature, for
example, can decrease the resolution for low boiling compounds and increase the resolution for
high boiling compounds due to sharpened peaks and improved peak shapes.

4.5.2 Integrating GC with Mass Spectrometry (MS)

GC by itself is a powerful separation technique, but its value is multiplied if GC is coupled to a mass
spectrometer which has become a universal detector for gas chromatography. The mass spectrometer
receives the injected material, ionizes it in high vacuum, focuses these ions and their fragmentation
products through a magnetic mass analyser and then collects and measures the amount of each ion.
This yields the mass spectrum for each separated analyte in the chromatogram, which provides two
modes of identification, by the retention time and the mass spectrum. These data have entirely
different molecular origins, and the use of this combination enhances the reliability of the identifica-
tion. The molecular peak and fragment peaks form a typical pattern based on the chemical structure of
the compound, and the pattern can be assigned to a specific structure by an experienced mass
spectroscopist or by comparison against large spectral libraries. However, the analysis of a
corresponding standard compound to confirm both the mass spectrum and retention time is
recommended. Despite the power of GC–MS analysis, the volatility, polarity and thermal stability
of the compounds to be analysed are some parameters that may limit the gas chromatographic
analysis. Another limitation is that sample preparation is, in most cases, required before GC analysis.
Derivatization can be utilized to make the samples more thermally stable, less polar and/or more
volatile, but it adds a further step in the sample preparation.
4.6 Qualitative and Quantitative Analysis by HPLC and GC 193

4.6 Qualitative and Quantitative Analysis by HPLC and GC

HPLC or GC provides both qualitative and quantitative information of the compounds to be analysed
(Guiochon and Guillemin 1988). The retention time can be used for identification, whereas the peak
area is an indication of the amount of analyte (Fig. 4.17). In plain GC or HPLC analysis, the
compounds are identified with the help of retention time. This is often an adequate tool to confirm
the presence or absence of specific compounds. However, the identification of totally unknown
compounds is not possible, and retention time-based identification lacks definite proof because
several compounds may have the same retention time. If GC or HPLC is coupled with MS, both
the retention time and the mass spectrum are available for each eluting peak, and this increases the
reliability of the peak identification for a totally unknown sample (Fig. 4.17) (McMaster 2005, 2011).
It should also be kept in mind that the absence of a peak for a certain compound can depend on an
error during sample preparation or on the unsuitability of the compound for the column and other
experimental parameters selected.
For a quantitative analysis, a calibration curve is required. The amount of analyte passing from
injector to GC column is usually related to the volatility of the analyte, but in HPLC, the detector

Fig. 4.17 Identification and quantification of compounds by chromatography and/or mass spectrometry
194 4 Chromatographic Analysis of Polymers

Fig. 4.18 External calibration curve where the detector response is plotted against the amount of the calibration
standard

response is typically dependent on the chemical structure of the compound. This means that, ideally, a
calibration curve should be established separately for each compound. The simplest calibration is
performed with external or internal standards. In the external standard method, solutions with
different amounts of the standard compound are prepared and analysed by GC, and the detector
response is plotted against the amount of standard compound (Fig. 4.18). This response should
preferably be linearly dependent to the amount of analyte. This type of calibration is suitable for
simple gaseous or homogeneous liquid samples. An internal standard can be added to the sample to be
extracted to correct for errors or variations during extraction and injection.
For more complicated liquid matrices or for solid matrices (e.g. during headspace sampling), it is
necessary to use other calibration methods such as multiple headspace extraction (Kolb and Pospisil
1977; Kolb and Ettre 1997; Hakkarainen 2007) or a standard addition to correct for different matrix
effects. Kolb and Pospisil first presented a technique called discontinuous gas extraction (Kolb and
Pospisil 1977), which was later called multiple headspace extraction (MHE) (Kolb and Ettre 1997).
The method eliminates matrix effects and makes possible the quantitative determination of analytes
in solid matrixes by headspace techniques. In this method, multiple subsequent headspace extractions
from the same sample are performed, and the total amount of analyte is then theoretically calculated
(Kolb and Ettre 1997; Hakkarainen 2007). For hydrogen bonding solid matrices, such as polyamides,
it can be difficult to achieve a reliable analysis even with these methods (Gr€oning and Hakkarainen
2004).

4.7 Sample Preparation Before GC or HPLC Analysis

Sample preparation is commonly required prior to a GC and HPLC analysis because most samples
are either complex, dilute, contaminated or incompatible with the particular chromatographic system
used (Moldoveanu and David 2015; Albertsson and Hakkarainen 2008). This is a crucial step for
obtaining correct results. Incorrect sample preparation can exclude compounds of interest or change
the proportions of the components in the sample and thus lead to false conclusions. The extraction
techniques are divided into solvent extraction techniques (liquid–solid and liquid–liquid extractions)
and techniques that use no or only a small amount of solvent (Nerin et al. 2009). The choice of method
depends on the nature of the analytes and the matrix from which the analyte is extracted (e.g. water,
air, solid). In the case of polymers, it can be of interest to analyse compounds present in the sample
matrix or compounds that are migrating or have migrated to the surrounding environment. Depending
4.7 Sample Preparation Before GC or HPLC Analysis 195

on the application, this could, for example, be outdoor or indoor air, food, medicine or cosmetics
inside the packaging, the human body, agricultural soil, oceans, etc. The extraction and quantitative
analysis of volatile and semi-volatile substances in solid or complex matrices is an analytical
challenge. The nature of the matrix and the products to be extracted determine the choice of
extraction method. An ideal extraction method is quantitative, selective, and rapid, uses little or no
solvent and does not require expensive instrumentation.
Traditional liquid–solid extractions (LSE) include Soxhlet, dissolution–precipitation, sonication
and microwave-assisted extractions (MAE). Liquid–liquid extraction (LLE) can be utilized for the
extraction of compounds from aqueous solutions. These techniques often require rather large amounts
of solvent, especially the Soxhlet and LLE procedures. Other methods such as headspace extraction,
solid-phase extraction (SPE) and microextractions such as solid-phase microextraction (SPME)
developed by Arthur and Pawliszyn (1990) require little or no solvent. Several books deal with
liquid-phase extraction (Poole 2020) and solid-phase extraction including SPME (Poole 2020;
Pawliszyn 2012).
Dissolution–precipitation is a simple technique where the polymer is first fully dissolved. After
dissolution, a second solvent is added to precipitate the polymer, while additives remain soluble due
to their lower molar mass. Soxhlet extraction is another traditional method for the extraction of low
molar mass products and additives from solid polymers, where the sample is placed in ‘a sock’ or
thimble, which is bathed with hot solvent. The insoluble matrix remains in the thimble, while the
analytes are transferred to the solvent. The extracting solvent is repeatedly vapourized, condensed
and allowed to percolate through the solid sample. This method is time-consuming (2 or 3 days is not
uncommon), is non-selective, uses large volumes of solvent and is often not quantitative. Some
analytes may also be damaged by the prolonged exposure to hot solvent. Ultrasonication-assisted
extraction produces agitation which increases the rate of transfer across the solid–liquid boundary
layer. Microwave-assisted extraction (MAE) is another effective method of extracting (semi-)
volatiles from solid matrices. It is suitable, for example, for the extraction of additives from solid
polymer matrices (Alin and Hakkarainen 2010) or for the extraction/fractionation of complex
biopolymers such as lignin (Cederholm et al. 2020). Samples are heated and immersed in a
microwave-absorbing solvent in closed vials. Microwave radiation can heat the solvent to a tempera-
ture above its boiling point providing rapid and effective extraction. Advantages compared to Soxhlet
extraction are the significantly faster extraction and the lower solvent consumption. Also, several
samples can be extracted simultaneously. MAE has been used particularly for the extraction of
antioxidants and degradation products from packaging materials.
Liquid–liquid extraction (LLE) is the traditional way to extract organic compounds from water,
where low molar mass compounds are transferred from one liquid phase to another immiscible liquid
by shaking in a separation funnel. LLE generally has low selectivity. It is also labour-intensive, uses
large amounts of organic solvents and requires a concentration step. Solid-phase extraction (SPE) is a
related technique somewhat analogous to LLE, but it utilizes a solid sorbent inside a small column
(Fig. 4.19). SPE is much more selective than. LLE and uses only a small amount of organic solvent
(Waters Corporation 2014). SPE can separate complex mixtures to facilitate GC or LC analysis. The
selectivity of an extraction depends upon the chemical structure of the analyte and the properties of
the solid phase and sample matrix. Compounds that have a stronger interaction with the solid phase
than with the sample matrix are retained in the column. They can then be eluted by a small amount of
a strong solvent that is able to break the interaction. This results in simultaneous elution and
concentration of the compounds of interest. The solid phase is similar to the stationary phase inside
the HPLC column. The specific interaction leading to the retention of analytes is defined by the
functional groups bonded on the surface of the silica substrate. The types of interaction between the
analyte and the sorbent include non-polar, polar and ionic interactions. In polymer analysis, LLE and
SPE can be used to extract degradation products or additives that have migrated from polymers in
contact with aqueous solutions, such as simulated body fluids, aqueous foods or simulated compost
(Hakkarainen et al. 2000, 2007).
196 4 Chromatographic Analysis of Polymers

Fig. 4.19 Solid-phase extraction of analytes from liquid samples

Headspace extraction can be used for the analysis of volatile compounds in solid or liquid samples
such as polymers, sewage, oil, food, blood and environmental materials (Hachenberg and Schmidt
1985). Static headspace analysis indirectly determines the amounts of volatiles in a liquid or solid
sample by analysing the vapour phase that is in thermodynamic equilibrium with the sample in a
closed system. The sample is closed in a vial that is heated for a predetermined time at a
predetermined temperature to establish equilibrium between the compounds in the solid and liquid
phases and in the headspace above. A predetermined thermostating time can be used for solid samples
for which it may be difficult and time-consuming to reach equilibrium. A portion of the gas phase is
then subjected to GC or GC–MS analysis. At equilibrium, the distribution of the individual
compounds between the two phases depends on their saturation pressures at the temperature of the
vial and their concentrations in the liquid or solid phase. Headspace analysis is very useful for the
analysis of volatile substances in solid polymer samples (Gr€ oning and Hakkarainen 2008), but it has
some drawbacks such as non-selectivity, volatility requirements and the influence of the matrix on the
partitioning of the analytes to the headspace, and these facts complicate the quantitative analysis. The
dynamic headspace procedure involves sampling from the flowing gas above the liquid sample.
Solid-phase microextraction (SPME) is an inexpensive, rapid, solvent-free technique for the
extraction of organic compounds from liquid or solid samples (Pawliszyn 1997). SPME is based on
a thin fused silica fibre coated with a stationary polymeric phase mounted in a modified syringe. The
fibre is immersed directly into aqueous samples or into the headspace over the liquid or solid sample
matrix. The analytes are adsorbed onto the fibre and then thermally desorbed in the injection port of
GC. Figure 4.20 shows the SPME extraction and injection steps. SPME is an equilibrium technique,
and the analytes are therefore not completely extracted from the matrix. The method resembles
headspace extraction, but it is more selective and more sensitive, and it often has a higher extraction
efficiency for less volatile compounds. In comparison to static headspace extraction, the analyte is
removed from headspace due to sorption of the analyte onto the fibre, which leads to further
partitioning of the analyte to the headspace. In addition, for the extraction of less volatile compounds
from liquid samples, the direct immersion mode can be utilized. There are several coating materials
that can be used to influence the selectivity of the extraction. SPME has been demonstrated as a
powerful tool for the extraction of degradation products and additives from solid polymer matrices or
for the extraction of polymer migrants from aqueous solutions (Hakkarainen et al. 1997, Hakkarainen
2008a).
4.8 Application of GC and HPLC for the Analysis of Polymers 197

Fig. 4.20 Left: solid-phase microextraction and the subsequent desorption of the analytes directly into the GC column.
Right: different SPME modes with fibres immersed in a liquid phase or extraction from the headspace above the liquid
or solid samples

4.8 Application of GC and HPLC for the Analysis of Polymers

Low molar mass compounds are present in polymeric materials as additives or as reaction products
from the degradation of the polymer component and from deterioration of the additives. These
reactions occur at all stages of the polymer life cycle, although the high-temperature exposures
during melt processing and the long-term service exposure are probably the most influential periods.
The analysis of these compounds is of interest for multiple regulatory, safety and scientific reasons
(Fig. 4.21) (Bart 2006; Albertsson and Hakkarainen 2008). Furthermore, analysis can verify compo-
sition, help to reconstruct recipes of unknown materials and solve manufacturing problems. Many
industries and branches have regulations concerning the total allowed concentration of volatile
compounds, as well as the additives allowed in specific applications. Examples of such compounds
are residual monomers, additives, impurities and degradation products formed during synthesis and
processing of the product (Andersson et al. 2002).
One unique feature of the low molar mass species is that they have the potential to migrate from
the polymeric materials to the surrounding phases. The effect of such migration can be the spread of
substances causing odours, smells, irritants (cf. review by Bart 2006) and environmental hazards. GC
and HPLC have the capacity to detect low concentrations of a wide range of these substances
(covering a wide range of volatility) in the polymeric material itself and in the surrounding media.
The data obtained are important inputs to mathematical (numerical) models for the modelling the
transport properties (cf. Chaps. 5 and 7). The migration of hazardous additives such as some
plasticizers (Audourin et al. 1992; Jakubowicz et al. 1999; Clausen et al. 2004, 2007; Ekelund
et al. 2008, 2010; Hakkarainen 2008b; H€oglund et al. 2010; Bernard et al. 2014) and brominated
flame retardants (cf. review by van der Veen and de Boer 2012) to both air and aqueous media has
been studied by chromatographic methods. The migration of low molar mass compounds is especially
important in applications such as food packaging; recommended texts are those of Silva et al. (2006);
Barnes, Sinclair and Watson (2007); Nerin et al. (2013); Bhunia et al. (2013); and Rodriguez
Bernaldo de Quiros et al. (2019). Migration of compounds (antioxidant, degradation products from
198 4 Chromatographic Analysis of Polymers

Fig. 4.21 Reasons to for analysing low molar mass compounds in polymers

antioxidants, etc.) to fresh water in pipes is another important field which is largely based on
chromatographic analysis (Skjevrak et al. 2003; Wetton and Nguyen 2013; Mansoor Shaikh et al.
2018). The migration of low molar mass compounds from biomedical polymers in which chromatog-
raphy plays a vital role is a huge field which is dealt with in several excellent books and articles:
Jenkins (2007); Labarre et al. (2010); Chu (2015); Perale and Hilborn (2016); and Nahan et al. 2020
(3D printed products). The potential presence of these compounds also complicates the recycling of
old plastics and synthetic fibre, like old carpets, to high value applications (Horodytska et al. 2020;
Strangl et al. 2020).
One of the most effective ways of minimizing the environmental impact of production and use of
products and materials is to extend the lifetime of the products. Plastic pipes are an example of an
infrastructural product that should last for a very long period of time, 50–100 years. The lifetime of
polyethylene hot water pipes is controlled by the rate of migration of antioxidants from the pipe to the
surrounding media, mostly towards the aqueous phase (Karlsson et al. 1992; Smith et al. 1992). The
concentration of antioxidants in the polymer can be assessed by various methods, of which HPLC is a
prominent example. However, the solubility of phenolic antioxidants in polyethylene is low, and, in
fact, the polymer is in many cases supersaturated with antioxidant after production. Therefore, a
combination of methods is the most efficient: oxidation induction time (OIT) recording by DSC
(cf. Chap. 1) and HPLC. The first process involves internal precipitation of antioxidant (Viebke et al.
1996a) which is followed by slow migration of antioxidant to the surrounding media, in most cases
primarily to the aqueous phase; the distinction between these two processes can be assessed by
combining OIT–DSC with HPLC. By numerical fitting of diffusion models based on these data,
diffusivity and solubility parameters can be assessed and used for lifetime prediction (Smith et al.
1992; Viebke and Gedde 1998). Monte Carlo simulation has yielded data consistent with the molar
mass distribution obtained by SEC of thermally aged polyethylene samples (Viebke et al. 1994,
1996b; cf. Chap. 5).

4.9 Summary

Chromatographic methods play an important role in polymer science and engineering. They present
vital information about molar mass, molecular architecture and the nature of low molar mass species
present in polymeric materials which in turn is relevant for the assessment of long-term performance
and for lifetime predictions, the safety of the environment and human lives. Size exclusion chroma-
tography has evolved during recent decades as the most important tool for the assessment of molar
mass distribution. It is a technique which yields a comprehensive picture of the molar mass distribu-
tion and, with the use of more advanced detectors, the molecular architecture. Polymeric materials
4.10 Exercises 199

also contain low molar mass compounds, additives, which are intentionally added to bring new
character to the polymer. Low molar mass compounds can also emerge due to degradation or to
uptake from the surroundings. HPLC and GC analysis alone or coupled with mass spectrometry are
versatile and powerful tools to identify these compounds or to follow their formation or migration
from materials in contact with different environments. This provides information concerning the
long-term performance and potential safety issues related to the presence or migration of low molar
mass compounds. Identification of these compounds is also important for identifying degradation
mechanisms and understanding the interactions of materials with different environments.

4.10 Exercises

4.1. Calculate Mn, Mw and Ð for a polypropylene sample that contains 3 g of molecules with
molar mass 20,000 g mol–1, 5 g with molar mass 50,000 g mol–1 and 2 g with molar mass
90,000 g mol–1.
4.2. Assume that a monodisperse polymer, M1 ¼ 200,000 g mol–1, is mixed with a monodisperse
oligomer, M2 ¼ 500 g mol–1; 1 wt.% of oligomer after mixing.
(a) Calculate Mn and Mw.
(b) Why are Mn and Mw affected so differently?
4.3. Three different polymer samples (A–C) have been analysed by SEC (Fig. 4.22). Data from the
instrument are shown in the left-hand columns for the three samples (1–3). Which data belongs
to which sample? Explain the reasons for your answer for each combination.
4.4. Two narrow dispersity polymers have the same molar mass, but Polymer A has good solubility
in the SEC solvent, whereas Polymer B has poor solubility. Which of the polymers will elute
first? Explain the reasons for your answer.
4.5. You have three different polymer samples (A–C) that you analyse by SEC (Fig. 4.23).
Unfortunately, you confuse the sample names and end up with the three chromatograms
below (1–3). Which chromatogram belongs to which sample? Explain the reasons for your
answer for each combination.
4.6. You aim to synthesize a block copolymer by living anionic polymerization. How can SEC
reveal whether what you obtained is a block copolymer poly(A-block-B) or a blend of two
homopolymers (polyA and polyB) or a mixture of all three products? You start the living

Fig. 4.22 Connect the

SEC data with the correct
polymer sample
200 4 Chromatographic Analysis of Polymers

Fig. 4.23 Connect the SEC chromatogram with the correct polymer sample.

polymerization by preparing homopolymer polyA (blockA) which is known to have a molar

mass of 80,000 g mol–1. In the second step, you add monomer B. The blockB or polyB is
expected to have a molar mass of 20,000 g mol–1. Draw a SEC chromatogram for three different
results: (a) 100% formation of poly(A-block-B), (b) 50% poly(A-block-B) and 50% of
homopolymers polyA and polyB and (c) only homopolymers polyA and polyB.
4.7. The information in Table 4.1 was obtained by SEC analysis of PMMA in tetrahydrofuran at
25  C.
The system was calibrated by narrow-dispersity PS run under the same conditions:
MPS1 ¼ 86,000 g mol-1 eluting at 30 min, MPS2 ¼ 10,000 g mol–1 eluting at 35 min and
MPS3 ¼ 1600 g mol–1 eluting at 39 min. Calculate Mn, Mw and Ð.
4.8. The use of universal calibration assumes that the hydrodynamic volume is the same for all
polymers with the same retention time, i.e. that the product [η]M is the same for all polymers
with the same retention time. If the retention time for a monodisperse polystyrene of 97,000 g
mol–1 is 15 min, what is the molar mass of monodisperse poly(methyl methacrylate) that has
the same retention time? The Mark–Houwink–Sakurada constants, K and a, for PS are
1.0310–2 mL g–1 and 0.74. The K and values for PMMA are 5.710-3 mL g–1 and 0.76.
4.9. It is suspected that a fraction of polystyrene that should be monodisperse is in fact polydisperse.
Results from SEC analysis are reported in Table 4.2.
Calibration versus guaranteed monodisperse polystyrene standards gave the graph shown in
Fig. 4.24.
The requirement for monodispersity of the first sample was that Mw/Mn < 1.1. Was this
requirement fulfilled?
4.10 Exercises 201

Table 4.1 SEC analysis of Retention time (min) Height (Hi)(mm)

PMMA sample
30 1.0
31 6.2
32 30
33 40
34 27
35 11
36 2

Table 4.2 SEC analysis of Retention time (min) Height of the graph (Hi)(mm)
polystyrene
35 2
36 19
37 136
38 184
39 59
40 24
41 8

Fig. 4.24 Log M as a function of retention time

4.10. A series of narrow dispersity polystyrene standards dissolved in chloroform were injected into
an SEC column at 35  C yielding a set of chromatograms. Table 4.3 gives the peak elution
volumes and corresponding molar masses.
(a) Using the above data for polystyrene standards, calculate the relative molar mass for
monodisperse polymer X run under the same conditions with a retention time of 38 min.
(b) Use the Mark–Houwink–Sakurada constants, K and a, and the fact that [η]M is constant for
samples with the same elution time to correct the error resulting from different hydrody-
namic volumes for polystyrene and polymer X in the above calculation. For polystyrene,
KPS ¼ 4.9103 cm3 g1 and aPS ¼ 0.79. For polymer X, KX ¼ 5.4 103 cm3 g–1 and
aX ¼ 0.77.
4.11. Aliphatic polyester, polylactide (PLA), is the most promising commercial biobased and
biodegradable thermoplastic material. Which chromatographic method would you choose for
the following analytical tasks. Explain the reasons for your answers.
202 4 Chromatographic Analysis of Polymers

(a) You have synthesized polylactide by ring-opening polymerization of lactide monomer and
wish to confirm that you have obtained a high molar mass polymer.
(b) You are testing the degradation rate of the synthesized PLA in an aqueous environment and
wish to confirm that the degradation products formed are, as expected, water-soluble lactic
acid oligomers with molar mass 90–1200 g mol–1.
(c) You are trying to develop a process for chemical recycling of PLA by thermal treatment
resulting in backbiting to original cycling lactide or large cycling oligomer with max molar
mass 350 g mol–1.
4.12. Predict and explain the order in which hexane, hexanol and hexanoic acid elutes under the
following conditions:
(a) GC analysis by mid-polar ‘universal’ column
(b) HPLC analysis in normal mode
(c) HPLC analysis in reverse mode

References

Albertsson, A.-C., & Hakkarainen, M. (Eds.). (2008). Chromatography for sustainable polymeric materials (Advances
in polymer science) (Vol. 211). Berlin/Heidelberg: Springer.
Alin, J., & Hakkarainen, M. (2010). Journal of Applied Polymer Science, 118, 1084.
Andersson, T., Wesslén, B., & Sandstr€om, J. (2002). Journal of Applied Polymer Science, 86, 1580.
Arthur, C. L., & Pawliszyn, J. (1990). Analytical Chemistry, 62, 2145.
Audourin, L., Dalle, B., Metzger, G., & Verdu, J. (1992). Journal of Applied Polymer Science, 45, 2091.
Barnes, K. A., Sinclair, R., & Watson, D. H. (Eds.). (2007). Chemical migration and food contact materials.
Cambridge: Woodhead Publisher.
Bart, J. C. J. (2006). Polymer additive analytics: Industrial practice and case studies. Firenze: Firenze University Press.
Barth, H. G., Boyes, B. E., & Jackson, C. (1998). Analytical Chemistry, 70, 251R.
Bartle, K. D., & Myers, P. (2002). TrAC Trends in Analytical Chemistry, 21, 547.
Berek, D. (2010). Journal of Separation Science, 33, 315.
Bernard, L., Decaudin, B., Lecoeur, M., Richard, D., Bourdeaux, D., Cueff, R., & Sautou, V. (2014). Talanta, 129, 39.
Bhunia, K., Sablani, S. S., Tang, J., & Rasco, B. (2013). Comprehensive Reviews in Food Science and Technology, 12,
523.
Block, C., Wynants, L., Kelchtermans, M., De Boer, R., & Compernolle, F. (2006). Polymer Degradation and Stability,
91, 3163.
Boyd, R. H. & Phillips, P. J. (1993) The science of polymer molecules. Cambridge: Cambridge University Press.
Brandrup, J., Immergut, E. H., & Grulke, E. A. (Eds.). (1999). Polymer handbook. New York: Wiley.
Buchard, W. (1999). Advances in Polymer Science, 143, 113.
Cazes, J. (Ed.). (2010). Encyclopedia of chromatography (3rd ed.). Boca Raton: CRC Press.
Cederholm, L., Xu, Y., Tagami, A., Sevastyanova, O., Odelius, K., & Hakkarainen, M. (2020). Industrial Crops and
Products, 145, 112152.
Chu, C.-C. (2015). Biodegradable Polymers (New Biomatertial advancement and challenges) (Vol. 2). Ithaca: Nove
Science Publisher.
Churms, S. C. (1996). Journal of Chromatography A, 720, 151.
Clausen, P. A., Hansen, V., Gunnarsen, L., Afshari, A., & Wolkoff, P. (2004). Environmental Science and Technology,
38, 2531.
Clausen, P. A., Xu, Y., Kofoed-Sorensen, V., Little, J. C., & Wolkoff, P. (2007). Atmospheric Environment, 41, 3217.
Einstein, A. (1906). Annalen der Physik, 19, 289.
Ekelund, M., Azhdar, B., Hedenqvist, M. S., & Gedde, U. W. (2008). Polymer Degradation and Stability, 93, 1704.
Ekelund, M., Azhdar, B., & Gedde, U. W. (2010). Polymer Degradation and Stability, 95, 1789.
Ettre, L. S., & Horvath, C. (1975). Analytical Chemistry, 47, A422.
Fanali, S., Haddad, P. R., Poole, C., & Riekkola, M.-L. (2017). Liquid chromatography, fundamentals and instrumen-
tation (2nd ed.). Amsterdam: Elsevier.
Gaborieau, M., & Castignolles, P. (2011). Analytical and Bioanalytical Chemistry, 399, 1413.
Gaborieau, M., Gilbert, R. G., Gray-Weale, A., Hernandez, J. M., & Castignolles, P. (2007). Macromolecular Theory
and Simulations, 16, 13.
References 203

Gedde, U. W., & Hedenqvist, M. S. (2019a). Conformations in polymers (chap. 2). In Fundamental polymer science.
Cham: Springer.
Gedde, U. W., & Hedenqvist, M. S. (2019b). The molten state (chap. 6). In Fundamental polymer science. Cham:
Springer.
Gedde, U. W., & Hedenqvist, M. S. (2019c). Morphology of semicrystalline polymers (chap. 7) and crystallization
kinetics (chap. 8). In Fundamental polymer science. Cham: Springer.
Gedde, U. W., & Hedenqvist, M. S. (2019d). Polymer synthesis (chap. 10). In Fundamental Polymer Science. Cham:
Springer.
Ginzburg, A., Macko, T., Dolle, V., & Brüll, R. (2011). European Polymer Journal, 47, 319.
Graessley, W. W. (1983). Viscoelasticity and flow in polymer melts and concentrated solutions. In J. E. Mark (Ed.),
Physical properties of polymers. Washington, DC: American Chemical Society.
Graessley, W. W. (2004). Polymeric liquids and networks: Structure and properties. London: Garland Science.
Gr€oning, M., & Hakkarainen, M. (2004). Journal of Chromatography A, 1052, 61.
Gr€oning, M., & Hakkarainen, M. (2008). Advances in Polymer Science, 211, 51.
Grubisic, Z., Rempp, P., & Benoit, H. (1967). Journal of Polymer Science: Polymer Letters, 5, 753.
Guiochon, G., & Guillemin, C. L. (1988). Quantitative gas chromatography for laboratory analyses and on-line
process control, journal of chromatography library – Volume 42. Amsterdam: Elsevier Science Publishers.
Hachenberg, H., & Schmidt, A. P. (1985). Gas chromatographic headspace analysis. New York: Wiley.
Hakkarainen, M. (2007). Journal of Biochemical and Biophysical Methods, 70, 229.
Hakkarainen, M. (2008a). Advances in Polymer Science, 211, 23.
Hakkarainen, M. (2008b). Advances in Polymer Science, 211, 159.
Hakkarainen, M., & Karlsson, S. (2000). Encyclopedia of analytical chemistry, chapter 9. Chichester: Wiley.
Hakkarainen, M., Albertsson, A.-C., & Karlsson, S. (1997). Journal of Environmental Polymer Degradation, 5, 67.
Hakkarainen, M., Karlsson, S., & Albertsson, A.-C. (2000). Polymer, 41, 2331.
Hakkarainen, M., H€ oglund, A., Odelius, K., & Albertsson, A.-C. (2007). Journal of the American Chemical Society,
129, 6308.
Hiller, W., Hehn, M., Hofe, T., & Oleschko, K. (2010). Analytical Chemistry, 82, 8244.
Hiller, W., Sinha, P., Hehn, M., & Pasch, H. (2014). Progress in Polymer Science, 29, 979.
H€oglund, A., Hakkarainen, M., & Albertsson, A.-C. (2010). Biomacromolecules, 11, 277.
Horodytska, O., Cabanes, A., & Fullana, A. (2020). Chemosphere, 251, 126373.
Horváth, C. (Ed.). (1988). High performance liquid chromatography: Advances and perspectives (Vol. 5). San Diego:
Academic Press.
Jakubowicz, I., Yarahmadi, N., & Gevert, T. (1999). Polymer Degradation and Stability, 66, 415.
James, A. T., & Martin, A. J. P. (1952). Biochemical Journal, 50, 679.
Jenkins, M. (2007). Biomedical polymers. Cambridge: Woodhead Publisher.
Jennings, W., Mittlefehldt, E., & Stremple, P. (1997). Analytical gas chromatography (2nd ed.). San Diego: Academic
Press.
Jones, R. G., Kahovec, J., Stepto, R., Wilks, E. S., Hess. M., & Kitayama, T. (2009). Compendium of polymer
terminology and nomenclature – IUPAC recommendations 2008. International Union of Pure and Applied Chemis-
try. Published by The Royal Society of Chemistry, Cambridge.
Jovic, K., Nitsche, T., Lang, C., Blinco, J. P., De Bruycker, K., & Barner-Kowollik, C. (2019). Polymer Chemistry, 10,
3241.
Karlsson, K., Smith, G. D., & Gedde, U. W. (1992). Polymer Engineering and Science, 32, 649.
Kausch, H. H. (1978). Polymer fracture. Berlin/Heidelberg/New York: Springer.
Kolb, B., & Ettre, L. S. (1997). Static headspace-gas chromatography: Theory and practice. New York: Wiley.
Kolb, B., & Pospisil, P. (1977). Chromatographia, 10, 705.
Konstanski, L. K., Keller, D. M., & Hamielec, A. E. (2004). Journal of Biochemical and Biophysical Methods, 58, 159.
Kramers, H. A. (1946). Journal of Chemical Physics, 14, 415.
La Mantia, F. P., & Acierno, D. (1985). Polymer Engineering and Science, 25, 279.
Labarre, D., Ponchel, G., & Vauther, C. (2010). Biomedical and pharmaceutical polymers. London: Pharmaceutical
Press.
Lin, Y.-H. (1990). Macromolecules, 23, 5292.
Liu, P., Liu, W., Wang, W.-J., Li, B.-G., & Zhu, S. (2017). Macromolecular Reaction Engineering, 11, 1600012.
Mansoor Shaikh, M., Al Suhaimi, A. O., Hanafiah, M. M., Aqeel Ashraf, M., & Norliyana Harun, S. (2018).
Desalination and Water Treatment, 112, 344.
McMaster, M. C. (2005). LC/MS: A practical User’s guide. Hoboken, NJ: Wiley.
McMaster, M. C. (2011). GC/MS: A practical User’s guide (2nd ed.). New York: Wiley-VCH.
Moldoveanu, S., & David, V. (2015). Modern sample preparation for chromatography. Amsterdam: Elsevier.
Montaudo, G., Samperi, F., & Montaudo, M. (2006). Progress in Polymer Science, 31, 277.
Moore, J. C. (1964). Journal of polymer science. Polymer Chemistry, 2, 835.
Mori, S., & Barth, H. G. (1999). Size exclusion chromatography. Berlin/Heidelberg/New York: Springer.
204 4 Chromatographic Analysis of Polymers

Mourey, T. H., Turner, S. R., Rubinstein, M., Frechet, J. M. J., Hawker, C. J., & Wooley, K. L. (1992). Macromolecules,
25, 2401.
Moyassari, A., Gkourmpis, T., Hedenqvist, M. S., & Gedde, U. W. (2019a). Polymer, 161, 139.
Moyassari, A., Gkourmpis, T., Hedenqvist, M. S., & Gedde, U. W. (2019b). Macromolecules, 52, 807.
Murgasova, R., & Hercules, D. M. (2003). International Journal of Mass Spectrometry, 226, 151.
Musl, O., Sulaeva, I., Bacher, M., Mahler, A. K., Rosenau, T., & Potthast, A. (2020). ChemSusChem, 13, 1.
Nahan, K., Sussman, E. M., Oktem, B., Schultheis, L., & Wickramasekara, S. (2020). Talanta, 212, 120464.
Nerin, C., Salafranca, J., Aznar, M., & Batlle, R. (2009). Analytical and Bioanalytical Chemistry, 393, 809.
Nerin, C., Alfaro, P., Aznar, M., & Domeno, C. (2013). Analytica Chimica Acta, 775, 14.
Orski, S. V., Kassekert, L. A., Farrell, W. S., Kenlaw, G. A., Hillmyer, M. A., & Beers, K. L. (2020). Macromolecules,
53, 2344.
Pasch, H. (2013). Polymer Chemistry, 4, 2628.
Pasch, H., & Trathnigg, B. (2013). Multidimensional HPLC of polymers (Springer laboratory manuals in polymer
science). Berlin/Heidelberg: Springer.
Pawliszyn, J. (1997). Solid phase microextraction: Theory and practice. New York: Wiley-VCH.
Pawliszyn, J. (Ed.). (2012). Handbook of solid phase microextraction. Amsterdam: Elsevier.
Perale, G., & Hilborn, J. (2016). Bioresorbable polymers for biomedical applications. Cambridge: Woodhead
Publisher.
Pirok, B. W. J., Gargano, A. F. G., & Schoenmakers, P. J. (2018). Journal of Separation Science, 41, 68.
Plüschke, L., Ndiripo, A., Mundil, R., Merna, J., Pasch, H., & Lederer, A. (2020). Macromolecules, 53, 3765.
Poole, C. F. (2003). The essence of chromatography. Amsterdam: Elsevier Science.
Poole, C. F. (2012). Gas chromatography. Amsterdam: Elsevier.
Poole, C. F. (Ed.). (2020). Liquid-phase extraction handbooks in separation science. Amsterdam: Elsevier.
Porath, J., & Flodin, P. (1959). Nature, 183, 1657.
Raust, J. A., Brüll, A., Moire, C., Farcet, C., & Pasch, H. (2008). Journal of Chromatography A, 1203, 207.
Rodriguez Bernaldo de Quiros, A., Lestido Cardama, A., Sendon, R., & Garcia Ibarra, V. (2019). Food contamination
by packaging: Migration of chemicals from food contact materials. Berlin/Boston: De Gruyter.
Rubinstein, M., & Colby, R. H. (2003). Polymer physics. Oxford: Oxford University Press.
Scholte, T. G. (1983). Characterization of long-chain branching in polymers. In J. V. Dawkins (Ed.), Developments in
polymer characterisation – 4. London/New York: Applied Science Publishers.
Silva, A. S., Garcia, R. S., Cooper, I., Franz, R., & Losada, P. P. (2006). Trends in Food Science and Technology, 17,
535.
Simal-Gandara, J., Damant, A. P., & Castle, L. (2002). Critical Reviews in Analytical Chemistry, 32, 47.
Skjevrak, I., Due, A., Gjerstad, K. O., & Herikstad, H. (2003). Water Research, 37, 1912.
Smith, G. D., Karlsson, G. D., & Gedde, U. W. (1992). Polymer Engineering and Science, 32, 658.
Stadler, F. J., Piel, C., Klimke, K., Kaschta, J., Parkinson, M., Wilhelm, M., Kaminsky, W., & Munstedt, H. (2006).
Macromolecules, 39, 1474.
Steinkoenig, J., Rothfuss, H., Lauer, A., Tuten, B. T., & Barner-Kowollik, C. (2017). Journal of the American Chemical
Society, 139, 51.
Strangl, M., Ortner, E., Fell, T., Ginzinger, T., & Buettner, A. (2020). Journal of Cleaner Production, 260, 121104.
Striegel, A. M., Kirkland, J. J., Yau, W. W., & Bly, D. D. (2009). Modern size-exclusion liquid chromatography-
practice of gel permeation and gel filtration chromatography. Hoboken: Wiley.
Tackx, P., & Tackx, J. C. J. F. (1998). Polymer, 39, 3109.
Uliyanchenko, E. (2017). Chromatographia, 80, 731.
Uliyanchenko, E., van der Wal, S., & Schoenmakers, P. J. (2012). Polymer Chemistry, 3, 2313.
Van der Veen, I., & de Boer, J. (2012). Chemosphere, 88, 1119.
Viebke, J., & Gedde, U. W. (1998). Polymer Engineering and Science, 38, 1244.
Viebke, J., Elble, E., Ifwarson, M., & Gedde, U. W. (1994). Polymer Engineering and Science, 34, 1354.
Viebke, J., Hedenqvist, M., & Gedde, U. W. (1996a). Polymer Engineering and Science, 36, 2896.
Viebke, J., Elble, E., & Gedde, U. W. (1996b). Polymer Engineering and Science, 36, 458.
Viktor, Z., & Pasch, H. (2020). Analytica Chimica Acta, 1107, 225.
Vitha, M. F. (2017). Chromatography: Principles and instrumentation. Hoboken: Wiley.
Waters Corporation. (2014). Beginner’s guide to SPE: Solid-phase extraction. New York: Wiley.
Wellings, D. A. (2005). A practical handbook of preparative HPLC. Amsterdam: Elsevier.
Wetton, A. J., & Nguyen, T. (2013). Critical Reviews in Environmental Science and Technology, 43, 679.
Winkler, M., Montero de Espinosa, L., Barner-Kowollik, C., & Meier, M. A. R. (2012). Chemical. Sciences, 3, 2607.
Wyatt, P. J. (1993). Analytica Chimica Acta, 272, 1.
Zimm, B., & Stockmayer, W. (1949). Journal of Chemical Physics, 17, 1301.
Chapter 5
Simulation and Modelling of Polymers

5.1 Introduction

Theory and experiment are historically the two main tools of material science, but during the last few
decades, computer simulation has emerged as an increasingly important complement. In polymer
science, simulations can be used to develop polymeric materials with improved properties, to
optimize the geometries of macroscopic constructions, to study polymeric materials under experi-
mentally inaccessible conditions, to explain experimentally observed phenomena and to reduce the
number of required experiments. Many simulation techniques exist, and the choice of simulation
strategy depends on the characteristic time and length scales of the computational problem. Some
phenomena are preferably simulated with atomistic simulation techniques, whereas others are better
modelled with macroscopic methods. Multiscale modelling combines simulation methods on differ-
ent time and length scales. The aim of this chapter is to provide a brief overview of the simulation
techniques used in material science.
Atomistic material problems where the influence of electrons must be explicitly included can be
handled with quantum chemistry (QC) techniques such as molecular orbital (MO) strategies, which
include ab initio or semi-empirical MO methods, or density functional theory (DFT). All QC methods
solve simplified versions of the Schr€odinger equation numerically. The applications of QC techniques
include the prediction of electron spectra, electron structures, electron transport, chemical reactions
and force fields. The computational cost of explicitly including the effect of individual electrons in an
atomistic simulation is usually very high, and this considerably limits the number of atoms that can be
included in QC simulations. Examples of QC computer programs include Gaussian, MOPAC,
ONETEP, DMol3, CASTEP, Dalton, DIRAC, GAMESS, DFTB+ and MOLCAS.
Atomistic material problems in which the influence of individual electrons can be neglected can be
handled with molecular modelling (MM) techniques, either molecular dynamics (MD) or Monte Carlo
(MC) methods. In MD, Newton’s equation of motion is solved as functions of time, whereas in MC
small random movements are introduced until the molecular system has reached equilibrium. All MM
techniques use atom position-dependent force fields to describe the potential energy of a molecular
structure. Force fields can be determined experimentally or from QC simulations. Some force fields use
coarse-graining, meaning that small groups of nearby atoms are grouped into larger ‘superatoms’.
Coarse-graining leads to fewer atoms and thus increases the speed of the simulation, but the accuracy of
the solution can be somewhat reduced. MM can, for instance, be used to simulate small-molecule
diffusion in polymeric structures or elastic deformations due to external stresses. Examples of MM
simulation programs include Avogadro, CHARMM, GROMACS, LAMMPS, HyperChem,
MacroModel, Chimera, Abalone, Ascalaph Designer, Materials Studio, NAMD/VMD and EMC.

# Springer Nature Switzerland AG 2021 205

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_5
206 5 Simulation and Modelling of Polymers

Material models that are too coarse to include individual atoms (or coarse-grained ‘atoms’) but too
detailed to treat the material as a continuum are called ‘mesoscale models’. Such methods can, for
instance, be used to model molten polymer mixtures, large crystal structures and composites.
Examples of mesoscale techniques are dissipative particle dynamics (DPD), Brownian dynamics
(BD), Lattice Boltzmann (LB) and effective media models, and examples of mesoscale programs
include Espresso, Gromacs-DPD, LAMMPS and DL-meso.
Computational problems where the materials can be treated as homogeneous continua rather than
mixtures of atoms, molecules, nanoparticles and other small sub-units can be handled with continuum
simulation methods, such as finite element methods (FEMs), finite difference methods (FDM), finite
volume methods (FVM) and spectral methods. Continuum methods are typically used to solve partial
differential equations, like the Navier–Stokes equations, the heat equation, the diffusion equation or
the Poisson equation, on complex macroscopic geometries. Multiphysics FEM modelling can be used
to solve several coupled differential equations simultaneously. A wide variety of continuum level
programs exist, for instance, Abaqus, Autodesk, COMSOL Multiphysics, Ansys and Nastran.
Statistical methods, such as the quantitative structure–activity relationship (QSAR) and group-
contribution (GC) methods, can also be used to relate macroscopic properties to microscopic
structural data or to other macroscopic properties. Such methods are often expressed by relatively
simple analytical equations, with parameters determined by a statistical analysis of data from material
databases, compiled from experimental or simulation data. Techniques like data mining, machine
learning and artificial neural networks can improve the performance of statistical methods.
Other material modelling methods exist, including lattice fluid methods and image analysis
techniques, but the techniques here presented are currently the most widely used methods in polymer
science. A summary of suitable time and length scales for some of the most important polymer
modelling techniques is given in Fig. 5.1.

Fig. 5.1 Simulation methods with different time and length scales. The MD figure is from Moyassari et al 2017 with
permission from AIP Publishing, the Meso-scale figure from Nilsson et al 2016 with permission from Elsevier and the
FEM figure from Nakamura et al 2018 with permission from Elsevier
5.2 Quantum Chemistry (QC) 207

5.2 Quantum Chemistry (QC)

5.2.1 QC: Overview

Quantum chemistry (QC) methods are atomistic simulation methods which include electron effects
(Cramer 2004; Leach 2001; Szabo and Ostlund 1996). They can be used to calculate, for instance,
dipole moments, electrostatic potentials, spectra, optimized geometries, bond strengths and (some-
times) even the outcome of chemical reactions including bond breaking and bond formation. All
quantum methods seek to find approximate, numerical solutions to simplified versions of the
Schr€odinger equation. Quantum chemistry can be divided into two major classes of methods:
molecular orbital (MO) methods and density functional theory (DFT) methods. The former optimize
wave functions, whereas the latter optimize electron densities, but the classes are closely related.
Some QC methods, like ab initio MO methods, introduce very few simplifications to the Schr€ odinger
equation and are therefore accurate but computationally expensive. Other techniques, like semi-
empirical MO or linear-scaling DFT, are computationally cheaper due to the incorporation of
experimental data or because of mathematical simplifications. Compared to molecular dynamics
(MD) and other molecular mechanic methods, all QC methods are however still computationally
expensive. Some hybrid QC–MD techniques have been developed to combine the best features of
MD (speed, time dependence) and QC (accuracy, chemical reactions), but extensive research is still
ongoing in this field.

5.2.2 Computational Quantum Mechanics: Formalism

The most fundamental equation in quantum mechanics is the time-dependent Schr€

odinger equation
(Schr€odinger 1926):

b ðr, tÞ ¼ iℏ ∂Φðr, tÞ
HΦ ð5:1Þ
∂t
b it the Hamilton operator and Φ is a wave function which depends on position r and time t.
where H
∂Φ/∂t is the partial derivative of Φ with respect to t. The Hamilton operator is given by:
 2 
b ℏ 2 b
H¼ ∇ þ V H ðr, tÞ ð5:2Þ
2m

where the first term is the kinetic energy operator (with mass m) and the second term (i.e. VH(r,t)) is
the potential energy operator. The nabla operator (∇) symbolizes partial derivatives with respect to
position, such that ∇ ¼ ∂/∂x and ∇ 2 ¼ ∂2/∂x2 in 1D Euclidean geometry. In 3D Euclidean
geometry, the squared nabla operator becomes:
 2 2 2
2 ∂ ∂ ∂
∇ ¼ þ þ ð5:3Þ
∂x2 ∂y2 ∂z2

If the potential energy is approximately time-independent, i.e. VH(r) rather than VH(r,t), then the
time-dependent Schr€odinger equation can be decoupled into a time-dependent and a time-
independent part. If Φ is defined as the product of its spatial- and time-dependent factors Ψ(r) and
Θ(t),
208 5 Simulation and Modelling of Polymers

Φðr, tÞ ¼ Ψðr ÞΘðtÞ ð5:4Þ

The Schr€odinger equation can then be solved as a separable partial differential equation. A
constant E which is the energy of the system is first defined, and the time-dependent and time-
independent parts of the separable Schr€odinger equation are derived by adding (E-E)Φ ¼ 0 to Eq. 5.1,
dividing by Φ and separating the r-dependent and t-dependent factors into different equations:

b ðr ÞΘðtÞ ¼ iℏ ∂Ψðr ÞΘðtÞ )

HΨ ð5:5Þ
∂t

b ðr Þ ¼ Ψðr Þiℏ ∂ΘðtÞ

ΘðtÞHΨ þ ðE  EÞΨðr ÞΘðtÞ ) ð5:6Þ
∂t

∂ΘðtÞ
iℏ ¼ EΘðtÞ ð5:7Þ
∂t

b ðr Þ ¼ EΨðr Þ
HΨ ð5:8Þ

The time-dependent part of the separable Schr€odinger equation (Eq. 5.7) is easily solved. With
starting time t0, the solution is:
 
iEðt  t0 Þ
ΘðtÞ ¼ 1 exp  ð5:9Þ
ℏ

The time-independent Schr€odinger equation (Eq. 5.8) is more difficult to solve, and most quantum
chemistry models primarily address this part of the Schr€ odinger equation. For simplicity, relativistic
effects are almost always neglected. The rest of this chapter is therefore devoted to techniques for
finding approximate solutions to the non-relativistic, time-independent Schr€ odinger equation:
b ðr Þ ¼ EΨðr Þ
HΨ ð5:8Þ

In QC calculations, atomic units are normally used instead of SI units, in order to simplify the
calculations and reduce the number of constants. When atomic units are used, the mass unit is
me ¼ 9.109∙1031 kg, the length unit is a0 ¼ 5.292∙1011 m, the charge unit is e ¼ 1.602∙1019 C,
and the energy unit is Eh ¼ 4.360∙1018 J. In SI units, the Hamilton operator, also known as the
Hamiltonian, of an atom with a single electron is:
 
b ℏ 2 ∇2 ℏ 2 ∇2 e2 Z
H¼    ð5:10Þ
m0 2 MA 2 4πε0 r

where Z is the atomic number, e the electron charge, me the electron mass, MA the nuclei mass and
r the distance between the electron and the centre of the nucleus. In atomic units, the expression can
be simplified to:
 
b ∇2 ∇2 Z
H¼    ð5:11Þ
2 2MA r

Typical molecular systems containing several atoms and several electrons have five energy
contributions to the Hamiltonian:
5.2 Quantum Chemistry (QC) 209

Xn
∇2i X
Nc
∇2A X n X Nc
ZA X X 1 X
n n Nc XNc
ZA ZB
b¼
H   þ þ ð5:12Þ
i¼1
2 A¼1
2M A i¼1 A¼1
r iA i¼1 j>i
r ij A¼1 B>A
RAB

where the indexes i and j are for the n electrons and the indexes A and B are for the Nc nuclei. MA and
MB are the masses of nuclei with indexes A and B. The distances between electrons, between nuclei
and between electron and nucleus are denoted rij, RAB and riA, respectively. The five terms on the
right-hand side of Eq. 5.12 correspond, respectively, to the kinetic energy of the electrons, the kinetic
energy of the nuclei, the potential energy of electron–nuclei attractions, the potential energy of
interelectronic repulsions and the potential energy of internuclear repulsions.
In more complex situations, for instance, when external magnetic or electrical fields are present or
when relativistic effects arise, additional terms are added to the Hamilton operator, but for most
molecular systems, these terms are sufficient. In most MQ simulations, it is even possible to ignore
two of the terms: the kinetic energy of the electrons and the potential energy of the internuclei
repulsions. This important simplification is called the Born –Oppenheimer approximation (Born and
Oppenheimer 1927) and is motivated by the fact that the speed of the electrons is much greater than
the speed of the nuclei. The remaining Hamiltonian becomes:
Xn
∇2i X n X Nc
ZA X X 1
n n
b¼
H  þ ð5:13Þ
i¼1
2 r
i¼1 A¼1 iA
r
i¼1 j>i ij

and the Born–Oppenheimer version of the Schr€ odinger equation, where most of the nuclei
contributions are neglected, is called the electronic Schr€
odinger equation:
bel Ψel ¼ Eel Ψel
H ð5:14Þ

The subscript el, indicating that this is a property of the electronic Schr€
odinger equation, is usually
omitted. The total energy of the time-independent Schr€ odinger equation equals the electronic energy,
computed from Eq. 5.14, plus the nuclear energy:
XNc XNc
ZA ZB
Etot ¼ Eel þ ð5:15Þ
A¼1 B>A
RAB

Some mathematical definitions are required for the computations in this chapter. Therefore,
consider a general Schr€odinger-type equation of the form:
b ¼ Eψ
Hψ ð5:16Þ

The left-hand side of the equation, where the Hamilton operator acts on the function ψ, must equal
the function ψ multiplied by the energy E. All the functions ψ that can fulfil this criterion are
eigenfunctions of H, whereas the corresponding prefactors E are the eigenvalues of these solutions. In
general, many different ψ will fulfil the Schr€odinger equation, but some of the solutions result in
lower energies than others. The product ψ*ψ ¼ |ψ 2| is called the probability density of ψ, and this
product is the probability of finding the particle within an infinitely small volume element dV. The
symbol “*” is the complex conjugate. All functions ψ i should be normalized, which means that the
integral of their probability density is unity when integrated over an infinitely large volume V1:
210 5 Simulation and Modelling of Polymers

ð
ψ i ∗ ψ i dV ¼ 1 ð5:17Þ
V1

Two vectors ψ i and ψ j are orthogonal (perpendicular) to each other if, when i 6¼ j, they fulfil the
equation:
ð
ψ i ∗ ψ j dV ¼ 0 ð5:18Þ
V1

Two functions ψ i and ψ j are orthonormal if they are both orthogonal and normalized. Using
Kronecker’s delta δij, where δij ¼ 1 if i ¼ j and δij ¼ 0 otherwise, this can be formulated:
ð
ψ∗i ψ j dV ¼ δij ð5:19Þ
V1

The expectation value of any operator, in this case xb, is defined as:
ð
< xb >¼ ψ ∗ xbψdV ð5:20Þ
V1

If ψ i and ψ j are orthonormal, the expectation value of the Hamilton operator of the time-
independent electronic Schr€odinger equation becomes:
D E ð ð ð
Hb ¼ H ij ¼ ψ ∗i
b j dV ¼ ψ ∗ E j ψ j dV ¼ E j ψ ∗ ψ j dV ¼ E j δij
Hψ i i ð5:21Þ
V V V

The expectation value of the Hamilton operator is thus Ei if j ¼ i, and otherwise zero. Ei is an
energy and H an energy operator. Only for the simplest molecular systems is it possible to derive the
exact analytical solutions ψ exact and Eexact to the Schr€
odinger equation, but it is fortunately possible to
obtain reasonably good approximations.

5.2.3 Molecular Orbital (MO) Methods, One-Electron Technique

In the case of molecular structures, one strategy for finding approximate solutions to the Schr€
odinger
equation is to construct molecular orbital (MO) wave functions as weighted sums of atomistic orbitals
(AO) and to minimize the energies with respect to the weights of the molecular orbitals (Lennard-
Jones 1929). Methods that optimize molecular orbitals are classified as molecular orbital
(MO) techniques, including the ab initio MO and semi-empirical MO methods.
Early MO methods (one-electron techniques) assumed that the contributions of the electrons could
be calculated independently and then added together. The accuracy was however significantly
improved when electron–electron interactions (Coulombic repulsions), electron spin and electron
anti-symmetry were also explicitly taken into account, which was done in later MO methods (many-
electron techniques). Further improvements have been introduced in more recent MO methods. To
reduce the mathematical complexity and increase the clarity of the presentation, the one-electron MO
technique is first presented here.
5.2 Quantum Chemistry (QC) 211

Assume that a set of N-independent atomistic one-electron orbitals ψ j is available and that you want to
construct a corresponding set of N molecular orbitals Ψ i for the electrons in a molecule. For each
electron, an approximative molecular orbital Ψ i, which is a one-electron wave function, can be
constructed as a linear combination of atomic orbitals (LCAO), with adjustable (scalar) coefficients cj:
X
N
Ψi ¼ c jψ j ð5:22Þ
j¼1

The total molecular wave function is computed as a Hartree product Ψ Total ¼ Ψ 1Ψ 2Ψ 3. . .Ψ N when
electron–electron interactions are neglected, otherwise as an anti-symmetric linear combination of
MO products.
According to the variational principle, the expectation value of the Hamiltonian, using an
approximative wave function Ψ , is always greater than or equal to the exact energy value Eexact,
which is the lowest energy eigenvalue of the system:
Ð ∗
b
Ψ HΨdV
 Ee  Eexact
V
Ð ∗ ð5:23Þ
Ψ ΨdV
V

Thus, even if the exact energy Eexact cannot be assessed, reasonably good approximations can be
achieved if Ψ is well chosen. Better approximations of the wave function give lower values of the
computed energy. The minimum energy is obtained by finding the adjustable coefficients ci that
minimize the energy E, e either by using Lagrange multiplier technique or by differentiating Ee with
respect to ci and setting the derivative to zero:

∂Ee
¼ 0, i¼1:N ð5:24Þ
∂ci

Before this derivative is calculated, the resonance integrals Hij and the overlap integrals Sij are
introduced and computed, assuming that the approximate one-electron orbitals ψ j are known:
ð
b j dV
Hij  ψ i Hψ ð5:25Þ
V

ð
Sij  ψ i ψ j dV ð5:26Þ
V

Insertion of the LCAO approximation (Eq. 5.22) into Eq. 5.23 gives:
! !
Ð ∗ Ð P P
b
Ψ HΨdV ci ψ i Hb c j ψ j dV
V V i j
Ð ∗ ¼ ! ! ¼
Ψ ΨdV Ð P P
V ci ψ i c j ψ j dV
V i j ð5:27Þ
P Ð P
b j dV
ci c j ψ i Hψ ci c j Hij
¼ Ee
ij V ij
¼ P Ð ¼ P
ci c j ψ i ψ j dV ci c j Sij
ij V ij

which can be rewritten:

212 5 Simulation and Modelling of Polymers

X X
ci c j H ij  Ee ci c j Sij ¼ 0 ð5:28Þ
ij ij

Differentiation with respect to the weights ci gives N equations:

N 
X 
e ij c j ¼ 0,
H ij  ES i¼1:N ð5:29Þ
j¼1

Non-trivial solutions to such a system of equations require that the determinant is zero, resulting in
the secular equation:
 
e 11 H 22  ES
 H11  ES e 22 . . . H 1N  ES e 1N 
 
 e 21 H 22  ESe 22    H 2N  ES e 2N 
 H21  ES
 ¼0 ð5:30Þ
 .. 
 ⋮ ⋮ . ⋮ 
 
 e n1 H N2  ESe N2    H NN  ES e NN 
H N1  ES

In principle, the secular determinant can be written as a N’th order polynomial with polynomial
roots Ei corresponding to energy levels of different electrons. Each root gives rise to a separate set of
N weights. If several energies have the same value, these are degenerated energies. In practice, the
secular equation, which is an eigenvalue problem, is however solved with numerical linear algebra.
The resulting eigenvalues are the energies Ei, and the eigenvectors are the weights cij. Equation 5.30
can then be written as a matrix equation:
HC ¼ SCE ð5:31Þ

where H is a known symmetric resonance integral matrix, S is a known symmetric overlap integral
matrix, C is an unknown eigenvector matrix with components cij and E is an unknown (diagonal)
energy matrix with eigenvalues Ei on the diagonal. All four matrices have size N*N. Such eigenvalue
problems can be solved numerically with standard methods. For instance, in MATLAB, the compu-
tation is written: [E,C] ¼ eig(H,S). If the basis functions ψ are orthonormal, S becomes an identity
matrix I, and the matrix equation is simplified to:
HC ¼ CE ð5:32Þ

This simplified equation can also be solved with numerical methods for eigenvalue problems, for
instance, with the command [E,C] ¼ eig(H) in MATLAB. When the weight matrix C with
components cij has been computed, the optimized LCAO wave function for orbital i is obtained:
X
N
Ψi ¼ cij ψ j ð5:33Þ
j¼1

The basic one-electron MO–LCAO technique is summarized by Eqs. 5.21, 5.22, 5.23, 5.24, 5.25,
5.26, 5.27, 5.28, 5.29, 5.30, 5.31, 5.32 and 5.33. If the basis functions ψ are cleverly chosen and the
matrices Hij and Sij are accurately computed, this technique gives reasonably good approximate
solutions to the Schr€odinger equation.
The Hückel model (Hückel 1931) is an early example of the one-electron MO–LCAO strategy.
This technique was primarily developed to study electronic energies and charges in hydrocarbons.
5.2 Quantum Chemistry (QC) 213

Only π electrons from sp2-hybridized atoms are included in the Hückel calculations, which means
that adjacent ψ are orthonormal and that the overlap matrix S is a diagonal unit matrix, i.e. Sij ¼ δij.
The ionization energy of a methyl radical and the stabilization energy of a half π-bond are denoted α
and β, respectively. In the resonance integral, all diagonal terms are Hii ¼ α. The non-diagonal terms
are Hij ¼ β if i and j are adjacent π electrons; otherwise Hij ¼ 0. For linear hydrocarbons, the secular
determinant is a symmetric tridiagonal N*N matrix, where N is the number of π electrons in the
molecule:
 
α  E β ... 0 
 
 β α  E  0 

 ¼0 ð5:34Þ
 .. 
 ⋮ ⋮ . ⋮ 
 
 0 0  α  E

If the hydrocarbon were to contain branches or rings, the secular determinant would still be
symmetric, but no longer tridiagonal, since all zeros that correspond to adjacent π electrons would be
replaced with β. The energies Ei are the eigenvalues of the secular determinant, and since the
electrons are considered to be independent, the total energy can be computed as the sum of all Ei.
Even though the Hückel theory includes many simplifications, it can still be used to compute energies
for chemical reactions, to explain why aromatic hydrocarbons have a lower energy than the
corresponding linear hydrocarbons and to predict bond order and charges for different atoms in
hydrocarbon molecules. Its relative simplicity also makes it useful for educational purposes. For
practical applications, more advanced methods are however needed.
The extended Hückel method, which gave the inventor Hoffman the Nobel Prize in 1963
(Hoffmann 1963), included the effect of σ-orbitals in addition to that of the π-orbitals. In contrast
to the original Hückel method, the non-diagonal overlap matrix elements Sij are computed from
Eq. 5.26 rather than being set to zero. The components of the secular determinant are:
ð
b i dV
H ii ¼ ψ i Hψ ð5:35Þ
V

 
1:75 Hii þ H jj Sij
H ij ¼ ð5:36Þ
2

The extended Hückel method is not sufficiently accurate to compete with modern state-of-the-art
quantum models, but it is often used to generate a starting guess in more accurate, iterative
algorithms.

5.2.4 Molecular Orbital (MO) Methods, Many-Electron Technique

The main fault with the Hückel method is that it neglects (Coulombic) electron–electron repulsions.
More accurate quantum calculations can thus be obtained by including interactions between electrons
in the MO calculations. Two effects that then must be explicitly considered are the effects of electron
spin and the Pauli principle of electron anti-symmetry. (When the Coulomb repulsion is neglected, the
only effect of electron spin and the anti-symmetry requirement is that a maximum of two electrons is
allowed in the same orbital. This requirement is enforced in the Hückel method.)
214 5 Simulation and Modelling of Polymers

Electrons are characterized not only by position but also by spin, which can be either up (+1) or
down (1). In so-called closed shell systems, two electrons reside in each spatial electron orbital Ψ i,
one with up-spin and one with down-spin. As described in Eq. 5.33, the spatial electron orbitals can be
expanded into a linear combination of presumably known basis functions ψ. Each spin orbital χj is
defined as:
Ψi ðx, y, zÞϖðþÞ or
χj ¼ ð5:37Þ
Ψi ðx, y, zÞϖðÞ

where ϖ(+) and ϖ() indicate up- and down-spin, respectively. The jth spin orbital occupied by the
ith electron is denoted χ j(i) and number of electrons, as well as of spin orbitals, is n. If the spatial
orbitals are orthonormal, the spin orbitals are also orthonormal.
The Pauli principle states that the wave function for a system of electrons must be anti-symmetric
with respect to the interchange of any two electrons. Hartree (1928) initially wrote the total wave
function as a product of molecular orbitals, i.e. Ψ total ¼ χ 1χ 2χ 3. . .χ n, but that expression only holds
true when electron–electron interactions can be neglected. Fock (1930) and Slater (1929) indepen-
dently introduced anti-symmetric wave functions, which were anti-symmetric linear combinations of
spin orbitals. The use of these wave functions transformed Hartree’s method to the Hartree–Fock
method. The new wave functions fulfilled the anti-symmetry condition and were written as Slater
determinants (Slater 1929):
 
 χ 1 ð 1Þ χ 2 ð 1Þ . . . χ n ð nÞ 
 
 χ ð 2Þ χ ð 2Þ    χ ð nÞ 
 1 2 n 
ΨSlater ð1, 2 . . ., nÞ ¼  ..

 ð5:38Þ
 ⋮ ⋮ . ⋮ 
 
 χ ð nÞ χ ð nÞ    χ ð nÞ 
1 2 n

The Hartree–Fock self-consistent field method (HF–SCF) (Fock 1930; Hartree 1928) is an itera-
tive LCAO technique which is the basis of many molecular orbital quantum chemistry models. The
Hartree–Fock method ultimately includes interactions between all n electrons, but in each iteration,
the n-electron system is approximated by n one-electron systems. Electron spin effects are included
by using Slater determinants. If the variational principle is utilized, the energy can be minimized with
respect to the spin orbitals, giving the one-electron Fock operator for electron i ¼ 1 (Hartree 1928):
n  
∇2 X ZA X b
Nc
Hb1 ¼  1 
F
þ J j ð1Þ  Kb j ð1Þ ð5:39Þ
2 r
A¼1 1A j¼1

ð  2
ψ j ð 2Þ 
Jbj ð1Þ ¼ dr2 ð5:40Þ
r 12
ð 
ψ j ð2Þψ i ð2Þ
Kb j ð1Þψ i ð1Þ ¼ dr2 ψ j ð1Þ ð5:41Þ
r 12

The Fock operator is an approximation of the Hamilton operator, whereas Jbj ð1Þ and Kb j ð1Þ are the
Coulomb and exchange operators, respectively. The Coulomb operator describes the average poten-
tial acting on an electron at position r1 from an electron in orbital j, whereas the exchange orbital
ensures that the wave function becomes anti-symmetric.
The Hartree–Fock equation is a Schr€odinger-type equation where the ordinary Hamilton operator
is replaced by the Fock operator. For each electron i, the equation can be written:
5.2 Quantum Chemistry (QC) 215

bF χ i ¼ Ei χ i
H ð5:42Þ
i

The Hartree–Fock simplifications give energies slightly higher than those given by the true
Schr€odinger equation. By analogy with the one-electron Hückel theory, the analytical Hartree–Fock
equation can be transformed with the MO technique into a matrix equation like Eqs. 5.30 and 5.31
(i.e. HC ¼ SCE), by assuming a closed shell system and Slater-determinant wave functions. This
matrix equation is called the Roothaan–Hall equation (Hall 1951; Roothaan 1951). If the number of
electrons is n and the number of basis functions is N ¼ n/2, the components Sij of the N*N overlap
matrix S are computed as:
ð
Sij ¼ ψ i ∗ ψ j dV ð5:43Þ
V

After some simplifications, the components Hij of the one-electron Hartree–Fock approximation of
the N*N Hamiltonian matrix, termed the Fock matrix (Fock 1930), become:
ð !
∇2 X
Nc
ZA
Hij ¼ ψ i   ψ j dVþ
2 r
A¼1 iA
V
0Ð Ð 1
ψ i1 ∗ ψ j1
2
ψ λ2 ∗ ψ σ2 dr 1 dr 2  ð5:44Þ
X
N X
N B j r1  r2 j C
þ Pλσ B
@Ð Ð
C
A
σ¼1 ∗ 1 ∗
λ¼1 ψ i1 ψ λ1 ψ j2 ψ σ2 dr 1 dr 2
j r1  r2 j

In Hij, the first term is a one-electron integral, which consists of operators for kinetic energy and
electron–nuclei interactions. The second term, which is a double sum, contains two-electron double
integrals for Coulomb and exchange interactions. The variable ψ i1 is the spatial basis function of the
first of the two electrons in orbital i, and ψ i2 is the basis function of the second of these electrons. The
variables r1 and r2 are the positions of electrons 1 and 2. The variables i, j, λ and σ are indexes ranging
from 1 to N, and the elements of the charge density matrix P are computed from the eigenvector
matrix C:

X
n=2
Pλσ ¼ 2 cλi cσi ð5:45Þ
i¼1

In order to solve the eigenvalue equation HC¼SCE and to compute the full eigenvector matrix C
and the diagonal eigenvalue matrix E, the matrices H and S must first be computed. But H is a
function of C! How can that be handled? The solution is to first make an initial guess of C and then
iteratively solve the equation with increasingly close approximation until the difference between two
iterations is sufficiently small. This iterative procedure is the Hartree–Fock self-consistent field
algorithm (HF–SCF). If the aim is to optimize the geometry of a molecular structure, an additional
outermost loop is also required. The geometry and energy are iteratively updated until the gradient of
the energy, with respect to the coordinates, is close to zero. This criterion means that the energy
difference between two iterations is small and indicates that the geometry is sufficiently good.
A pseudocode for the HF–SCF algorithm, including geometry optimization, is presented in
algorithm 1.
216 5 Simulation and Modelling of Polymers

Algorithm 1, Hartree-Fock Self-Consistent-Field (QC)

set_basis_set(); %Choose basis functions
calc_geometry(); %Calc initial geometry
while geometry_is_good¼¼false %Geometry loop
calc H,S, P(0),P(1) %Calc H,S,P (Eqs. 30-34)
n¼1 %Initiate index n
while ||P(n)-P(n-1)||>maxError %Hartree-Fock loop
n¼n+1 %Update index n
calc H,S %Calculate H and S
[E,C]¼eig(H,S); %Solve HC¼SCE
calc P(n) %Calculate P
end %End Hartree-Fock loop
check_geometry(); %Calc energy of geometry
if geometry_is_good¼¼true %Check if geometry ok
return %Finish computation
else %Else
calc_geometry() %Update geometry
end %End
end %End

For all N2 components Hij in the approximate Hamiltonian, the double sum requires a total of N2
two-electron integrals to be computed, and this leads to a total computational complexity of Ordo(N4)
to solve a single iteration of the Roothaan–Hall SCF algorithm. A naive implementation of the
equation, without significant simplifications, would thus result in an extremely rapid increase in the
computational cost with increasing number of basis functions N and thus also with the number of
electrons n and the number of atoms. However, the accuracy tends to increase with increasing N, and
this makes it desirable to have a high N-value. To meet this difficulty, two main branches of molecular
orbital (MO) techniques have emerged. Ab initio methods, which are based on first principle,
generally use the exact Hamiltonian and introduce approximations only to the wave function.
Semi-empirical methods simplify the HF equation considerably and use empirical experimental
data to tune parameters in the models. For a given level of accuracy, semi-empirical methods tend
to be faster than ab initio methods, but the latter have a more fundamental foundation, and they can
thus be better for modelling molecular systems where abundant experimental data is not available.

5.2.5 Semi-empirical MO Methods

In semi-empirical molecular orbital methods, the strategy is to simplify the matrices H and S in the
Roothaan–Hall equation so that the computational cost is limited but the accuracy still acceptable. In
fact, an advanced semi-empirical MO method, with parameters determined from a large set of
experimental data, can be as accurate as the corresponding ab initio method.
The earliest semi-empirical MO method was the Hückel theory (Hückel 1931), which was
followed by the extended Hückel theory (Hoffmann 1963). These methods include severe
simplifications, but they can still be used for the qualitative prediction of the outcome of a chemical
reaction or to obtain initial values for more advanced iterative MO methods. To obtain more accurate,
quantitative predictions, less severe simplifications of the Roothaan–Hall equation are however
required. Since the two-electron contributions of the Fock matrix H are the most computationally
expensive, the main aim is to improve H, whereas the overlap matrix S is as usual defined by
5.2 Quantum Chemistry (QC) 217

Ð
Sij ¼ ψ iψ jdV. A trio of historical semi-empirical methods, with increasing accuracy and complexity,
are the CNDO, INDO and NDDO methods.
CNDO (Complete Neglect of Differential Overlap) theories (Pople et al. 1965) assume that for
orbitals in the same atom, ψ i*ψ j ¼ 0 unless i ¼ j. Then S becomes the identity matrix (S¼I), the
double-sum integrals in H are replaced by a single function which depends on the interatomic distance
between the two atoms, and the off-diagonal terms in H become proportional to Slater-type overlap
integrals S with experimentally determined prefactors. Since the double sums in H are removed, the
total computational cost is reduced from Ordo(M4) to Ordo(M2), and the costs of the individual
integrals are also significantly reduced by the incorporation of various simplifications.
INDO (Intermediate Neglect of Differential Overlap) theories (Pople et al. 1967) add double sum
integrals on the diagonal of H, but only if the orbitals are positioned on the same atom. This enables
spin effects to be included. Some parameters calculated by ab initio methods were included in the
original model. An enhanced Modified INDO method (MINDO/3) (Bingham et al. 1975) was later
developed, in which the most important improvements were the construction of parameters from an
extensive database with experimental data rather than from crude ab initio calculations and the
inclusion of a geometry optimization routine in the implementation of the model. Another improved
version is the symmetrical orthogonalized INDO (SINDO1) (Nanda and Jug 1980), which included d-
orbitals in addition to s- and p-orbitals and thus enabled heavier atoms to be modelled.
NDDO (Neglect of Diatomic Differential Overlap) theories (Pople et al. 1970) include a larger
portion of the double sum integrals in H, in particularly all the integrals where the first two orbitals are
on one atom and the last two orbitals are on another atom. The computation of these integrals
increases the number of calculations by a constant factor, which is 100 if only s- and p-orbitals are
considered and 2025 if d-orbitals are also included. In spite of their high computational cost, NDDO
methods are currently the most widely used MO methods. MNDO (Modified Neglect of Diatomic
Overlap), AM1 (Austin Model 1) and PM3 (Parameter Model 3) are three such models that are often
implemented in quantum chemistry programs like MOPAC and AMPAC. Of these models, AM1 is
typically the fastest, whereas PM3 has the largest number of systematically determined experimental
parameters. MNDO, AM1 and PM3 include only s- and p-orbitals, whereas more recent NDDO
methods, like PM5 (Parameter Model 5) and MNDO/d, also include d-orbitals and can thus be used
for elements with a higher atomic number in the periodic table. SRP (specific reaction parameter)
methods are techniques where the standard parameters of a model are perturbed to optimize the model
for a specific class of problem.
Significant efforts have also been made to develop algorithms for semi-empirical MO methods that
are linear scaling, i.e. Ordo(N ). One important application for linear-scaling quantum techniques,
which are cheap but less accurate, is to include them in hybrid quantum mechanics-molecular
mechanics (QM/MM) methods (Warshel and Levitt 1976). Such hybrid methods can, for instance,
be used to simulate chemical reactions in large molecular systems. Linear scaling quantum algorithms
(Lee et al. 1996) can, of course, also be used to model large molecular systems and to compute the
properties of many small molecules in order to construct usable material databases.

5.2.6 Ab Initio Methods

In contrast to the semi-empirical MO methods, ab initio methods use the exact Hamiltonian and only
approximate the wave function. Hartree–Fock is the simplest of the ab initio methods. (Frenkel and
Smit 2002; Leach 2001). The philosophy is that it is better to develop a robust and general
fundamental methodology than to develop methods that are more specific and rely on experimental
data. Ab initio methods are computationally expensive and are today primarily used to model
218 5 Simulation and Modelling of Polymers

relatively small molecular systems, but with increasing computer power in the future, the methods
will be usable for larger systems. Surprisingly, for particular problems, semi-empirical models can
sometimes give results in better agreement with experimental data than ab initio models, the reason
being that the semi-empirical models are calibrated against experimental data, whereas the ab initio
methods are based on first principle calculations. The goal of ab initio methods is to come as close as
possible to the exact solution of the Schr€odinger equation. The energy difference between the exact
solutions of the Hartree–Fock equations and the Schr€ odinger equation is called the electron correla-
tion energy, and a multitude of techniques have been developed to assess this value. The energy
difference is roughly 1%, which can cause serious troubles in certain applications.
The formal computational complexity of Hartree–Fock is Ordo(N4), where N is the size of the
basis set. For comparatively large molecular systems, it is possible to improve the performance
somewhat, for instance, by utilizing the fact that many integrals involving distant particles are
approximately zero or by using fast multipole techniques to reduce the computational cost of the
Coulomb integrals. If these features are included in the implementation, the method scales as Ordo
(N3).
The choice of basis functions in the computations significantly affects on the performance of the
models with regard to the accuracy of the result and to the computational expense. Slater-type orbitals
(STO) are in principle desired, but in order to increase the computational efficiency, the Slater orbitals
are often approximated as linear combinations of Gaussian-type orbitals (Hehre et al. 1969). Three
Gaussian functions are the optimal trade-off between accuracy and speed, which is often abbreviated
STO-3G, ‘Slater type orbitals approximated by three Gaussian functions’. By default, the STO-3G
basis set is constructed as a minimal basis set, with one basis function for each valence orbital. It is
however also possible to add additional basis functions in order to increase the accuracy of the
calculation and to obtain results closer to the Hartree–Fock (HF) limit. This value can be estimated by
iteratively solving the HF equations with an increasing number of basis functions and using these
solutions to extrapolate to the corresponding value for an infinite number of basis functions, i.e. the
HF limit. An extensive database with basis sets, for both molecular orbital and DFT methods, can be
found on the ‘basis set exchange’ homepage, www.basissetexchange.org.
In quantum chemistry methods, an atom or a molecule is generally assumed to be in the electronic
ground state. Phenomena like excited electrons, the loss or gain of electrons, polarization and
chemical reactions can be handled, but accurate electron correlation techniques must be included;
otherwise serious errors can be introduced, even with calculations close to the HF limit.

5.2.7 Density Functional Theory (DFT)

An alternative route to obtain approximate solutions to the Schr€ odinger equation is to search for
optimal electron densities ρ(r) rather than for optimal wave functions. The former strategy is utilized
in density functional theory (DFT) simulations and the latter in the previously described
Hartree–Fock computations. DFT should in principle converge towards the exact solution of the
Schr€odinger equation, whereas an infinitely accurate HF implementation will still deviate from the
exact value. The numerical errors for real DFT implementations are almost identical as for HF, but
there is a fundamental problem with DFT; since the functionals Ψ are unknown, approximate
functionals have to be used. Therefore, DFT results can differ significantly depending on the choice
of functionals.
The theoretical foundations of DFT (Hohenberg and Kohn 1964) are the Hohenberg–Kohn
existence theorem, which shows that the electron density determinesÐ both the Hamilton operator
and the wave function, and the Hohenberg–Kohn variational theorem ψHψ > Eexact, which states
that energies obtained with the variational principle are always greater than or equal to the exact
5.2 Quantum Chemistry (QC) 219

energy. A practical realization of DFT (Kohn and Sham 1965) is made possible by utilizing the
assumption that the Hamiltonian can be written as the sum of energy contributions from
non-interacting electrons together with an energy correction term Exc. For the non-interacting part
of the Hamiltonian, the eigenfunctions can be written as Slater determinants of single-electron
eigenfunctions and the eigenvalues as sums of single-electron eigenvalues. For orthonormal
one-electron orbitals χ i, the electron density then becomes:
X
n
ρð r Þ ¼ jχ i j2 ð5:46Þ
i¼1

Using this definition, the approximate Kohn–Sham energy for a molecular system with n electrons,
Nc nuclei and nuclei charge ZA is:
!
Xn ð ð
∇2i X ρðr 0 Þdr 0
Nc
ZA
Eðρðr ÞÞ ¼ χi  þ þ χ dr þ Exc ðρðr ÞÞ ð5:47Þ
i¼1
2 r
A¼1 iA
2 j ri  r0 j i

The three terms within the integral sign are the energy contributions for kinetic energy of
non-interacting electron, electron–nuclei Coulomb energy and electron–electron Coulomb energy
(also known as the Hartree electrostatic energy). In DFT, the minimum energy is obtained by
differentiating the approximate energy E with respect to the electron density ρ(r) rather than with
respect to the basis function coefficients as in the molecular orbital methods:
∂Eðρðr ÞÞ
¼0 ð5:48Þ
∂ρðr Þ

However, Kohn–Sham DFT and HF use essentially an identical procedure, and in both cases, the
coefficients that give lowest energy are found. The exchange correlation functional Vxc is defined as
the derivative:
∂Exc ðρðr ÞÞ
V xc ðr Þ ¼ ð5:49Þ
∂ρðr Þ

The Kohn–Sham equation for a single electron is written:

ð ! !
∇ 2 XNc
Z ρ ð r 0
Þdr 0
b Ψi ¼
H
KS
 i
þ A
þ þ V xc ðr i Þ Ψi ¼ Ei Ψi ð5:50Þ
i 2 r
A¼1 iA
j ri  r0 j

and this equation can easily be extended to take electron spin into account. Such DFT versions are
called the local spin density functional theory (LSDFT).
Typically, the Kohn–Sham orbitals are written as linear sums of orthonormal, atom-centred basis
functions ψ j, as in MO theory:
X
N
Ψi ¼ cij ψ j ð5:51Þ
j¼1

These basis functions can be Slater-type orbitals (STO), Gaussian-type orbitals (GTO), plane
waves (PW), splines, wavelets or numerical basis functions. Basis set free versions using grids also
exist. If the Kohn–Sham orbitals are expressed as linear combinations of such basis functions, the
220 5 Simulation and Modelling of Polymers

Kohn–Sham equation can, like the Roothan–Hall equation, be expressed as a linear system of
equations, which can be solved as an eigenvalue problem with numerical linear algebra:
HC ¼ SCE ð5:52Þ
ð
Sij ¼ ψ i ∗ ψ j dV ð5:53Þ
V

ð ð ! !
∇2i X ρðr 0 Þdr 0
Nc
ZA
H ij ¼ ψ i  þ þ þ V xc ðr i Þ ψ j dr ð5:54Þ
2 r
A¼1 iA
j ri  r0 j

The choice of exchange–correlation energy Exc and of the exchange–correlation functional Vxc has
a considerable influence on the accuracy of any DFT simulation. The most straightforward strategy is
the ‘local density approximation’ (LDA), where Exc and Vxc depend only on the position of the
electron. Corresponding methods where spin is included are abbreviated LSDA. More advanced and
accurate functionals also take into account the gradient, i.e. the first spatial derivative. Such gradient-
corrected functionals are called ‘generalized gradient approximations’ (GGA). Examples include the
Becke (B) (Becke 1988) and the Becke–Lee–Yang–Parr (BLYP) exchange functionals. The second
spatial derivative can also be included, yielding meta-GGA (MGGA) functionals. Other strategies for
deriving accurate functionals also exist, in particular hybrid Hartree–Fock/DFT functionals like the
classic B3LYP method (Kim and Jordan 1994). Hybrid functionals typically give more accurate
results than basic functionals.
Some components in the Kohn–Sham matrices depend upon the electron density, which has to be
computed by an iterative process. The algorithm for the Hartree–Fock self-consistent field method
(HF– SCF) can be used for that purpose, but with the matrices H and S computed from the
Kohn–Sham rather than from the Roothaan–Hall equations.
The third term in Hij is a double integral which can be expanded into a double sum, as in the
corresponding Hartree–Fock expression. A naı̈ve DFT implementation would thus also scale as Ordo
(N4), where N is the number of basis functions. Fortunately, these expressions can easily be simplified
in DFT by approximating the charge density with another basis set expansion, so that the algorithm
scales with Ordo(N3). For larger molecular structures, linear scaling Ordo(N ) DFT is achievable by
using divide and conquer strategies and by using multi-pole expansion to compute long-range
interactions in repeating structures. Efficient DFT implementations are typically much faster than
HF codes, especially when they are applied on large molecular structures and when reasonable initial
electron densities are used. Semi-empirical DFT methods, like the density functional tight-binding
(DFTB) method (Seifert et al. 1986), also exist, and these include parameters optimized from
empirical experimental data.
Example: DFT on an Antioxidant Molecule Quantum chemistry can be used to predict where on a
molecule it is most probable that other atoms will attach. When a polymer like polyethylene is
chemically crosslinked with peroxides, crosslinking by-products like acetophenone arise. In this
example, a DFT simulation was performed on an acetophenone molecule. The iso-surfaces (Fukui
functions) (Ayers et al. 2009) show in red the regions with the highest probability of electron
attraction. As anticipated, the region near the oxygen molecule is the most electrophilic region
(Fig. 5.2).
5.3 Molecular Dynamics (MD) 221

Fig. 5.2 A DFT-generated Fukui function, which shows in red the most electrophilic regions of an acetophenone
molecule

5.3 Molecular Dynamics (MD)

5.3.1 MD: Overview

Molecular dynamics (MD) is a technique for computing the positions of the atoms in a molecular
system as a function of time t. If the initial positions r(t) and velocities v(t) of the atoms are available,
it is possible to determine the positions at a later time if the acceleration a(t) is known. In MD,
accelerations are computed with Newton’s equation of motion F ¼ ma where F is the force and m the
mass. (Newton 1687):
2
1 ∂ rðt Þ 1 ∂UðrðtÞÞ
aðtÞ ¼ FðrðtÞÞ , ¼ ð5:55Þ
m ∂t 2 m ∂r

The acceleration a(t) is the second time derivative of r(t), i.e. a(t) ¼ ∂2r(t)/∂t2. U(r) is the potential
energy, which is a known function of the atom positions, and F(r) is the spatial derivative of U(r),
i.e. F ¼ ∂U/∂r. New atom positions, for the next time step, are obtained by integrating a(t) twice
with respect to t.
Atomistic motion in MD is described by classical mechanics (Newton’s equation) rather than by
quantum mechanics (Frenkel and Smit 2002; Leach 2001). The neglect of quantum effects can lead to
errors for light, rapidly vibrating atoms, but in practice such errors can be neglected for most material
systems. MD is a deterministic technique, in that once the starting positions, velocities, accelerations
and force fields are set, the molecular system evolves in a predestined way with time, provided that a
suitable numerical algorithm is used. MD is also a molecular mechanics (MM) method, in the sense
that the forces in the molecular system are computed using force-field (FF) parameters, correlating a
geometrical molecular arrangement with the molecular potential energy and force. Certain Monte
222 5 Simulation and Modelling of Polymers

Carlo (MC) methods are also molecular mechanics methods, and most of the theory described in this
section (ensembles, force fields, coarse-graining, etc.) is valid also for molecular MC, at least after
small modifications. The development of MD started in the late 1950s (Alder and Wainwright 1959;
Fermi et al. 1955; Gibson et al. 1960; McCammon et al. 1977; Rahman 1964; Stillinger and
Rahman 1974), but the technique has evolved considerably since then (Gartner and Jayaraman
2019; Kremer and Grest 1990; Páll et al. 2014; Rapaport 1996) partly due the rapid advance of
computer science.
Properties that can be calculated with MD include density, elastic modulus, diffusivity, the radial
distribution function and hydrogen bonding. Applications of MD include protein folding, drug
design, diffusion of small molecules through polymer membranes, mechanical deformations of
semi-crystalline polymers and simulations of polymeric phase transitions (crystallization, glass
transitions and melting). Interactions between polymers and nanoparticles (Karatrantos et al. 2016),
copolymer blends, material interphases, plasticisers in polymers, the sorption of penetrant molecules
and solvent effects are also phenomena that can be analysed with MD.
Every MD simulation involves three steps: (1) generate a suitable starting configuration of the
molecules, (2) relax the initial structure until it reaches a stable and representative low-energy
configuration, and (3) perform production runs on the equilibrated system to compute the properties
of interest.
Initial structures for systems with small molecules are easily generated by placing the molecules
on a lattice, but for polymers, it is computationally better to generate an initial random molecular
configuration that is close to that of the final relaxed system utilizing a Monte Carlo (MC) simulation.
Significant speed-up can be achieved by using MC-generated starting structures instead of using
naı̈ve all-trans starting configurations of the polymer chains.
Relaxation of a MD system is often initiated with a non-dynamic geometry optimization to remove
high-energy atom configurations, i.e. nearly overlapping atoms, strange angles, etc., before the actual
MD simulation. Examples of numerical optimization techniques include the conjugate gradient
method and Newton-type methods. During the first part of the MD relaxation, the force fields can
be turned on gradually to avoid non-physically high initial accelerations. The MD equilibration run
must be long enough to traverse a sufficient amount of the ‘phase space’; otherwise the system
generated will be an inaccurate representation of the real material. A reasonable criterion for
sufficient equilibration is that the energy and density no longer fluctuate significantly with time.
The simulated density should also resemble the experimental density. Realistic MC-generated
starting structures and special techniques such as forced annealing (cyclic increase and decrease of
the temperature) can speed up the relaxation runs and reduce the risk for confusing local and global
energy minima. Note that long polymer chains require long equilibration times.
Production runs in which where material properties of interest are computed constitute the final
step of a MD simulation. Numerous properties can be computed, including diffusivity, solubility,
viscosity, density, glass transition temperature and elastic modulus.
Limitations of the MD technique however exist. The available computer capacity, for example,
limits both the number of atoms in a MD cell and the length of a MD run. Compared to most real-
world experiments, the acceptable number of atoms in MD is small, and the maximum simulated time
is short, even when a high-performance computer is available. Real-world polymer phenomena such
as relaxation, crystallization, melting and stretching take place over extended timescales (minutes,
days, years), whereas the maximum affordable MD simulation time for a typical polymer simulation
is a fraction of a second (nanoseconds, milliseconds). Processes modelled with MD are often
artificially rapid, especially when coarse-graining techniques are used. Some of these difficulties
can be dealt with using periodic boundary conditions, coarse-graining and extrapolation techniques
or by inserting the MD results in a kinetic Monte Carlo (kMC) model, where extended timescales can
be achieved within more affordable CPU time (Andersen et al. 2019; Chatterjee and Vlachos 2007;
Young and Elcock 1966).
5.3 Molecular Dynamics (MD) 223

Fig. 5.3 A periodic MD box showing how molecules are replicated to mimic a continuum material with an almost
infinite number of atoms

Periodic boundary conditions are routinely used in most MD simulations, and they can to some
extent resolve the problem of having a limited number of atoms in a MD cell compared to the number
in a real material. The periodic replications of the atoms from the original simulation cell make it
possible to mimic systems with an infinite number of atoms (Fig. 5.3) (Allen and Tildesley 1989). For
a molecular system with small molecules, a small simulation box is often sufficient to evaluate the
properties of interest within a reasonable computation time. Polymers are unfortunately not small. To
avoid errors from self-interactions from the period images, a simulation cell with dimensions at least
twice as large as the contour length of the longest molecule in the cell is recommended. This
condition is difficult to fulfil for realistic polymers chains (~1E6 atoms), and as a consequence,
shorter polymer segments (oligomers) are often used in MD to represent polymers. Shorter chains
however result in lower elastic moduli of the materials. This can be taken into account by adjusting
the force fields or by extrapolating the results from a series of successively longer oligomer chains to
the desired polymer chain length.
Coarse-graining can also increase the speed of a MD simulation where full atomistic detail is not
required. A system in which all the atoms, including hydrogen atoms, are explicitly shown is called an
all-atom (AA) system, whereas a system where the hydrogen atoms are only implicitly represented is
called a united-atom (UA) system, and a system where an even larger group of atoms is represented as
a single unit is called a coarse-grained (CG) system. Coarse-graining leads to a tremendous speed-up
224 5 Simulation and Modelling of Polymers

of the simulations, but the simplifications can make the results less accurate than with the
corresponding all-atom simulation. A trade-off between speed and accuracy is always involved.
There are many MD simulations programs in existence, inclusive Gromacs, CHARMM, Amber,
GROMOS, NAMD, VOTCA, LAMMPS and Materials Studio.

5.3.2 Force-Field Potentials

A central part of setting up MD simulation is to choose suitable force fields (FF) for the atoms and
molecules in the model. A force field describes how the potential energy U of a molecular system
depends on the positions, bonds, charges and masses of the atoms therein. The force F acting on a
particular atom can be computed by differentiating the potential energy. Traditionally, force fields are
described by analytical equations with material parameters depending on atom type and molecular
arrangement, although it is also possible to describe force fields with numerical interpolation tables.
Material parameters for MD can be derived either from experimental data or from a quantum
chemistry simulation. A large range of force fields, i.e. force-field equations with material parameters
or numerical interpolation tables, have been developed, including the CHARMM, GROMOS, Amber,
Martini, OPLS, TrappEE, Universal and COMPASS force fields. Some force fields result in slow but
accurate simulations, whereas others are rapid and less accurate. Some are particularly good for
organic molecules, whereas others are better suited for metal-containing compounds, and some have
large sets of pre-computed material parameters available, whereas others have smaller amounts. The
contributions to the potential energy can be divided into intra- and intermolecular potentials, the
former taking into consideration contributions from nearby atoms in the same molecule and the latter
taking into consideration all other contributions.
Intramolecular contributions (between nearby atoms inside a single molecule) to the force field in
a polymer system include (1) bond stretching, (2) bond angle bending, (3) dihedral torsions, and
(4) van der Waals interactions between pairs of neutral atoms. Intermolecular contributions (between
two molecules) include van der Waals interactions and (5) electrostatic interactions (Fig. 5.4).
Bond stretching, acting on two connected atoms, increases the potential energy of the molecule.
Stretching is most readily modelled with the equation for an elastic spring (i.e. the harmonic
oscillator), but other stretching equations, including the Morse potential (Morse 1929), the cubic
bond stretching potential or Lagrange multiplier constraints, are also possible. If the equilibrium
distance between the two atoms is r0, the current distance is r and the spring constant is kb, the elastic
spring potential energy contribution is (Fig. 5.5a):

Ustretch ðr Þ ¼ kb ðr  r 0 Þ2 =2 ð5:56Þ

Angle bending can be modelled with a harmonic function resembling the bond-stretch potential.
This bending potential describes the potential energy which arises when the angle between three
atoms in a polymer chain is bent. Other equations, like cosine-based angle potentials or Urey–Bradley
potentials (Urey and Bradley Jr 1931), can also be used in order to reduce the computational cost of
the calculations. If θ is the current angle, θ0 is the equilibrium angle and kθ is the bending constant, the
harmonic angle bending contribution is:

U bend ðθÞ ¼ kθ ðθ  θ0 Þ2 =2 ð5:57Þ

Dihedral torsion contributes to the potential energy by changing the dihedral angles which form
between four nearby atoms in a polymer chain. It is common to distinguish between improper
dihedrals Utorsion1, which are forced to remain within certain bounds, and proper dihedrals Utorsion2,
5.3 Molecular Dynamics (MD) 225

Fig. 5.4 Classical

potential energy
contributions which can be
included in a force field for
polymers: bond stretching,
bond angle bending,
dihedral angle torsion, van
der Waals interactions and
electrostatic interactions

Fig. 5.5 (a) Bending (or stretching) energy, (b) torsion energy as a function of dihedral angle ϕ for an alkene polymer,
(c) Lennard–Jones energy for a pair of atoms
226 5 Simulation and Modelling of Polymers

which can take any value. The former, which are typically used for aromatic rings and similar
structures, can be modelled with harmonic potentials similar to the stretching and bending potentials:

U torsion1 ðϕÞ ¼ kϕ ðϕ  ϕ0 Þ2 ð5:58Þ

whereas the latter can be modelled as a sum of trigonometric functions. One expression for a proper
dihedral, which is often used for alkanes, is the Ryckaert–Bellemans equation, where ϕ is the current
dihedral angle and the Cn’s are torsional material constants (Fig. 5.5b):
X
5
U torsion2 ðϕÞ ¼ Cn ð cos ðϕn  π ÞÞ ð5:59Þ
n¼0

van der Waals potentials are interactions between pairs of non-bonded atoms, but excluding
possible electrostatic contributions. They are assigned to atoms which are either resident in different
molecules or which reside in the same molecule but are separated by two or more other atoms in the
same chain. Lennard–Jones (LJ) potentials (Lennard-Jones 1924) are often used to model van der
Waals potentials. Other potentials such as the Buckingham potential (Buckingham 1938) also exist,
but the LJ potential is better known, is computationally cheaper and has sufficient accuracy for most
applications. For reasons of efficiency, the Lennard–Jones potentials for atoms exactly three atoms
apart are often handled separately. The LJ potential can model all pairwise long-range atom
interactions apart from electrostatic interactions, which are handled separately. If r is the distance
between atom i and j, εLJ is the depth of the potential well for these two atoms, and σLJ is the distance
where the potential curve ULJ crosses the zero line (U ¼ 0) and then the Lennard–Jones potential
ULJ(r) and the force in the x-direction FLJx(r) are given by:
   6 
σ LJ 12 σ
U LJ ðr Þ ¼ 4εLJ  LJ ð5:60Þ
r r
    6 
∂U LJ ðr Þ σ LJ 12 σ x
FLJx ðr Þ ¼  ¼ 4εLJ 12  6 LJ ð5:61Þ
∂ðxÞ r r r2

The factors εLJ and σ LJ are material-dependent constants that are determined from experimental
data. An approximate value of σ LJ is obtained by adding the van der Waals radii of the two atoms. A
distinct feature of the LJ potential is that it contains one repulsive Pauli contribution and one
attractive (dispersive) van der Waals contribution. The former dominates when the atoms are very
close to each other, whereas the latter dominates when the atoms are slightly more separated. Both
terms rapidly approach zero at longer separation distances (Fig. 5.5c).
LJ forces should in principle be applied to all the atom pairs in the simulation cell, and the
distances between all these atom pairs should therefore be calculated in each time step. For a
non-periodic system with N atoms, the distance between each atom and (N-1) other atoms must be
computed, giving a total of N(N-1)/2 comparisons during each iteration. This corresponds to a
computational complexity of Ordo(N2), which indicates a very high computational cost for a polymer
system with many atoms. In a periodic system, the situation is even worse, due to the repeating
boundary conditions. Each atom is then in principle surrounded by an infinitely large number of other
atoms! Simplifications must obviously be made to handle this situation.
The first key simplification is to use the minimum image convention for a system with periodic
boundary conditions, so that for each atom, only the nearest image of each other atom contributes to
the applied force. This reduces the computational complexity of the LJ contributions to Ordo(N2) in
the case of periodic boundary conditions. Note that this simplification destroys the conservation of
5.3 Molecular Dynamics (MD) 227

energy and thus gives rise to seriously incorrect simulation results unless it is combined with the
cut-off technique below.
The second key simplification is to utilize the fact that the LJ potential decreases rapidly with
increasing distance between the atoms in the pair potential. It is therefore reasonable to truncate the
potential and calculate only the contributions from nearby atoms within a given radius. A LJ cut-off
radius of rc ¼ 2.5σ is often chosen. To avoid an abrupt change in potential energy at rc, which could
cause numerical instability, the LJ potential is often both truncated and shifted so that it goes
smoothly to zero at the cut-off radius:
U LJ ðr Þ  ULJ ðr c Þ, r  r c
Utruncated
LJ ðr Þ ¼ ð5:62Þ
0, r > rc

An additional term can if necessary be added to ensure that the first spatial derivative of U (i.e. the
force F) is continuous. The cut-off radius rc must never exceed half the box side in a periodic cell
when the minimum-image convention is applied; otherwise the energy will not be conserved!
When a cut-off is used, it is sufficient to compute LJ forces for atoms within rc. For small
molecular systems, the atom coordinates can be stored in a Verlet list (Verlet 1967), but for large
systems, a faster strategy for finding nearby atoms is required. The most straightforward technique is
a grid search, where the domain is divided into small boxes and vicinity checks are performed only for
atoms in nearby boxes (Fig. 5.3). If the number of boxes is M3 and the total number of atoms is N, the
average number of atoms per box is Nbox ¼ N/M3. If the shortest box size is greater than rc, only the
nearest 33 ¼ 27 boxes (or fewer) need to be checked to find all the atoms within rc. This requires on
average 27N/(2M3) comparisons and a complexity of Ordo(N ) if M is chosen wisely. The atom
coordinates can be stored in a linked list.
Electrostatic Coulomb interactions are the final contributions to the potential energy. These
pairwise, long-range interactions handle the influence of charges in the systems. The basic form of
a Coulomb potential between two atoms i and j with charges qi and qj, permittivity in vacuum ε0 and
dielectric permittivity εr is (Coulomb 1785):
 
1 qi q j
U coloumb ðr Þ ¼ ð5:63Þ
4πε0 εr r

The Coulomb potential decreases with increasing distance between the atoms, but the decrease is
much slower than that of the LJ potential, the growth factors being ~1/r and ~ 1/r6, respectively, which
is a huge difference. It is not thus possible to use a simple cut-off for the Coulomb interactions
without introducing large numerical errors. Fortunately, there are ways of handling this. All the
Coulomb potentials in a periodic system can be efficiently summed with, for instance, fast multipole
methods, Ewald summation, particle-mesh Ewald (PME) summation, particle–particle/particle–mesh
(PPPM) algorithms or reaction field methods (Leach 2001). For a sufficiently large system, the fast
multipole method should in theory be the fastest with a computational complexity of Ordo(N ). For a
medium-sized system, particle–mesh Ewald summation with a complexity of Ordo(Nlog(N )) is often
recommended, and for small systems, the traditional Ewald summation with Ordo(N3/2) or Ordo(N2)
is often the fastest and most accurate.
The total potential energy is described by combing Eqs. 5.56, 5.57, 5.58, 5.59, 5.60, 5.61, 5.62 and
5.63, with atom–atom distance r, σij ¼ (σi + σj)/2 and εij ¼ (εiεj)1/2:
228 5 Simulation and Modelling of Polymers

X  2 X X
U ðr, θ, ϕÞ ¼ kb r ij  r 0 þ k θ ðθ  θ 0 Þ2 þ k ϕ ð ϕ  ϕ0 Þ 2 þ
bonds angles torsion1

X X X  12  6 !
5
σ ij σ ij
þ Cn ð cos ðϕ  π Þ þ 4εij  þ
torsion2 n¼0
r ij r ij
nonbonded
ð5:64Þ
pairsði, jÞ
X  
1 qi q j
þ
4πε0 εr r ij
nonbonded

pairsði, jÞ

The terms in the equation can be formulated in several different ways, but these are the components
that are usually included in a force-field potential for a polymeric system. Hydrogen bonds, which are
important in many (bio)polymers, can either be handled explicitly, by adding an extra term to Eq. 5.64
(Gedde and Hedenqvist 2019a), or implicitly, by adjusting the parameters of the equation.

5.3.3 Coarse-Graining

Force-field potentials can either be all-atom (AA), united-atom (UA) or coarse-grained (CG). In
all-atom systems, each atom is represented explicitly, including hydrogen atoms. In united-atom
systems, most of the hydrogen atoms that are connected to carbons are represented implicitly by
changing the properties of the nearby carbon atom. In coarse-grained systems, adjacent pairs or
groups of non-hydrogen atoms are merged into united atoms with approximately equivalent
properties (Kmiecik et al. 2016; Levitt and Warshel 1975) (Fig. 5.6). The material parameters for
AA force fields are derived from experimental data or from quantum chemistry simulations, but UA
potentials are typically derived from AA MD simulations and CG potentials from UA MD
simulations. Several techniques for deriving CG force-field potentials are possible, including the
Iterative Boltzmann Inversion (Moore et al. 2014; Reith et al. 2003) or the inverse Monte Carlo
(Lyubartsev 2017) method. Small changes in the coarse-grained potential are made until the coarse-
grained force field can reproduce key material parameters from corresponding AA or UA MD
simulations, such as the density and the radial distribution function.

Fig. 5.6 Different

molecular MD
representations of
polyethylene (a) all-atom
(AA), (b) united-atom
(UA), (c) coarse-grained
(CG)
5.3 Molecular Dynamics (MD) 229

Coarse-graining techniques (both UA and CG) can speed up MD simulations, thus enabling larger
MD systems and longer simulations. The speed-up for a coarse-grained system is typically larger than
that which is intuitively assumed. If, for instance, each group of three atoms is merged into a single
super atom, the number of atoms is reduced by a factor of three, but the speed is increased by a factor
much greater than three. The primary reasons are (1) that the MD algorithm may scale worse than
linearly with an increasing number of atoms and (2) that, when hydrogen atoms are removed, most of
the high-frequency vibrations of the system are also removed, which makes it possible to take larger
time steps in the algorithm. Coarse-graining techniques are therefore essential when very large
polymer systems are simulated with almost realistic numbers of atoms. Unfortunately, there are
also problems associated with the use of coarse-grained potentials. The main problem is that the
coarse-graining artificially increases the speed of many processes compared to the speed of a
corresponding all-atom simulation. For instance, diffusion is faster and the elastic modulus decreases.
To some extent, it is possible to neglect this problem, e.g. by scaling procedures or by using relative
comparisons rather than absolute comparisons between different coarse-grained simulations, but it is
not guaranteed that a linear relationship exists. Reverse mapping can be used to coarse-grain a system
to reach equilibrium quickly and then transfer it back to an AA or UA representation, and this can be
an advantage. A huge leap within the scientific field of coarse-graining and multiscale simulations has
been made since the pioneering work (Levitt and Warshel 1975; McCammon et al. 1977; Müller-
Plathe 2002), but additional research is still required. Several coarse-graining MD softwares exist,
including Espresso, LAMMPS, Magic, VOTCA, Culgi, BOCS and Yup.

5.3.4 The Basic MD Algorithm

A MD program in its most basic form can be summarized with Program 2 shown below, where a
simplified MD algorithm is described with pseudocode. A molecular structure is first generated, a
suitable force field is chosen, and initial positions and velocities are assigned to all the particles. A
time loop is then initiated, using small time steps. During each time step, the forces acting on the
particles are determined, and new particle positions are computed by integrating Newton’s equations,
and the properties of interest are evaluated. This loop continues until the maximum time tMax has
been reached.

Algorithm 2, Molecular Dynamics (MD)

Initiate_MD %Prepare initial structure

t¼0 %Set initial time to zero
While (t < tMax) %Iterate
Compute_Forces %Compute all forces
Integrate_Newtons_Equation %Integrate Newton's Eq.
Calculate_Properties %Calculate all properties
t¼t+dt %Update the time
end

Three subroutines are also presented: the initiation of the MD program, the calculation of forces
and the updating of atom positions and velocities by the integration of Newton’s equation of motion.
These subprograms can be maximally simplified if all the ‘molecules’ consist of a single atom, only
Lennard–Jones potentials are used to determine the forces, and all the atoms have the same mass mi.
The first subroutine describes how a MD simulation can be initiated. Vectors xV, yV and zV
containing the initial coordinates of the atoms are first generated, e.g. by inserting the atoms on a grid.
It is also possible to iteratively suggest random off-lattice positions, which are accepted only if the
230 5 Simulation and Modelling of Polymers

distance to previous atoms is sufficiently small, and to iterate until the desired number of particles
N has been inserted into the simulation domain, which is DX*DY*DZ in size. Random velocities in the
x- y- and z-directions, with values between 0.5 and 0.5, are then assigned to all the particles. In order
to conserve the momentum of the system, it is important that the average initial velocity in each
direction is equal to zero. Therefore, for each particle, the average velocity is subtracted from the
velocity in all three spatial directions. The total kinetic energy Ukinetic is computed as the sum of the
individual kinetic energies for each particle:
XN
mi v2i N eff kB T
U kinetic ¼ ¼ ð5:65Þ
i¼1
2 2

where kB is Boltzmann’s constant, T is the temperature and Neff ¼ 3 N-3 for a typical periodic MD
system in 3D.
The initial velocities can be scaled with a constant kscale to give the system the desired starting
temperature T ¼ T0. A slight reformulation of the previous equations yields, for the initial
configuration:

1X  2 N eff kB T 0
N
U kinetic ¼ mi vi kshift ¼ ) ð5:66Þ
2 i¼1 2

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffi u
T0 u X N
kshift ¼ ¼ tN eff kB T 0 = mi v2i ð5:67Þ
T i¼1

The coordinates (in the x-direction) for the time step preceding the initial time step are (with time
step Δt and velocity v) given by:
xold ¼ x  vΔt ð5:68Þ

with corresponding expressions in the y- and z-directions. To avoid particle positions outside the
periodic domain, the mathematical modulus (mod) of the coordinates is computed to ensure that the
values are within the simulation box (Frenkel and Smit 2002).
MD Subroutine 1, Initiate_MD

xV; %Inital x-positions, pss y & z

vxV¼(rand(N,1)-0.5); %Random velocities x, pss y & z
sum_vxV¼sum(vxV); %Sum of velocities x, pss y & z

mv2V¼mV.*(vxV.^2+vyV.^2+vzV.^2); %Energy contributions

U_kinetic¼sum(mv2V)/2; %Kinetic energy
Neff¼3*N-3; %Degrees of freedom
T=2*U_kinetic/(Neff*kB) %Temperature now
k_scale¼sqrt(T0/T); %Scaling coeff (T to T0)

vxV¼(vxV-sum_vxV)*k_scale; %Scaled x-velocity, pss y & z

xVold¼xV-vxV*dt; %Previous x-position, pss y & z
xVold¼mod(xVold+DX,DX); %Periodic BC, x-pos, pss y & z
5.3 Molecular Dynamics (MD) 231

The second subroutine computes the Lennard–Jones potentials (Eq. 5.60) and the corresponding
forces (Eq. 5.61) in the x- y- and z-directions for each atom. For each atom, the distance to each other
atom is calculated, and if the distance is shorter than the cut-off radius rc, the Lennard–Jones potential
ULJ and the forces Fx, Fy and Fz are calculated and updated. Truncation and shifting of the
LJ-potential is applied (Eq. 5.62). This trivial implementation is not optimized for computational
efficiency. For larger systems it is thus necessary to implement a framework that enables grid
searching, and it is also easy to vectorize the inner loop of the program, which also speeds up the
calculations considerably.
The third subroutine suggests new particle positions by integrating Newton’s equations of motion
numerically. A Verlet algorithm (Verlet 1967) is used, where dt is the size of the time step, m is the
mass of the atom, F(t) is the applied force (from the LJ potential), r(t) is the current particle position, r
(t-dt) is the previous position, r(t + dt) is the next position and v(t + dt) is the next velocity.

FðtÞ 2  
rðt þ ΔtÞ ¼ 2rðtÞ  rðt  ΔtÞ þ Δt þ Ordo Δt4 ð5:69Þ
m

vðt þ ΔtÞ ¼ ðrðt þ ΔtÞ  rðt  ΔtÞÞ=2Δt ð5:70Þ

This commonly used Verlet algorithm is derived by taking the difference between the Taylor
expansions of r(t + dt) and r(t-dt). It is equivalent to the Leapfrog algorithm, is time-reversible, is fast,
requires little memory, has third-order accuracy and enables good long-term energy conservation, but
it does not facilitate particularly large time steps (Van Gunsteren and Berendsen 1990). This third
routine also monitors the current kinetic energy (Eq. 5.65), the temperature T and the total energy Utot
of the system:

1 X
N
T¼ m v2 ð5:71Þ
kB N eff i¼1 i i

U tot ¼ U LJ þ U kinetic ð5:72Þ

This implementation is very simplified, but nevertheless it provides a good starting point for
understanding the basic concepts of MD.
MD Subroutine 2, Compute_Forces

U¼0; %Set potential energy U¼0

Fx¼zeros(N,1); Fy¼Fx; Fz¼Fx; %Set x,y,z forces ¼0
src6¼(sigma^6/rc^6); %Shorthand for (sigma/rc)^6
U_rc¼4*ep*(src6^2-src6); %LJ-Energy at cut-off radius

for i¼1:N-1 %Loop over all atoms

for j¼i+1:N %Check neighbors of all atoms

dx¼xV(i)-xV(j); %Distance in x-dir, pss y & z

dx¼dx-DX*round(dx/DX); %Periodic BC x-dir, pss y & z

r2¼(dx*dx+dy*dy+dz*dz); %Dist between atoms i and j

if r2<rc2
sr6¼(sigma^6/r2^3); %Shorthand: (sigma/r)^6
U_LJnew¼4*ep*(sr6^2-sr6); %Compute new LJ energy
232 5 Simulation and Modelling of Polymers

U_LJ¼U_LJ+U_LJnew-U_rc; %Update potential energy

F0¼48*ep*(sr6^2-0.5*sr6)/r2; %Force calculations
Fx(i)¼Fx(i)+F0*dx; %Force i, x-dir, pss y & z
Fx(j)¼Fx(j)-F0*dx; %Force j, x-dir, pss y & z
end
end
end

MD Subroutine 3, Integrate_Newtons_Equation

xNewV¼2*xV-xOldV+Fx*dt^2/m; %New x-positions, pss y & z

vxNewV¼(xNewV-xOldV)/(2*dt); %New x-velocitys, pss y & z
xOldV¼xV; %Update old x-pos, pss y & z
xV¼xNewV; %Update x-pos, pss y & z

U_kinetic¼sum(mV.*(xV.^2+yV.^2+zV.^2))/2; %Kinetic energy

U_total¼U_LJ+U_kinetic; %Total energy
Tnow¼2*U_kinetic/(Neff*kB); %Temperature

5.3.5 Ensembles

Thermodynamic systems are typically modelled with three of the model variables such as tempera-
ture T, pressure P, volume V, total energy E, enthalpy H, number of atoms N or chemical potential μ
fixed. In macroscopic experiments, it is common to keep the temperature, the pressure and the number
of atoms fixed. Some of the most commonly used thermodynamic ensembles are:
• The microcanonical ensemble (fixed NVE)
• The canonical ensemble (fixed NVT)
• The isothermal–isobaric ensemble (fixed NPT)
• The grand canonical ensemble (fixed μVT)
• Gibbs ensemble (fixed μPT)
• The isoenthalpic–isobaric ensemble (fixed NPH)
A default MD code always generates the NVE ensemble, but additional ensembles can be achieved
by utilizing more clever strategies. For instance, a constant temperature NVT ensemble can be
achieved by introducing a thermostat such as the Andersen (1980) or Nose–Hoover thermostat
(Hoover 1985; Nosé 1984). The Andersen thermostat models constant T, corresponding to a heat
bath, by stochastically adding impulsive forces to some random atoms in each time step. The new
velocities are randomly picked from the Maxwell–Boltzmann distribution that corresponds to the
chosen temperature, and this means that the new velocities can be either greater than or less than those
in the previous time step. Andersen’s thermostat is time-efficient and gives accurate static properties
but inaccurate time-dependent properties. The Nose–Hoover thermostat is slower but gives more
accurate time-dependent properties. Other thermodynamic ensembles can be achieved in a similar
manner. Constant pressure, and the NPT ensemble, can be achieved by introducing a barostat, for
instance, the Andersen barostat (Andersen 1980) or by more complex methods. The pressure P(t) at
time t can be calculated as (Leach 2001):
5.3 Molecular Dynamics (MD) 233

!
1XX
N N
1
Pð t Þ ¼ NkB T ðtÞ þ F r ð5:73Þ
V ðtÞ 3 i j>i ij ij

where V(t) is the volume, N is the number of atoms, kB is the Bolzmann constant and Fij and rij are the
pairwise forces and distances between the atoms

5.3.6 The Phase Space

To completely describe a classical molecular system with N atoms, the position ri ¼ (xi, yi, zi) and
momentum pi ¼ mivi ¼ ( pxi, pyi, pzi) must be determined for each atom i. The vectors r ¼ (r1, r2,. . .rN)
and p ¼ ( p1, p2,. . .pN) contain the positions and momenta of all the atoms in the system. Together, r
and p form a point in the phase space which defines the current state of the system (Cramer 2004).
The equilibrium ensemble average of a property A(r,p) in the (canonical) NVT ensemble can in
principle be calculated as the double integral over all possible positions and momenta:
ðð
< A >¼ Aðr, pÞPps ðr, pÞdrdp ð5:74Þ

where the phase-space probability factor Pps(r,p) has a Boltzmann-distributed relation to the energy
Eps(r,p) of the present phase-space position, with normalization constant kps:
 
Pps ¼ kps exp Eps =kB T ð5:75Þ

In practice, it is impossible to integrate A over the whole phase space of a molecular system
(i.e. over all r and q), because the number of potential configurations rapidly becomes astronomically
large when N increases. Fortunately, since only ensemble averages are requested, it is sufficient to
calculate A on a small representative subsection of the phase space. High-energy configurations have
low probabilities and can be sampled with lower probability. Such ensemble average calculations are
used in (molecular) Monte Carlo models.
In molecular dynamics, the time average of the property A is computed instead of the ensemble
average. If the starting structure has a conformation with reasonably low energy, the phase-space
trajectories will probably penetrant relevant parts of the phase space. For a MD simulation with
starting time t0 and ns time steps ti, the time average of the (now time-dependent) property A becomes:

1X
ns
< A >¼ Aðti Þ ð5:76Þ
n i¼1

After an infinitely long sampling time, the equation can be written:

t0ðþt
1
< A >¼ lim ðt ! 1Þ AðτÞdτ ð5:77Þ
t
t0

In this limit, where the time average no longer depends on the starting point t0, the time average
coincides with the ensemble average, according to the ergodic theorem. In all equilibrium MD
simulations, the aim is to come sufficiently close to this limit, so that reliable steady-state statistics
234 5 Simulation and Modelling of Polymers

can be obtained. If the starting atom positions and momenta are Boltzmann distributed, the distribu-
tion will (in principle) be preserved during the simulations and sampling will be carried out in the
NVT ensemble.
A rule of thumb ensuring that the simulation time is sufficiently long simulation is that the polymer
chains on average should move a distance greater than (a specific fraction) of their radius of gyration
Rg (Auhl et al. 2003). Rg increases linearly with increasing chain length, whereas the distance moved
increases as the square root of time. A 10-fold increase in chain length should therefore give a
100-fold increase in CPU time, even if the total number of atoms is unchanged. Otherwise the ergodic
hypothesis will easily be violated. It is consequently almost impossible to reach ergodic conditions
for long polymer chains with standard MD equilibration techniques. During the last decades, a
multitude of more advanced techniques have been developed to handle this problem, often involving
MD–MC hybrid methods, kinetic MC and/or multiscale modelling (Auhl et al. 2003; Karayiannis
et al. 2003; Lemarchand et al. 2019; Masubuchi and Uneyama 2018; Sliozberg et al. 2016).

5.3.7 A Practical MD Example: Diffusion

A basic MD simulation is presented here to show how a simple MD simulation can be carried out.
Several of the steps are significantly simplified (shortened, etc.) to increase the clarity of the
simulation.
An initial polymer structure, including penetrant molecules, was first generated with a Monte
Carlo technique using the amorphous builder toolbox in Materials Studio. An all-atom simulation
force field (COMPASS) was used with the target temperature 300 K and target density 0.9 g/cm3
(i.e. amorphous PE). A total of ca. 8000 atoms was used, corresponding to a box size of ca. 40 Å3. The
total number of methane penetrant molecules was 20, the number of polymer chains was 25 and the
length of each polymer chain was 100 carbon atoms. The initial structure was geometry optimized
and equilibrated using 5 cycles of NVE forced annealing (300 K–500 K), until the energy E had
stabilized. (Fig. 5.7a) After this equilibration, a production run in the NVT ensemble with the Nose-
Hoover thermostat was performed for 50 ps, and the coordinates of all the atoms were stored as
functions of time. When the simulation had been completed, the mean square displacement <Δr(t)2 >

Fig. 5.7 MD diffusion simulation (a) non-equilibrated polymer system with penetrant molecules, (b) mean square
displacement versus time
5.4 Monte Carlo Methods (MC) 235

of the penetrant atoms was calculated and plotted as a function of time (Fig. 5.7b). Finally, the
diffusivity D was estimated with the Einstein relation in 3D (Einstein 1905), where the diffusivity D is
approximated by the slope of the line for the mean square displacement versus time:
D E
1 X
N
Δr ðtÞ2 ¼ ðr ðtÞ  r ðt0 ÞÞ2 ð5:78Þ
N t¼0

D E
2
1 ∂ Δr ðtÞ
D¼ ð5:79Þ
6 ∂t

More advanced approximations of D can be obtained by fitting the mean square displacement with
more complex equations that include the initial cage-like diffusion in the beginning of the curve. The
diffusivity can also be computed from the diffusivity autocorrelation function (Hansen and
McDonald 1986).
Even though this diffusion example is very simple compared to current research front (van der
Giessen et al. 2020), it still demonstrates the main steps of a typical MD simulation.

5.4 Monte Carlo Methods (MC)

5.4.1 MC: Overview

In a broad sense, Monte Carlo (MC) techniques are methods that use numerous random numbers to
predict the properties of interest (Metropolis 1987). For example, the volume of a complex geometry
can be approximated with a naive MC algorithm by inserting the object in a container and randomly
placing dots inside this domain. The fraction of dots inside the object then corresponds to the volume
fraction of the object. Another example is that a spherulitic crystal structure in semi-crystalline
polyethylene can be modelled with mesoscale MC by mimicking how the crystals grow, rotate and
branch, using experimental input parameters for branch angles, crystal aspect ratios, etc. (Fig. 5.8)
(Nilsson et al. 2009).
In computational polymer chemistry, the main application of MC is to model chemical structures
such as polymer chains at a molecular level (Leach 2001). Molecular MC can be used to generate and
equilibrate polymeric structures and to evaluate their time-independent equilibrium properties, like
density and glass transition temperature. Interactions between different polymers and between
polymers and other chemical species can also be assessed.
In contrast to molecular dynamics (MD), the chain movements with MC are stochastic rather than
deterministic, with the local chain movements being governed by random numbers and probabilities.
In MD, Newton’s equations of motions are solved as functions of time, giving time-dependent
solutions that are deterministically dependent on the starting conditions of the simulation set-up. If
identical starting configurations are used in two different MD runs, the results of the two simulations
will be identical. In contrast, MC generates time-independent solutions, and since random numbers
are used, two MC simulations with identical starting conditions will give different results. Standard
MC cannot therefore be used to simulate time-dependent properties such as diffusion. More recent
hybrid methods like kinetic Monte Carlo can overcome this deficiency (Andersen et al. 2019).
In spite of the differences between MC and MD, their steady-state solutions will on average
be the same, provided that they use the same force fields and that a sufficient fraction of the phase
space has been covered so that the systems are close to their global energy minima. This is a
236 5 Simulation and Modelling of Polymers

Fig. 5.8 A polyethylene spherulite generated with mesoscale MC. From Nilsson et al. 2009, with permission from
Elsevier

consequence of the ergodicity theorem which states that energy minimization in time is ultimately
equal to the energy minimization in space. Both MD and force-field-controlled molecular MC thus
belong to the class of molecular force-field methods, also called molecular mechanics (MM) methods
(Frenkel and Smit 2002).
One advantage of MC is that MC algorithms can break physical laws but still give a correct
solution, whereas MD must follow the laws of physics. For instance, in the grand canonical ensemble,
the number of atoms fluctuates. The sudden insertion or deletion of atoms is clearly non-physical, but
it can still be simulated with MC. The Widom insertion–deletion algorithm, which is used to compute
the solubilities of penetrant molecules in polymers, is one such example (Widom 1963). The
generation of initial polymer systems for MD or MC simulations is also a classical MC application.
Such structures can be obtained by modelling polymerization either with self-avoiding polymer
chains or with ghost chains. The orientation of each such polymer segment is randomly chosen
with a probability related to the change in potential energy. To increase the computational efficiency
of MC, special strategies like biased Monte Carlo techniques are typically utilized.

5.4.2 Importance Sampling: Metropolis Hastings and Biased Sampling

Energy minimization of a molecular polymer system at a given temperature T, volume V and particle
number N (i.e. the canonical NVT ensemble) is a typical MC task. The probability of a specific chain
movement is then related to the energy difference before and after the move. A move leading to a
5.4 Monte Carlo Methods (MC) 237

molecular configuration with lower energy than before should have a high probability and vice versa.
Sometimes, however, it is necessary to accept movements that temporarily result in higher energies;
otherwise it is impossible to move away from the local energy minima and approach the global energy
minimum. This can be achieved with importance sampling in contrast to naı̈ve random sampling. The
first importance sampling algorithm, the standard Monte Carlo Metropolis–Hastings (MCMH)
algorithm (Hastings 1970; Metropolis et al. 1953), is a classic technique that is still widely used in
many fields. If a suggested MCMH move leads to a lower potential energy than the previous iteration,
the move is automatically accepted. If the energy is higher, the probability of accepting the move is
small (20–50%), but non-zero (Mountain and Thirumalai 1994). The probability that the move is
accepted is given by:
paccept ðold ! trialÞ ¼ min ð1, exp ððEold  Etrial Þ=kB T ÞÞ ð5:80Þ

where Eold is the potential energy of the system before the trial move, Etrial is the energy after the trial
move, kB is Boltzmann’s constant and T is the temperature. Thus, if the random number (between
0 and 1) is lower than the temperature-dependent energy, the move is accepted, otherwise not.
Spherical particles, such as solitary atoms, have three degrees of freedom; the translations in the
x-, y- and z-directions. Their possible moves can be written:
xnew ¼ xold þ ð2  rand ð0, 1Þ  1ÞΔxmax ð5:81Þ

ynew ¼ yold þ ð2  rand ð0, 1Þ  1ÞΔymax ð5:82Þ

znew ¼ zold þ ð2  rand ð0, 1Þ  1ÞΔzmax ð5:83Þ

where Δxmax ¼ Δymax ¼ Δzmax is the largest allowed translational change in a single iteration and
rand(0,1) is a random number between 0 and 1. Eqs. 5.80, 5.81, 5.82 and 5.83 illustrate the
foundations of the Monte Carlo concept. In the following MC example, which is a mesoscale rather
than a molecular simulation model, the construction of a composite structure containing a large
fraction of spherical particles is demonstrated. Representative composite structures like these can be
used to compute effective composite properties with FEM or to construct three-dimensional
visualizations of two-dimensional SEM composite micrographs (Pallon et al. 2016).
Example: MC of Spherical Particles The task is to construct a representative periodic section of a
composite containing a high fraction (>50 vol.%) of randomly positioned spherical fillers. The
process involves three steps. Firstly, spherical solid particles with suitable radii are inserted randomly
one by one into an initially empty cubic domain until a sufficiently high intermediate filler fraction
has been obtained. This intermediate filler fraction can, for instance, be 20% of the final target filler
fraction. If a new particle collides with an already existing particle, the insertion trial is rejected, and a
new trial at another random position is tested. Secondly, the radius of all the spherical particles is
multiplied by a constant until their total volume corresponds to the desired final target filler fraction.
This step will inevitably result in particle overlap, i.e. high-energy positions. Thirdly and finally, the
MHMC procedure of Eqs. 5.80, 5.81, 5.82, 5.83 and 5.84 is initiated. A suitable simple force field is
assigned to all the particles, where non-overlapping particles have no potential energy, whereas
overlapping particles have a very high potential energy that decreases slightly with increasing inter-
particle distance. The step size (Δxmax) is adjusted so that the acceptance ratio of a move is ca. 50%.
The MHMC algorithm is then iterated until the total potential energy of the system is smaller than the
target energy Eaim, which is zero if no overlapping particles are allowed. The third step, i.e. the
MHMC step, is summarized in algorithm 3, and a typical generated geometry is shown in Fig. 5.9.
238 5 Simulation and Modelling of Polymers

Fig. 5.9 A representative

periodic box of a composite
containing 50 vol.% of
spherical fillers, generated
with the MC technique

Algorithm 3, Metropolis-Hastings Monte-Carlo (MC)

Compute the initial system energy (Eold)

While (Eold > Eaim)
Chose a random particle
Suggest a random trial move of that particle (Eq. 2-4)
Calculate the system energy (Etrial) after trial move.
If Etrial<Eold
Enew¼Etrial
else
If rand(1)exp((Eold-Etrial)/kBT)
Enew¼Etrial
else
continue
end
end
Eold¼Enew
end

Non-spherical rigid objects, such as frozen molecules, have up to three rotational degrees of
freedom in addition to their three translational degrees of freedom. The orientation of a rigid object
can be described by its orthonormal basis vectors x, y and z. In the Cartesian coordinate system, these
vectors are initially x ¼ [1 0 0], y ¼ [0 1 0] and z ¼ [0 0 1]. If standard Euler angles are used, the
rotations in three dimensions can be described by the rotation matrix R (Leach 2001):

2¼
R 3
cos ϕ cos ψ  sin ϕ cos θ sin ψ sin ϕ cos ψ þ cos ϕ cos θ sin ψ sin θ sin ψ
6 7 ð5:84Þ
6  cos ϕ sin ψ  sin ϕ cos θ cos ψ  sin ϕ sin ψ þ cos ϕ cos θ cos ψ sin θ cos ψ 7
4 5
sin ϕ sin θ  cos ϕ sin θ cos θ
5.4 Monte Carlo Methods (MC) 239

where the variables ϕ, θ and ѱ are the Euler angles and Δϕ, Δθ and Δѱ are the changes in the
angles of orientation in this iteration, i.e. Δϕ ¼ ϕnew-ϕold, etc. Euler angle rotations correspond to
successive rotations around the local z-, x-, and z-axes. For a rigid object with an old orientation
vector vold ¼ [v1old, v2old, v3old]’, the new orientations are computed as:
vnew ¼ Rvold ð5:85Þ

To achieve sampling with an equal probability for all orientations, the sampling of θ must be
treated with extra care; otherwise some orientations are given a greater probability. If the sampling of
θ is done with respect to cos(θ), the probabilities become correct. Completely random rotations,
suitable for MHMC moves, can then be computed:
ϕnew ¼ ϕold þ 2  ðrand ð1Þ  1ÞΔϕmax ð5:86Þ
 
θnew ¼ a cos cos θold þ ð2  rand ð1Þ  1ÞΔð cos θÞmax ð5:87Þ

ψ new ¼ ψ old þ 2  ðrand ð1Þ  1ÞΔψ max ð5:88Þ

The variables Δϕmax, Δ(cosθ)max and Δѱmax are constants defining the maximum allowed change
in these angles during a single iteration, corresponding to the definition of Δxmax. Using Eqs. 5.80,
5.81, 5.82, 5.83, 5.84, 5.85, 5.86, 5.87, 5.88 and 5.89, MHMC simulations of arbitrary rigid objects
can be simulated, similarly to the MHMC simulation of spheres.
Although the examples hitherto in this paragraph relate to spherical particles and rigid objects
rather than to polymers and other molecular structures, the basic concept of Metropolis– Hastings
importance sampling is clearly demonstrated.
Biased Monte Carlo (Rosenbluth and Rosenbluth 1955) techniques are often used to improve the
computational efficiency of MC simulations. For macromolecules, the basic form of the MHMC
algorithm unfortunately often leads to high-energy states that are almost always rejected, resulting in
slow convergence of the algorithm. One strategy for dealing with this situation is to sample high-
energy particles with higher probability, i.e. to weight the probability of accepting a trial move by a
factor related to the energies before and after the trial move. This important technique is called biased
Monte Carlo. If f is a factor depending on the potential energy of the system, the probability of
accepting the trial move is:
  
f ðEold Þ ðEold  Etrial Þ
paccept ðold ! trialÞ ¼ min 1, exp ð5:89Þ
f ðEtrial Þ kB T

Cleverly constructed biased Monte Carlo strategies can be several orders of magnitude faster than
standard MCMH when applied to polymers and other macromolecules. Several biased MC models
are available, including orientational bias MC (Rosenbluth and Rosenbluth 1955), configurational
bias MC (Siepmann and Frenkel 1992), force-bias MC (Pangali et al. 1978; Rao and Berne 1979) and
Smart MC (Rossky et al. 1978). Some of these techniques were first applied to on-lattice geometries,
but nowadays off-lattice strategies are mainly used (Sliozberg et al. 2016).

5.4.3 Macromolecular Starting Configurations with MC

A key application of MC in polymer modelling is for the efficient generation of start geometries that
are reasonably close to the final equilibrated structure. Early attempts to generate temperature-
240 5 Simulation and Modelling of Polymers

Table 5.1 Probabilities of achieving a specific orientation of the next atom if the orientation of the previous atom is
known
Probability T (new) G (new) G’ (new)
T (old) 1/(1 + 2σ) σ/(1 + 2σ) σ/(1 + 2σ)
G (old) 1/(1 + ѱσ + ωσ) ѱσ/(1 + ѱσ + ωσ) ωσ/(1 + ѱσ + ωσ)
G’ (old) 1/(1 + ѱσ + ωσ) ωσ/(1 + ѱσ + ωσ) ѱσ/(1 + ѱσ + ωσ)

dependent polymer structures with locally realistic morphologies were made by Flory (1989). Flory’s
strategy, which is described in Fundamental Polymer Science (Gedde and Hedenqvist 2019b),
simulated the polymerization of linear alkanes in a manner such that the orientation of the next
atom in the chain depended only on the positions of the three previous atoms. This is thus a ghost
chain strategy which generates chains that are not self-avoiding. The radius of gyration for a Flory
polymer chain is correct for a molten polymer in the theta state. For simplicity, only the three local
energy minima of each rotation are considered: the trans (T), gauche (G) and anti-gauche (G’)
positions. The old orientation affects the probability of the new state as described in Table 5.1:
Table 5.1 describes the orientation probabilities for a general symmetric polymer molecule, with
constants ѱ and ω depending on molecular structure and with the variable σ depending on the
temperature T, gas constant R ¼ 8.31 and energy E:
σ ¼ exp ðE=RT Þ ð5:90Þ

For polyethylene, the constants are ѱ ¼ 1, ω ¼ 0, E ¼ 2.5 kJ/mol, ϕ ¼ 0 and θ ¼ 30 . In this
case, T, G and G’ correspond to torsion angles of 180 , 67 and 293 , respectively (Flory 1989; Nairn
2003). Mattice and Suter (1994) have tabulated material constants for a multitude of polymers.
Although molecular modelling has evolved considerably since Flory’s concept was developed, the
simplicity of his model still makes it useful for understanding how the polymerization of polymers
can be modelled. MC models based on the Flory concept have more recently been used to provide
starting structures for semi-crystalline polyethylene sandwich structures, which were energy
minimized numerically and relaxed with MD until the final realistic MD structures were obtained
(Fig. 5.10) (Moyassari et al. 2017).
Nowadays, many programs for the generation of initial polymer structures are based on algorithms
for self-avoiding random walks. The first random-walk algorithms were performed on lattice
structures (on-lattice), whereas most recent algorithms have no underlying lattice (off-lattice). In
self-avoiding random-walk polymerization models (Dijkstra 1997; Martin and Siepmann 1999), a
suggested new position is always rejected if the new atom is too close to already existing atoms. In
recoil growth algorithms (Alexandrowicz 1998; Meirovitch 1988), the outermost atoms in the chain
are removed when it is impossible to avoid a collision and the chain growth is restarted at an earlier
position. All possible chain-front configurations can be precalculated a couple of steps in advance to
ensure that routes leading towards low-density polymer regions have a higher probability of being
chosen. The initial polymer structures can be generated at a very low density (~50% of the final
density), and the process can gradually be ramped up to the desired temperature (with MD or MC).
Ramping can increase the probability of finding acceptable starting structures and thus makes the
algorithm more stable and reliable.
A growing number of commercial and free-ware programs have been developed for generating
initial polymeric structures (Lemarchand et al. 2019). Some of the more widely used programs are
Moltemplate, Materials Studio (amorphous cell toolbox) and Enhanced Monte Carlo (EMC) (in’t
Veld and Rutledge 2003]).
5.4 Monte Carlo Methods (MC) 241

Fig. 5.10 Simulation of a representative section of a semi-crystalline polyethylene structure. (a) Initial structure
generated with Flory-type MC technique, (b) The same structure after MD relaxation. From Moyassari et al. 2017, with
permission from AIP publishing

5.4.4 Macromolecular Energy Minimization with MC

When an initial period cell has been constructed containing polymers and other chemical species
(penetrant molecules, small nanoparticles, solid surfaces, etc), the next step is to relax and energy
minimize the system, either with MD or MC. After this equilibration step, a final production step
follows, where the properties of interest, such as diffusivity, solubility, Young’s modulus, density or
glass transition temperature, Tg, are calculated. If MC is used for the energy minimization, force fields
should be chosen that accurately mimic the properties of the materials. The energy contributions to
the force fields typically consist of the same terms as in MD, i.e. intramolecular energies (translations,
vibrations, rotations, dihedrals) and intermolecular energies (Lennard–Jones energies, van der Waals
energies, hydrogen bonds energies, Coulomb potentials). Force fields similar to those used in MD
(Amber, GROMOS, OPLS, TrappEE, etc.) can often be used in MC, both for all-atom (AA), united-
atom (UA) and coarse-grained (CG) molecular structures. Differences however exist, and a MD force
field will probably work better in a MD than in a MC simulation and vice versa. For example, the
bond lengths are typically constant in MC and variable in MD. Mixing force fields from different
sources is generally not recommended since this can decrease the quality of the simulation.
As an eventual first step before the MC energy minimization, the system can be energy minimized
with a numerical optimization algorithm, such as conjugate gradient methods, Newton–Raphson
methods, or generalized minimum residual methods (GMRES). This step will change high-energy
conformations to lower-energy states, and it is especially important if the probability of initial high-
energy conformations is very high, such as in ghost chain structures.
242 5 Simulation and Modelling of Polymers

The main step during a MC energy minimization is to systematically suggest different chain
movements and accept or reject them using an importance sampling MC technique. The choice of the
suggested movements is crucial for the performance of the MC algorithm. For a small molecule, like
ethane, it is easy to suggest suitable movements, for instance, (1) translation of the whole molecule,
(2) rotation around the molecular centre of gravity or (3) rotation around a single bond in the
molecule. For a small molecule, all the suggested movements will be accepted with high probability.
For a macromolecule it is much more demanding to construct good trial movements, because most
movements result in collisions with other atoms and therefore have a high probability of rejection.
The rotation of a bond angle in the middle of a macromolecule will result in a huge translational move
of the atoms at the end of the chain. Since non-physical movements can be tolerated in MC as long as
they ultimately reduce the potential energy of the polymer system, it should in principle be possible to
construct a MC that converges more rapidly than MD towards the global minimum. However,
without a particularly clever strategy of choosing trial movements, MD will be faster. Examples of
large trial movements that can be efficiently included are chain reptation, Utha-twisting, chain-
breaking, atom-atom reconnections, crankshaft and kink–jumps (Kremer and Binder 1988; Verdier
and Stockmayer 1962). Energy-minimizing MC programs include the Enhanced Monte Carlo (EMC)
program.
A standard MC cannot model dynamic, time-dependent properties, but this can be achieved with a
special type of MC, called kinetic Monte Carlo (kMC). The advantage of this technique is that
polymers can be studied over much longer timescales than is otherwise possible. The drawback is that
a kMC simulation requires the output from time-dependent (MD) simulations as input and that kMC
codes are less commonly available than standard MD or MC codes (Chatterjee and Vlachos 2007).

5.5 Mesoscale Modelling, Including Dissipative Particle Dynamics

Material models that don’t have a molecular detail level, but where the material is still not treated as a
continuum, are referred to as mesoscale models. Such models cover time and length scales between
the limits of microscopic and macroscopic models. In multiscale modelling, where attempts are made
to bridge all length scales from quantum mechanics to classical physics, the mesoscale transition is
often the most difficult to bridge. A wide range of models can be classified as mesoscale simulation
models, for instance, MC models for polyethylene spherulites (Fig. 5.8) or FEM effective media
models for composites (Fig. 5.16). In chemical science, mesoscale models typically handle the
dynamics of soft matter, such as the simulation of the mixing of polymer blends in the molten
state. Examples of mesoscale models for modelling soft matter include dissipative particle dynamics
(DPD) (Hoogerbrugge and Koelman 1992), Brownian dynamics (BD) (Allen and Tildesley 1989),
Lattice Boltzmann (LB) (Ladd 1993) and lattice gas cellular automata (LGCA) (Malevanets and
Kapral 2000). The representation of the material in these models is based not on individual atoms and
molecules but instead on soft spheres (DPD and BD) or lattice points (LB and LGCA) with properties
which, to some extent, can capture the nature of the simulated molten polymers. Both DPD and BD
are structured similarly to MD, with the difference that other force fields are used and that a fluid of
soft spheres is modelled rather than atoms and molecules. DPD conserves energy, in contrast to BD,
and it often generates more realistic results than the other models.
Dissipative particle dynamics uses force fields that include both conservative contributions, like
the Lennard–Jones potential, and random (non-conservative) contributions. The DPD technique is
thus more stochastic than MD but less stochastic than MC. It also includes a dissipative friction
contribution. The total force acting on particle i in a DPD simulation is (Frenkel and Smit 2002):
5.6 Statistical Methods, Including Group-Contribution Methods 243

X      
FDPD,i ¼ FC rij þ FR rij þ FD rij , vij Þ ð5:91Þ
j6¼i

with the spatial difference rij ¼ (ri-rj) and velocity difference vij ¼ (vi-vj) between the centres of the
soft spheres i and j. Corresponding normalized vectors are denoted b rij =rij =r ij and b vij =vij , where
vij =b
vij ¼ ||vij|| and rij ¼ ||rij||, i.e. rij is the distance between the points ri and rj. The random force FR and
the dissipative force FD are given by:
  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 
FR rij ¼ 2kB Tγϑ r ij  ξijb rij ð5:92Þ

     
rij b
FD rij, vij ¼ γϑ rij vij  b rij ð5:93Þ

where ξij is a Gaussian random variable, γ is a constant friction coefficient and ϑ(rij) is a distance-
dependent friction coefficient, which can be defined as:
( 2
  1  r ij =r c , r ij < r c
ϑR r ij ¼ ð5:94Þ
0, r ij  r c

The constant rc is a cut-off radius, similar to the cut-off radius for the LJ potential in MD. Finally, a
conservative force needs to be defined, for instance, as:
(  
  aij 1  r ij brij , r ij < r c
FC rij ¼ ð5:95Þ
0, r ij  r c

The variable aij adjusts the force, depending on the chemistry of the two spheres i and j. For a
typical two-component mixture with species A and B, aij can take three different values: one when
both spheres are of species A, one when both are of species B and one when one sphere is of A and
one of B.
DPD force fields are usable in MD simulation programs, and DPD simulations can be used to
model the mixing of soft matter on much longer time and length scales than traditional MD
simulations. Complex and heterogeneous viscous fluids, like molten polymer blends and polymer
micelles, are suitable for DPD simulations. Examples include the rheological analysis of deformable
soft matter like concrete and other multiphase flow simulations of deformable substances such as
blood cells (Blumers et al. 2017).

5.6 Statistical Methods, Including Group-Contribution Methods

In polymer science, statistical models can, in addition to standard statistical analysis of research
results, be used to predict polymer properties using large compilations of material data. This is called
material informatics. Databases with material properties can include data either from experiments or
from simulations. In general, the statistical predictions will become more reliable with increasing size
and accuracy of the database. Relevant databases include the Cambridge Structural Database and the
Materials Genome Initiative (Liu et al. 2020). A multitude of statistical material models exist, many
belonging to the large group of quantitative structure–activity relation (QSAR) models, which are
used to find relations between different material properties or between material structures and
244 5 Simulation and Modelling of Polymers

material properties. Group-contribution methods form an important QSAR subclass. Linear or

non-linear regression is used as the basis of most statistical methods, sometimes complemented
with advanced techniques such as machine learning (Kanazawa et al. 2019; Liu et al. 2017), data-
mining (Bock et al. 2019), artificial neural networks (Pyzer-Knapp et al. 2015), artificial intelligence
and other big data techniques. Common applications of statistical methods in polymer science are:
(1) predict a specific material property when the polymer structure and/or other material properties
are known, (2) design a polymer with desired properties or (3) choose the optimal polymer for a
certain task, for instance, with the program GRANTA Selector.
Group-contribution methods use statistics to study how different molecular subunits contribute to
the properties of a material. Structure–property relationships are extracted from material databases,
and this information is used to evaluate how different parts of a molecule influence the overall
properties of the material. Numerous material properties can be predicted using group-contribution
methods, including density, glass transition temperature, heat capacity, surface energy, viscosity,
dielectric polarization, magnetic susceptibility, solubility, permeation and elastic modulus. Group-
contribution simulations are very rapid in a computational perspective, and this makes them ideal for
finding promising material candidates from a large group of polymers, for estimating a missing
material parameter for a specific polymer, for predicting the properties of a new material or for
optimizing the molecular structure of a polymer molecule intended for a particular usage. The
drawbacks of group-contribution methods are that the accuracy is limited by the accuracy of the
underlying database and that the predictions are often imprecise outside the parameter range of the
database, for instance, outside a specific temperature range. In contrast to more detailed and time-
consuming simulations methods, group contributions can typically not predict phenomena that have
not already been observed among the materials in the underlying database. Some famous group-
contribution methods in polymer science are the UNIFAC model (Fredenslund et al. 1975), which is
used for predicting the chemical potentials and binding energies of polymer molecules; Hansen’s
solubility parameters (Hansen 2007), which can predict whether two chemicals are mutually soluble;
and van Krevelen’s group-contribution methods (Van Krevelen and Te Nijenhuis 2009), which
provide an extensive library of models for predicting a wide range of polymer properties. The rest
of this subchapter deals with methods related to van Krevelen’s findings.
van Krevelen models, and similar group-contribution models, divide the repeat unit of the polymer
into smaller pieces, such as hydroxyl or carbonyl groups. For each material property, an effective
contribution value is assigned to each of these pieces. Regression on large material databases is used
to compute the contributions.
Assume that the repeat unit of a particular polymer has been divided into n pieces, numbered
i ¼ 1..n, and that we wish to compute a property y of this polymer with a van Krevelen method. If gi is
the contribution of piece i, the simplest group-contribution equation for the determination of y is a
non-weighted sum of the contributing factors gi:

  X
n
!
y g ¼b gi ð5:96Þ
i¼1

If the contributions gi and the constant b are known, y is determined by identifying and summing
the group-contribution factors of a single repeat unit of the polymer.
Otherwise, if the gi values are unknown, they can be determined from databases containing known
y-values of polymers with known structures. The polymer structures are stored in a matrix A, where
each row corresponds to a single polymer. If row j reads A ¼ [2 0 1], it means that polymer j has two
units of piece one, zero of piece two, and one of piece three. When all polymers in the database are
combined, they together form an overdetermined system of equations:
5.6 Statistical Methods, Including Group-Contribution Methods 245

Fig. 5.11 Molar volumes for glassy amorphous (Va) and crystalline (Vc) polymers as functions of the van der Waals
volume (Vw). Drawn after data collected by (Van Krevelen and Te Nijenhuis 2009)

Ag ¼ yðgÞ=b ð5:97Þ

where y(g) is a known vector containing y-values for all polymers in the database, whereas g is an
unknown vector containing the group-contribution factors. Such a system of equations (with b ¼ 1) is
easily solved by typing g ¼ A\y in MATLAB.
Non-weighted sums have, for instance, been used to determine the amorphous and crystalline
molar volumes Va and Vc for polymers (Van Krevelen and Te Nijenhuis 2009):
X
n
V a ¼ 1:60 V w,i ð5:98Þ
i¼1

X
n
V c ¼ 1:435 V w,i ð5:99Þ
i¼1

The group-contribution parameters Vw,i are the van der Waal volumes of different molecular
groups (Bondi 1968), and the factors 1.60 and 1.435 were determined as the slopes of the linear
relationships between Va and Vw and between Vc and Vw for a large set of polymers (Van Krevelen and
Te Nijenhuis 2009) (Fig. 5.11).
The density ρsc of a semi-crystalline polymer with crystal fraction ϕc, repeat unit weight Mrw and
volume Vsc becomes:
Mrw Mrw
ρsc ¼ ¼ ð5:100Þ
V sc ϕc V c þ ð1  ϕc ÞV a

As an example, the density of amorphous (ϕc ¼ 0) poly(ethylene terephthalate) (PET) with

Mrw ¼ ∑Mi ¼ 192.2 g/mol can be determined using tabulated data (Table 5.2) in the above equations.
A PET repeat unit contains two (-CH2-), two (-COO-) and one (-C6H4- (ring)) groups, and these give
the amorphous PET density:
246 5 Simulation and Modelling of Polymers

Table 5.2 Group- Group (at 298 K) Mi (g/mol) Vw (cm3/mol) bi (-)

contribution parameters
-CH2- 14.03 10.23 1.33
-COO- 44.01 15.20 1.50
-aromat C6H4- 76.09 43.3 1.58
-aromat C6H5- 77.10 45.85 1.51
-CO- 28.01 11.7 1.40
-CH3 15.04 13.67 1.38
-CH< 13.02 6.8 1.30
-OH 17.01 8.0 1.52
>C¼C< 24.02 10.06 0.944
-H 1.01 3.44 0.40

Fig. 5.12 Densities calculated with Park and Paul’s group-contribution method (Park and Paul 1997) vs corresponding
experimental polymer densities. Drawn after data collected by Park and Paul 1997

Mrw 192:2
ρPET ð298Þ ¼ ¼ ¼ 1:33g=cm3 ð5:101Þ
Va 1:60∗ð2∗10:23 þ 2∗15:2 þ 43:3Þ

The predicted value was here close to the experimental density of PET. van Krevelen’s density
calculations are however often less precise than necessary, because errors are introduced both when
fitting Vw and when translating Vw to Va. Park and Paul (Park and Paul 1997) reduced the error
significantly (Fig. 5.12) by using a weighted sum for Va, with weights bi.
5.7 The Finite Element Method (FEM) 247

X
n
Va ¼ bi V w,i ð5:102Þ
i¼1

Weighted linear sums often give sufficiently good material property predictions, but even more
accurate forecasts can be achieved with non-linear equations, although the determination of the
group- contribution parameters then requires more care. For instance, for a polymer with density ρ,
the bulk modulus K can be parameterized with the Rau function UR (Van Krevelen and Te Nijenhuis
2009):
0P 1
U R,i 6
K ð U R Þ ¼ ρ@ A
i
ð5:103Þ
Mrw =ρ

Sometimes the property of interest depends on more than one material. For example, the diffusiv-
ity of a penetrant molecule in a polymer matrix depends both on the choice of penetrant and on the
polymer. Each combination of penetrant/polymer piece then results in a unique interaction parameter
gij, and the total number of parameters can easily become cumbersome. The well-known UNIFAC
model (Fredenslund et al. 1975) is a multicomponent group-contribution method.
To conclude, statistical methods, like group-contribution models, are computationally inexpensive
and can quickly give good predictions of material properties if reliable polymer databases are
available.

5.7 The Finite Element Method (FEM)

5.7.1 FEM: Introduction

The finite element method (FEM) is an important simulation technique for analysing macroscopic
phenomena like mechanical deformation, diffusion, heat transfer and fluid flow. Processes like these
can be described with differential equations, which are solved numerically. Complex geometries can
be handled by discretizing the computational domain into small, connected, non-overlapping
segments called finite elements. An example of a FEM model for calculating the stress distribution
in a tooth prototype subject to thermal stresses is presented in Fig. 5.13.
Since most polymer scientists prefer to use existing FEM packages such as COMSOL
Multiphysics rather than implementing their own codes for PDE solvers from scratch, a brief
introduction on how FEM is used in practice is presented before its mathematical description.

5.7.2 Applied FEM

When existing FEM programs like Ansys, COMSOL Multiphysics, Abaqus and Moldflow are used,
most of the numerical calculations are hidden so that the user can concentrate on defining his problem
in an optimal way. A typical FEM problem set-up involves the following steps: (1) define the
geometry; (2) define the materials, including their material parameters and mathematical material
models; (3) define the physics in terms of PDEs/ODEs with boundary and initial conditions; (4) define
248 5 Simulation and Modelling of Polymers

Fig. 5.13 FEM simulation of a tooth prototype subjected to stresses due to thermal expansion. The composite crown is
attached to a steel hub. (a) Tetrahedral mesh, (b) FEM solution showing the stress distribution, with red and blue
colours indicating high and low stress, respectively. Drawn after Nakamura et al. 2018

the mesh; (5) choose numerical settings and solve the problem; and (6) plot the solution and do post-
processing. A brief summary of each step is given below:
Geometries of complex macroscopic structures like aeroplanes or teeth can be constructed with
computer-aided design (CAD) tools or by importing computer tomography images, whereas repre-
sentative geometries of composite structures can be constructed using Monte Carlo (MC) techniques.
It is advisable to avoid having extremely small objects and thin or narrow sections in the geometries,
since such features can make it impossible to solve the discrete FEM system. If a small section of a
large structure is of particular interest, this can be handled by coupling several geometries together,
resulting in a multiscale FEM model. Three-dimensional geometries are often used in FEM
modelling, but it is sometimes possible to explore symmetries which make it possible to simplify
the problem into one or two dimensions, resulting in much faster iterations and shorter computation
times. Modern FEM tools enable advanced features to be simulated, such as moving and rotating
objects, objects in contact with each other, deformed objects, polymeric mould flow, etc. The latter is
very useful for simulating the injection moulding of polymers into intricately shaped forms.
(Fig. 5.14)
Materials are defined through material parameters and material models. A material parameter can
either take a single value or depend on one or more variables. For instance, the density ρ can depend
on the temperature T and pressure p, i.e. ρ(T,p). In this case, densities covering a wide range of
temperatures and pressures are tabulated, and values between the known data points can be obtained
by fitting the data with a smooth interpolation function, such as a cubic spline curve. Material models
define the constituent equations which are used to describe the material. For a linear elastic material
with elastic modulus E, the stress σ is linearly dependent on the strain ε through σ ¼ Eε, whereas for a
rubber, the stress can be described by the Mooney–Rivlin equation with constants C1 and C2 and
elongation λ ¼ 1 + ε:
  
2C 1
σ 11 ¼ 2C1 þ 2 λ2  ð5:104Þ
λ λ

The stress–strain relationship for a viscoelastic material, including most polymers, can be
described by the Zener model:
5.7 The Finite Element Method (FEM) 249

Fig. 5.14 FEM simulation of an electric tree in a block of LDPE. A high electric potential is assigned to the top surface,
the bottom surface is grounded, and the remaining four sides are insulated. The red lines are electrical field lines and the
colour on the electrical tree correspond to energy dissipation strength. Drawn after Pallon et al. 2017

 
η dε dσ
σ ¼ E1 ε þ ð E1 þ E2 Þ  E1 ð5:105Þ
E2 dt dt

where E1 and E2 are elastic moduli and η is a viscosity. The Zener model generalizes the simpler
Maxwell and Voigt–Kelvin models, which describe the viscoelastic behaviour of polymers by
combining the equations for feather and dashpot elements. There are however many other material
models for elasticity and other material properties.
Physics in FEM is typically defined through a combination of PDEs or ODEs, boundary conditions
and initial conditions. One example is the heat equation, where a typical initial condition is that the
object has a constant temperature and a typical boundary condition is that one surface section of the
object is heated while the remaining boundaries are insulated from heat, i.e. zero heat fluxes are
specified through them. Much more advanced possibilities however exist; it is, for instance, possible
to couple together conductive heating (modelled with the heat equation), convective heating
(governed by fluid flow models) and radiation-driven heating (modelled with ray-tracing algorithms)
in the same model. The technique of combining several kinds of physics in the same FEM model is
called multiphysics, and this is a very powerful technique.
Meshing strategies in FEM are strongly related to the topological features of the geometry and to
the type of physics. In order to obtain a solution with a high accuracy, a mesh with sufficiently small
mesh elements (locally) is required. A mesh adapted to the geometry is usually applied, in which the
fine details and crucial parts are discretised with smaller mesh elements than other regions. Most FEM
programs include mesh-generation algorithms for creating basic adaptive meshes, which can be
further optimized and fine-tuned by the user. Prisms, pyramids, 3D tetrahedrals and 3D hexahedrals
can be uses as mesh elements in 3D, whereas triangles, rectangles and rhomboidal tetrahedral
250 5 Simulation and Modelling of Polymers

Fig. 5.15 Meshes in 2D: (a) a triangular mesh, (b) a rectangular, structured (quad) mesh, (c) an adaptive triangular
mesh and (d) a triangular mesh with boundary elements

elements can be used in two-dimensional geometries. Boundary meshes, consisting of ca. 8–12 thin
and parallel mesh layers, can preferably be assigned to boundaries where large gradients are expected,
in order to achieve a more precise solution. (Fig. 5.15) Some types of PDEs, particularly those in fluid
flow equations like the Navier–Stokes equations, require a more narrow size distribution than the
average PDE which can have large differences in mesh size. A standard strategy for determining
whether or not a specific mesh is sufficiently dense is to solve the same problem several times with
gradually smaller mesh elements and check that the solution does not change up to a certain tolerance.
An alternative strategy is to examine the mesh-size histogram and check that it looks appropriate.
Solving a PDE with known boundary and initial conditions with FEM corresponds to discretizing
and eventually integrating the PDE with a FEM strategy utilizing Galerkin, Rayleigh–Ritz or
collocation techniques and then solving the resulting matrix equation Ax ¼ y numerically. A is the
coefficient matrix, x the vector of unknowns and y the load vector. Commercial FEM solvers often
have predefined settings that work sufficiently well, although they can typically be further optimized
to suit the particular problem at hand. Since the size of the matrix A is proportional to the number of
mesh elements, the time and memory requirements to solve a PDE almost always increase with
5.7 The Finite Element Method (FEM) 251

Fig. 5.16 FEM solution of the continuity equation on a periodic section of a short-fibre composite, where the colours
show the electric potential. The conductivity of the fibre particles is much higher than that of the surrounding polymer
matrix. Drawn after Nilsson et al. 2016

increasing number of mesh elements, although there are exceptions. If high accuracy is required, the
convergence of the solution is sometimes more rapid with a somewhat denser and more accurate
mesh. This is because it is difficult for a coarse mesh to provide a sufficiently precise solution.
Plotting and post-processing the solution to the problem are handled once the PDEs are solved.
The variables of interest in the solution, for instance, the temperature T when solving the heat
equation or the concentration c when solving the diffusion equation, can be plotted using a wide
range of plot types, e.g. as surface plots with colours corresponding to the numerical values directly in
the geometry. Complex phenomena, like large deformations and moving objects, can also be
visualized. The variables can also be analysed by, for instance, integrating them over the domain
and storing the averaged solution as a function of, e.g. time in a table. Derived quantities can also be
achieved from the post-processing. For instance, assume that the continuity equation:
∇  ðσ i ∇VÞ ¼ 0 ð5:106Þ

with electrical potential V is solved in a geometry resembling a cubic section of a two-component

composite and that the (local) electrical conductivities σ i are known for both materials. Then the
effective conductivity σ of the homogenized composite can be derived from the solution of the
differential equation (Fig. 5.16).
Example: Thermally Induced Stresses in a Tooth Prototype The aim is to compute the thermally
induced stresses in a manufactured composite tooth-crown prototype attached to a steel hub when the
tooth is subjected to rapid cooling, mimicking a person eating ice cream. To solve this problem, the
first step is to construct a realistic geometry which describes the geometry of the tooth. This is done by
importing a 3D CAD structure (e.g. a .stp file) from a computer tomography (CT) image. The original
252 5 Simulation and Modelling of Polymers

structure is simplified by removing small non-influential irregularities in order to facilitate the

computation. Since the expected deformations are small, the simplest mechanical model, Hooke’s
law (σ ¼ Eε), is assigned to both the composite crown and the steel hub, but with different values of
the elastic modulus E. Temperature-dependent values of the elastic modulus, the thermal conductivity
and other variables are inserted as tables in the model. Natural cubic splines are used to interpolate
between the data points and to extrapolate outside the range of available data points. Two PDEs are
solved simultaneously, the heat equation and the mechanical equation, which describe strains and
stresses as functions of the applied force. As an initial condition, the temperature of the tooth is set to
37  C, and as a boundary condition, the surface temperature of the tooth is set to 5  C. The geometry
is then meshed using a fine adaptive tetrahedral mesh (Fig. 5.13a) (Nakamura et al. 2018). Some
regions in the mesh have to be repaired and handled manually due to the fine details of the geometry.
Finally, the PDEs are solved as functions of time, and a steady-state solution is obtained, showing the
regions with the largest thermally induced stresses (Fig. 5.13b). These types of simulations can be
used both for choosing and evaluating materials and for determining suitable geometries for
manufactured tools.

5.7.3 FEM: Mathematical Background

On a continuum scale, when the influence of individual molecules can be neglected, homogenized
material models can be utilized, and physical properties of interest, such as stress, strain, temperature,
concentration and velocity, can then be calculated by solving appropriate differential equations
governing the laws of physics. In mathematical terms, a differential equation is an equation which
relates one or more functions to their derivatives. Ordinary differential equations (ODEs) contain a
single independent variable, whereas partial differential equations (PDEs) contain two or more
independent variables. For instance, Newton’s second law of motion states that the applied force
F(x(t)) on an object at position x(t) at time t equals the objects mass m multiplied by its acceleration
d2x/dt2:

d 2 xðtÞ
m ¼ FðxðtÞÞ ð5:107Þ
dt2

This is an ODE, because it has only one independent variable: the time t. In contrast, the heat
equation, which describes the change in temperature T with time when a material with thermal
diffusivity α is heated, is a PDE, because it has two independent variables (t and x):
2
∂T ðx, tÞ ∂ T ðx, tÞ
¼α ð5:108Þ
∂t ∂x2

PDEs can describe most continuum scale physics in materials science. Examples include the
diffusion equation, the heat equation, the wave equation, Maxwell’s equations (electrostatics and
electrodynamics), the Navier–Stokes equations (fluid dynamics), the Schr€odinger equation (quantum
mechanics) and equations for mechanical stresses and deformations. It is therefore extremely import
to have efficient methods for solving PDEs. The finite element method (FEM), which is also
denoted finite element analysis (FEA), is a versatile and commonly used numerical technique for
solving PDEs. Two other major techniques are the finite difference method (FDM) and the finite
volume method (FVM). The FDM is efficient and easy to implement, but the PDE needs to be
discretized on a topologically rectangular mesh that makes this method not suited for complex
geometries. In contrast, unstructured meshes can be used in FEM and FVM to discretize the geometry
5.7 The Finite Element Method (FEM) 253

with small, variable-sized elements like tetrahedral or hexahedral blocks. This feature makes the
methods efficient in handling complex geometries. FVMs were specifically developed for solving
hyperbolic PDEs, in particular many important fluid dynamics problems, but they are not well suited
for solving parabolic PDEs, like the heat equation, or elliptical PDEs, like the Laplace equation.
FEMs can handle all types of PDEs and boundary value ODEs, and they are therefore the focus of this
chapter. However, every simulation technique has both strengths and weaknesses, and FEMs have
comparatively poor treatment of boundary fluxes and a non-conservative nature. Therefore, commer-
cial FEM programs often switch seamlessly between FEM, FVM and other macroscopic techniques,
like ray-tracing algorithms, depending on the type of computational problem.
In order to describe different FEM strategies and to compare them with FDM, Poisson’s equation
in one dimension is used as an example. In one spatial dimension, it can, for example, describe the
vertical displacement u(x) of a thin horizontal rod, attached at its endpoints, when subject to gravity.
The displacement is zero at the fixed boundaries x ¼ 0 and x ¼ L, and the second derivative of u(x) is a
presumably known function f(x), which depends on the horizontal position x. This boundary value
ODE can be written:
9
2
∂ u =
¼ f ðx Þ
∂x2 ð5:109Þ
;
uð 0Þ ¼ uð L Þ ¼ 0

Finite difference methods (FDM) solve differential equations like Eq. 5.109 by first dividing the
1D domain into n intervals. These are often equally sized so that the domain becomes discretized with
n + 1 equidistant points at positions xi ¼ 0..L. If the length of each interval is Δxi ¼ xi + 1-xi, the first-
and second-order spatial derivatives can be approximated with three-point difference stencils using
Taylor series expansion as:
∂ui uiþ1  ui1
¼ ð5:110Þ
∂xi 2Δxi

2
∂ ui uiþ1  2ui þ ui1
¼ ð5:111Þ
∂x2i ðΔxi Þ2

This strategy makes it possible to replace Eq. 5.109 with:

9
uiþ1  2ui þ ui1 =
¼ f ð x i Þ, i ¼ 1::n  1
ðΔxÞ2 ð5:112Þ
;
ui¼0 ¼ ui¼n ¼ 0

Equation 5.112 contains two boundary values and n-1 unknowns in the domain and yields linear
equations, which can be combined into a tridiagonal linear system of equations of the form Au ¼ f.
Here, A is a tridiagonal (n-1)*(n-1) matrix, f is a known vector, and u is an unknown vector with n-1
elements, which are computed by solving the linear system of equations numerically. With the FDM
approach, the solution is thus computed at the nodes of the computational mesh. FDM requires the
differential equation to be on a strong form, like Eq. 5.109, such that the derivatives can be directly
replaced by their discrete equivalents using difference stencils. In contrast, FEM requires the
equations to be on a weak form, obtained by integrating the original PDE. The weak form allows
for approximations with special ‘ansatz functions’, which can be computationally useful, and it
requires a lower degree of regularity of the solution u than the strong form. Thus, in the present
example, u must be twice differentiable in the strong form, while only once in the weak form.
254 5 Simulation and Modelling of Polymers

Fig. 5.17 (a) Non-equidistant hat functions. (b) Equidistant Bell-splines

Finite element methods (FEMs) replace the unknown variables in differential equations such
as Eq. 5.109 with linear combinations some of (predefined) basis functions ψ(x) with unknown
weights ci:
X
n
uðxÞ ueðxÞ ¼ ci ψ i ðxÞ ð5:113Þ
i¼0

The basis functions ψ i(x) are required to be smooth (with non-zero second derivatives) and
vanishing at the boundaries. They are often chosen as hat functions or B-splines in 1D, where the
geometry of the problem is a line from 0 to L. The line is meshed with n + 1 grid points xi, where
x0 ¼ 0 and xn ¼ L, and the grid points need not to be equally distributed in space. In each interval [xi-1,
xi + 1], a basis function ψ i(x), which is zero outside the interval and has a finite degree of freedom, is
defined. If hat functions Hi are chosen, ψ i(x) ¼ Hi(x) is described by (Fig. 5.17a):
8
< ðx  xi1 Þ=ðxi  xi1 Þ, fxi1  x  xi g
>
H i ðxÞ ¼ ðxiþ1  xÞ=ðxiþ1  xi Þ, fxi  x  xiþ1 g ð5:114Þ
>
:
0, felsewhereg

When smoother basis functions with finite second derivatives are needed, third-order polynomials
can be used, for instance, basis-splines (B-splines). For one-dimensional problems with equidistant
spacing Δx ¼ xi + 1-xi, the i-th third-order B-spline ψ i(x) ¼ Bi(x) is described by (Fig. 5.17b):
8
>
> s3 =6, fxi2  x  xi1 , s ¼ ðx  xi2 Þ=Δxg
>
>
>
> 1=6 þ ð s þ s 2
 s 3
Þ=2, f i1
x  x  x , s ¼ ðx  xi1 Þ=Δxg
< i
Bi ðxÞ ¼ 1=6 þ ðs þ s  s Þ=2, fxi  x  xiþ1 ,
2 3
s ¼ ðxiþ1  xÞ=Δxg ð5:115Þ
>
>
>
> s =6,
3
fxiþ1  x  xiþ2 , s ¼ ðxiþ1  xÞ=Δxg
>
>
:
0, felsewhereg
5.7 The Finite Element Method (FEM) 255

By inserting the approximate finite element solution u into the original PDE, the residual R can be
obtained as the difference between the right-hand side and the left-hand side of the equation. For
instance, for Eq. 5.109, the residual becomes:
2
∂ ue
Rð x Þ ¼  f ðx Þ ð5:116Þ
∂x2

The FEM procedure aims at minimizing the residual to obtain a numerical solution of the original
solution. From the mathematical point of view, FEM can be implemented using three main
approaches; Galerkin, Rayleigh–Ritz and collocation. With Galerkin’s method, the residual is
assumed to be orthogonal to the basis function ψ. With Rayleigh–Ritz’s method, the energy is
minimized, and with the Collocation method, the residual is forced to be zero at some specific points.
Other related methods also exist, for instance, spectral and pseudo-spectral methods. The develop-
ment of FEM (Stein 2014) started with pioneering work by Leibniz (1697), Schellbach (1851),
Rayleigh (1877), Ritz (1909), Galerkin (1915) and Courant (1943), but with the development of
modern computers, progress has accelerated tremendously during the last decades.
The Galerkin FEM strategy, and other weighted residual techniques, uses the principle of virtual
work to transfer the differential equations to a form which is suitable for FEM calculations. At
equilibrium, the virtual work vanishes for all possible virtual displacements. As a starting point of the
Galerkin strategy, the two arbitrary functions g(x) and h(x) are defined as being orthogonal on the
interval a-b if:
ðb
gðxÞhðxÞdx ¼ 0 ð5:117Þ
a

Hence, the residual R(x) is required to be zero in a weighted sense, meaning that R(x) must be
orthogonal to all basis functions ψ i(x), i ¼ 0,1,2..n, that are vanishing at the boundaries:
ðb
RðxÞψ i ðxÞdx ¼ 0, i ¼ 0, 1, 2::n ð5:118Þ
a

If this strategy is applied to the boundary value problem Eq. 5.109, it can be formulated as:
ðL !
2
∂ ueðxÞ
 f ðxÞ ψ i ðxÞdx ¼ 0 ð5:119Þ
∂x2
0

By integrating by parts and utilizing ũ(0) ¼ ũ(L ) ¼ 0, this can be simplified to:
ðL 2 ðL L ðL ðL
∂ ue u ∂ψ i
∂e ∂e
u u ∂ψ i
∂e
ψ dx ¼ dx  ψ ¼ dx ¼ f ðxÞψ i dx ð5:120Þ
∂x2 i ∂x ∂x ∂x i 0 ∂x ∂x
0 0 0 0

From this, the weak (variational) form of Eq. 5.109 can be formulated: Find ũ(x), with
ũ(0) ¼ ũ(L ) ¼ 0, such that for all admissible ψ(x):
256 5 Simulation and Modelling of Polymers

ðL ðL
u ∂ψ i
∂e
dx ¼ f ðxÞψ i dx ð5:121Þ
∂x ∂x
0 0

Assuming hat-shaped basis functions ψ i(x) (Eqs. 5.114) inserted in Eq. 5.121, the equations for
elements i ¼ 1, 2, ... n-1 become:
9
ÐL ∂e
ðL Xn   X >
>
u ∂ψ i c j ∂ψ j ∂ψ i n
>
dx ¼   c j aij ¼ >
>
=
0 ∂x ∂x ∂x ∂x Xn
0
j¼0 j¼0
) c j aij ¼ f i ð5:122Þ
>
>
ÐL >
> j¼0
¼ f ðxÞψ i dx  f i >
;
0

Since the boundary values c0 ¼ cn ¼ 0 are known, the remaining n-1 unknown parameters cj are
found by solving the n-1 equations above. Together these form a (n-1)*(n-1) linear system of
equations:
Ac ¼ f ð5:123Þ

The matrix and vector elements are given by:

ðL
∂ψ i ðxÞ ∂ψ j ðxÞ
Aij ¼ dx ð5:124Þ
∂x ∂x
0

ðL
f i ¼ f ðxÞψ i dx ð5:125Þ
0

Once the elements in the matrix A and the vector f have been determined, the system of equations
can be solved numerically with respect to c, for instance, with the command c ¼ A\f in MATLAB. For
first-order elements utilizing hat functions, the equations are significantly simplified by the fact that
all elements ψ i, as well as their derivatives dψ i/dx, are zero outside the interval [xi-1, xi+1]. Hence, in
1D, the linear system of equations becomes tridiagonal, as in the FDM method. Knowing the basis
function and its derivative, the terms in i’th row of the equation system (except for the boundary
elements, i.e. the first and the last rows in the matrix) can be obtained as:
ai,i1 ci1 þ aii ci þ ai,iþ1 ciþ1 ¼ f i ð5:126Þ

xð
iþ1  2
∂ψ i 1 1
aii ¼ dx ¼ þ ð5:127Þ
∂x ðxi  xi1 Þ ðxiþ1  xi Þ
xi1

xð
iþ1  
∂ψ i ∂ψ i1 1
ai,i1 ¼ dx ¼  ð5:128Þ
∂x ∂x ðxi  xi1 Þ
xi1
5.7 The Finite Element Method (FEM) 257

xð
iþ1  
∂ψ i ∂ψ iþ1 1
ai,iþ1 ¼ dx ¼  ð5:129Þ
∂x ∂x ðxiþ1  xi Þ
xi1

ðxi xð
iþ1
! 1 1
fi ¼ f ðxÞðx  xi1 Þdx þ f ðxÞðxiþ1  xÞdx ð5:130Þ
ðxi  xi1 Þ ðxiþ1  xi Þ
xi1 xi

If the grid points are equidistantly spaced with Δx ¼ xi+1-xi, and if Dirichlet-type boundary
conditions c0 ¼ cn ¼ 0 are assumed, which affect the first and last rows of the linear system of
equations, the tridiagonal system becomes:
2 32 3 2 3
2 1 0  0 c1 f1
6 1 2 1    0 76 7 6 7
6 7 6 c2 7 6 f 2 7
1 66 .. 7 6 7 6 7
⋮ ⋮ .    7 6 c3 7 ¼ 6 f 3 7 ð5:131Þ
Δx 66 0 7 6 7 6 7
76 7 6 7
4 0 0 1 2 1 54 ⋮ 5 4 ⋮ 5
0 0 0 1 2 cn1 f n1

The Reighley–Ritz FEM strategy, and other variational methods, utilizes the principle of
minimum energy instead of the principle of virtual work. The minimum energy principle assumes
that at equilibrium, the energy of a system has its minimum value. Thus, if the ‘energy’ P depends on
a single variable x, a single function u and its first derivative du/dx ¼ ux, then the weak form is defined
as the integral of a function F of x, u and ux:
ð
PðuÞ ¼ Fðx, u, ux Þdx ð5:132Þ

This equation corresponds to the (strong form) Euler–Lagrange formulation:

∂F d ∂F
 ¼0 ð5:133Þ
∂u dx ∂ux

If F contains more variables than x or higher-order derivatives, additional terms are to be added to
the Euler–Lagrange equation. If the PDE contains more functions than u, each additional function
results in an additional Euler–Lagrange equation. The function F is constructed by identifying terms
in the relevant PDE. For Eq. 5.109 this transformation becomes:
9
∂F >
>
¼f =
∂u u2 ð x Þ
) Fðx, u, ux Þ ¼ x  f ðxÞuðxÞ ð5:134Þ
d ∂F ∂ u >
2 2
¼ >
;
dx ∂ux ∂x2

With this definition of the function F, and with u written as a linear combination of chosen basis
functions, the energy P(u) can be written:
0 !2 1
ðL X n Xn
1 ∂ψ i ðxÞ
Pð u Þ ¼ @ c  f ðxÞ ci ψ i ðxÞAdx ð5:135Þ
2 i¼0 i ∂x i¼0
0
258 5 Simulation and Modelling of Polymers

The minimum energy is achieved by minimizing P(u) with respect to all constants ci, i.e. the
derivatives with respect to each ci should be zero:

ðL ! ! ðL
∂PðuÞ X
n
∂ψ j ðxÞ ∂ψ i ðxÞ
¼0¼ cj dx  f ðxÞψ i ðxÞdx )
∂ci j¼0
∂x ∂x
0 0 ð5:136Þ
X
n
c j aij ¼ f i ) Ac ¼ f
j¼0

This results in exactly the same expression as the Galerkin method. The Rayleigh–Ritz and the
Galerkin methods are often considered as being equivalent, but there are exceptions, for instance,
when energy loss (e.g. due to friction) affect the system. Overall, Galerkin’s method is more general
and easier to implement.
The collocation FEM strategy uses delta Dirac functions δ(x) as basis function in Eq. 5.119. The
delta Dirac function is defined as being zero everywhere in the domain except close to specific points
xi, where δ(x) is very high. If δ(x) is applied on n + 1 different points xi, the Galerkin method can be
transformed into a collocation method, providing the exact solution at these particular points. When
ψ(x) ¼ δ(x) is inserted in Eq. 5.119, the expression becomes (at x ¼ xi):
ðL !
2 2
∂ ueðxÞ ∂ ueðxi Þ
 f ð xÞ δ ðx  x i Þdx ¼  f ðx i Þ ¼ 0 ð5:137Þ
∂x2 ∂x2
0

Since second-order derivatives act on the approximate function ũ, smoother functions than hat
functions must be used. If equidistant one-dimensional B-splines are employed, the equation
becomes:
! !
∂
2 Xn Xn 2
∂ Bi ðxi Þ
c i B i ðx i Þ ¼ ci ¼ f ðxi Þ ð5:138Þ
∂x2 j¼0 j¼0
∂x2

Twofold derivation of the equidistant B-spline (Eq. 5.115), with respect to x, yields:
2
∂ Bi ð x Þ 2
¼
∂x2 ðΔxÞ2
2 2
ð5:139Þ
∂ Bi1 ðxÞ ∂ Biþ1 ðxÞ 1
¼ ¼
∂x 2 ∂x 2
ðΔxÞ2

Further, insertion of Eq. 5.139 into 5.138 gives for the i-th interval:

ðci1  2ci þ ci1 Þ=ðΔxÞ2 ¼ f i ð5:140Þ

Once the boundary conditions are considered, the unknown constants ci are determined by solving
the resulting (sparse) tridiagonal system of equations Ac ¼ f, similarly as in the previous methods.
The collocation method can be quite powerful since it does not require the explicit calculation of
integrals. The choice of ψ(x) can be influenced by the choice of quadrature points. If the latter are
chosen to coincide with the nodal points, cubic splines are typically used. In case the quadrature
points are chosen (more effectively) to be positioned at two Gauss points inside the interval, cubic
5.8 Summary 259

Hermite polynomials are more effective. It can also be noted that many of the well-known
Runge–Kutta formulae for ODEs are derived by considering the collocation approach.
Spectral methods, pseudo-spectral methods and the spectral element method form a group of
methods closely related to FEM. Spectral methods (Gottlieb and Orszag 1977) use essentially
weighted sums of basis functions as approximate solutions to differential equations, just like FEM.
Classical spectral methods use sine functions as basis functions, which enable computationally
efficient fast Fourier transform (FFT) algorithms to be utilized in the computations. Sine functions
are however non-zero over almost all the computational domain, whereas FEM basis functions are
non-zero only locally (Cook 2007). In the case of problems with smooth solutions and simple
geometries, classical spectral methods are faster and more efficient than FEM, but they are not well
suited for modelling complex geometries or problems with non-smooth solutions, like shock propa-
gation phenomena. The speed of the computations can be increased by multiplying the standard basis
function with a cleverly chosen complementary basis, resulting in a so-called pseudo-spectral method
(Orszag 1972). Both spectral and pseudo-spectral methods are collocation methods. One special
spectral method, the spectral element method (Maday and Patera 1989), combines features from both
spectral and finite element methods. It often uses piecewise polynomials of high order, like Legendre
or Chebyshev polynomials, as basis functions. This is the essential ingredient of pseudo-spectral and
spectral element methods, because smooth functions can be very well approximated by short sums of
Chebyshev polynomials. The spectral element method can be useful for modelling seismic wave
propagation in fluid dynamics, but it is worse than FEM for modelling complex geometries.
To conclude, FEM and related techniques, such as spectral methods, are useful for simulating
many macroscopic phenomena, not at least within polymer physics, polymer engineering and
material science. Further reading on differential equations and FEM is available (Brenner and Scott
2008; Cook 2007; Deuflhard and Bornemann 2002; Eriksson 2002; Evans 2010; Hairer et al. 1993;
Hairer and Wanner 1996; Heath 1997; Larson and Bengzon 2013; LeVeque 2002; Quateroni et al.
2000; Reddy 1993; Strang 1986; Thomas 1999; Versteeg and Malalasekera 2007).

5.8 Summary

Computer simulations in polymer science can be used to predict, optimize and explain the material
properties of polymers and polymer composites.
Many simulation methods exist, and the choice of simulation technique depends largely on the
available computer resources and on the time and length scales of the problem to be solved.
Multiscale modelling can be used to cover wider time and length scales than otherwise possible.
The most commonly used simulation techniques in polymer science are briefly described in this
chapter, but several others exist.
Quantum chemistry (QC) methods are slow but accurate molecular techniques which explicitly
handle electron effects. Molecular orbital (MO) methods, which comprise ab initio and semi-empiri-
cal methods, and density function theory (DFT) are the primary classes of QC. In all molecular
simulation techniques, realistic polymer chain lengths are hard to achieve due to computational
limitations.
In molecular modelling (MM) methods, electrons are not included explicitly, but they are
implicitly introduced through force fields, which relate molecular arrangement to potential energy.
Time-dependent molecular dynamics (MD) and time-independent Monte Carlo (MC) are the main
MM methods. Force fields can be all-atom, united-atom or coarse-grained. In the latter case, nearby
atoms are grouped into larger superatoms, which reduce the computational costs but may eventually
introduce physical artefacts.
260 5 Simulation and Modelling of Polymers

Continuum methods like the finite element method (FEM), the finite difference method (FDM),
the finite volume method (FVM) and spectral methods treat materials as continuous structures rather
than collections of molecules. Such methods are used to find solutions to partial differential equations
such as the diffusion equation or the heat equation.
Mesoscale methods are techniques for time and length scales in intermediate between molecular
and continuum methods. Examples include dissipative particle dynamics (DPD), which can be used
to analyse the mixing of molten polymer blends, and effective media modelling, which can be used to
predict composite properties. The former technique is closely related to MD, whereas the latter can be
implemented with FEM.
Statistical methods such as quantitative structure–activity relationship (QSAR) methods and
group-contribution (GC) methods predict the material properties of unknown materials using large
material databases containing material properties and structural data for known materials. Such
methods are computationally very rapid, but they are limited by the quality of the database used as
input.

5.9 Exercises

5.1 Describe one strength and one weakness for each of QC, MD, MC, mesoscale models, FEM and
statistical methods.
5.2 Give one polymer application example for each of QC, MD, MC, mesoscale models, FEM and
statistical methods.
5.3 When simulating macromolecules with molecular simulation methods such as MD, which
difficulties arise due the extreme length of the chains? How can it be handled?
5.4 Describe pros and cons for using coarse-grained force fields in molecular mechanics
simulations.
5.5 Use Park and Paul’s group-contribution method to estimate the density of polystyrene (23  C).
The experimental density is 1.05 g/cm3.
5.6 Write down the Hückel determinant for benzene (C6H6). The Hückel orbital energies for
benzene are E1 ¼ α + 2β, E2 ¼ α + β, E3 ¼ α + β, E4 ¼ α-β, E5 ¼ α-β and E6 ¼ α-β. Up to
two electrons can occupy each orbital, starting with the lowest energy (i.e. E1 since α < 0 and
β < 0). Compute the energy of benzene, and compare it with the corresponding energy for three
ethane molecules. Explain the difference.
Miniprojects (Without Printed Solutions).

5.7 Use the Monte Carlo Metropolis–Hastings algorithm to create a representative section of a
composite with 50 vol.% randomly positioned spherical particles. Hint: add randomly the
suitable fraction of spherical particles to the cubical domain without considering eventual
overlap. Then set a high energy for overlapping spheres, and run the MCMH algorithm.
5.8 Compute an approximate value of π with a simple MC program.
5.9 Implement a simple MD program for solitary atoms with only the Lennard–Jones (LJ) force
field and a NVT ensemble. Run a simulation until the temperature and energy has stabilised, and
plot the geometry.
5.10 Update your simple MD code such that it works also for DPD for two species.
5.11 If you have access to a FEM modelling program: Create a simple but non-trivial 3D geometry
(like a dog bone structure), assign material data and a suitable material model for a polymer of
your choice to the structure, mesh the geometry, apply forces on suitable positions, and analyse
the deformations, stresses and strains.
References 261

5.12 If you have access to an advanced MD program, create a polymeric structure with small
penetrant molecules, equilibrate the system, run a production run while keeping track of the
atom positions, and compute the diffusivity from the Einstein equation.
5.13 If you have access to a QC program, create a small oligomer molecule, compute the electron
densities, and determine where it would be easiest for a nearby OH group to attach.

References

Alder, B. J., & Wainwright, T. E. (1959). Studies in molecular dynamics. I. General method. The Journal of Chemical
Physics, 31(2), 459–466.
Alexandrowicz, Z. (1998). Simulation of polymers with rebound selection. The Journal of Chemical Physics, 109(13),
5622–5626.
Allen, M., & Tildesley, D. (1989). Computer Simulation of Liquids. Oxford: Oxford University Press.
Andersen, H. C. (1980). Molecular dynamics simulations at constant pressure and/or temperature. The Journal of
Chemical Physics, 72(4), 2384–2393.
Andersen, M., Panosetti, C., & Reuter, K. (2019). A practical guide to surface kinetic Monte Carlo simulations.
Frontiers in Chemistry, 7, 202.
Auhl, R., Everaers, R., Grest, G. S., Kremer, K., & Plimpton, S. J. (2003). Equilibration of long chain polymer melts in
computer simulations. The Journal of Chemical Physics, 119(24), 12718–12728.
Ayers, P., Yang, W., & Bartolotti, L. (2009). Fukui function. In P. K. Chattaraj (Ed.), Chemical reactivity theory: A
density functional view (pp. 255–268). Boca Raton: CRC Press.
Becke, A. D. (1988). Density-functional exchange-energy approximation with correct asymptotic behavior. Physical
Review A General Physics, 38(6), 3098.
Bingham, R. C., Dewar, M. J., & Lo, D. H. (1975). Ground states of molecules. XXVI. MINDO/3 calculations for
hydrocarbons. Journal of the American Chemical Society, 97(6), 1294–1301.
Blumers, A. L., Tang, Y.-H., Li, Z., Li, X., & Karniadakis, G. E. (2017). GPU-accelerated red blood cells simulations
with transport dissipative particle dynamics. Computer Physics Communications, 217, 171–179.
Bock, F. E., Aydin, R. C., Cyron, C. J., Huber, N., Kalidindi, S. R., & Klusemann, B. (2019). A review of the application
of machine learning and data mining approaches in continuum materials mechanics. Frontiers in Materials, 6, 110.
Bondi, A. A. (1968). Physical properties of molecular crystals liquids, and glasses. New York: Wiley.
Born, M., & Oppenheimer, R. (1927). Zur quantentheorie der molekeln. Annalen der Physik, 389(20), 457–484.
Brenner, S., & Scott, L. (2008). The mathematical theory of finite element methods texts in applied mathematics. Berlin:
Springer.
Buckingham, R. A. (1938). The classical equation of state of gaseous helium, neon and argon. Proceedings of the Royal
Society of London. Series A: Mathematical and Physical Sciences, 168(933), 264–283.
Chatterjee, A., & Vlachos, D. G. (2007). An overview of spatial microscopic and accelerated kinetic Monte Carlo
methods. Journal of Computer-Aided Materials Design, 14(2), 253–308.
Cook, R. D. (2007). Concepts and applications of finite element analysis. New York: John Wiley & Sons.
Coulomb, C. A. (1785). Second mémoire sur l’électricité et le magnétisme. Histoire de l’Académie Royale des Sciences,
579, 578–611.
Courant, R. (1943). Variational methods for the solution of problems of equilibrium and vibrations. Bulletin of the
American Mathematical Society, 49, 1–23. https://doi.org/10.1090/S0002-9904-1943-07818-4.
Cramer, C. J. (2004). Essentials of computational chemistry: theories and models (2nd ed.). Chichester: John Wiley &
Sons.
Deuflhard, P., & Bornemann, F. (2002). Scientific computing with ordinary differential equations. New York: Springer
Science & Business Media.
Dijkstra, M. (1997). Confined thin films of linear and branched alkanes. The Journal of Chemical Physics, 107(8),
3277–3288.
Einstein, A. (1905). On the motion of small particles suspended in liquids at rest required by the molecular-kinetic
theory of heat. Annalen der Physik, 17(549–560), 208.
Eriksson, G. (2002). Kompendium i till€ ampade numeriska metoder. Stockholm: KTH Department of Numerical
Analysis and Computer Science.
Evans, L. (2010). Partial differential equations (2nd ed.). Providence, R.I: American Mathematical Society.
Fermi, E., Pasta, J., & Ulam, S. (1955). Los alamos report la-1940. Fermi. Collected Papers, 2, 977–988.
Flory, P. J. (1989). Statistical mechanics of chain molecules (Vol. 5). Munich: Hanser Publishers.
262 5 Simulation and Modelling of Polymers

Fock, V. (1930). N€aherungsmethode zur L€osung des quantenmechanischen Mehrk€ orperproblems. Zeitschrift für
Physik, 61(1–2), 126–148.
Fredenslund, A., Jones, R. L., & Prausnitz, J. M. (1975). Group-contribution estimation of activity coefficients in
nonideal liquid mixtures. AICHE Journal, 21(6), 1086–1099.
Frenkel, D., & Smit, B. (2002). Understanding molecular simulation: from algorithms to applications (2nd ed.). San
Diego: Academic Press.
Galerkin, B. (1915). Rods and plates. Series occurring in various questions concerning the elastic equilibrium of rods
and plates. Eng. Bull(Vestnik Inzhenerov) 19:897–908 (in Russian), (English Translation: 863–18925, Clearing-
house Fed. Sci. Tech. Info. 11963
Gartner, T. E., & Jayaraman, A. (2019). Modeling and simulations of polymers: A roadmap. Macromolecules, 52(3),
755–786.
Gedde, U. W., & Hedenqvist, M. S. (2019). Fundamental polymer science (2nd ed.). Cham: Springer.
Gedde, U. W., & Hedenqvist, M. S. (2019a). Introduction to polymer science (Chap. 1). In Fundamental Polymer
Science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Conformations in Polymers (Chap. 2). In Fundamental Polymer Science.
Cham: Springer Nature Switzerland AG
Gibson, J., Goland, A. N., Milgram, M., & Vineyard, G. (1960). Dynamics of radiation damage. Physical Review, 120
(4), 1229.
Gottlieb, D., & Orszag, S. A. (1977). Numerical analysis of spectral methods: Theory and applications. Philadelphia:
SIAM. https://doi.org/10.1137/1.9781611970425.
Hairer, E., & Wanner, G. (1996). Solving ordinary differential equations II stiff and differential-algebraic problems
(2nd ed.). Berlin: Springer Berlin Heidelberg.
Hairer, E., Nørsett, S. P., & Wanner, G. (1993). Solving ordinary differential equations I. Nonstiff problems (2nd ed.).
Berlin: Springer Series in Computational Mathematics.
Hall, G. (1951). The molecular orbital theory of chemical valency VIII. A method of calculating ionization potentials.
Proceedings of the Royal Society of London. Series A: Mathematical and Physical Sciences, 205(1083), 541–552.
Hansen, C. (2007). Hansen solubility parameters: A user’s handbook (2nd ed.). Boca Raton: CRC/Taylor & Francis.
Hansen, J., & McDonald, I. (1986). Theory of simple liquids (2nd ed.). London: Academic Press.
Hartree, D. R. (1928). The wave mechanics of an atom with a non-coulomb central field. Part II. Some results and
discussion. In: Mathematical Proceedings of the Cambridge Philosophical Society, vol 1. Cambridge University
Press, pp 89–110.
Hastings, W. K. (1970). Monte Carlo sampling methods using Markov chains and their applications. Biometrika, 57(1),
97–109.
Heath, M. (1997). Scientific computing: An introductory survey (2nd ed.). New York: McFraw-Hill Companies.
Hehre, W. J., Stewart, R. F., & Pople, J. A. (1969). Self-consistent molecular-orbital methods. i. Use of gaussian
expansions of Slater-type atomic orbitals. The Journal of Chemical Physics, 51(6), 2657–2664.
Hoffmann, R. (1963). An extended Hückel theory. I. hydrocarbons. The Journal of Chemical Physics, 39(6),
1397–1412.
Hohenberg, P., & Kohn, W. (1964). Inhomogeneous electron gas. Physical Review, 136(3B), B864.
Hoogerbrugge, P., & Koelman, J. (1992). Simulating microscopic hydrodynamic phenomena with dissipative particle
dynamics. EPL (Europhysics Letters), 19(3), 155.
Hoover, W. G. (1985). Canonical dynamics: Equilibrium phase-space distributions. Physical Review A, General
Physics, 31(3), 1695.
Hückel, E. (1931). Quantentheoretische beitr€age zum benzolproblem. Zeitschrift für Physik, 70(3–4), 204–286.
in’t Veld, P. J., & Rutledge, G. C. (2003). Temperature-dependent elasticity of a semicrystalline interphase composed
of freely rotating chains. Macromolecules, 36(19), 7358–7365.
Kanazawa, T., Asahara, A., & Morita, H. (2019). Accelerating small-angle scattering experiments with simulation-
based machine learning. Journal of Physics: Materials, 3(1), 015001.
Karatrantos, A., Clarke, N., & Kr€oger, M. (2016). Modeling of polymer structure and conformations in polymer
nanocomposites from atomistic to mesoscale: A review. Polymer Reviews, 56(3), 385–428.
Karayiannis, N. C., Giannousaki, A. E., & Mavrantzas, V. G. (2003). An advanced Monte Carlo method for the
equilibration of model long-chain branched polymers with a well-defined molecular architecture: Detailed atomistic
simulation of an H-shaped polyethylene melt. The Journal of Chemical Physics, 118(6), 2451–2454.
Kim, K., & Jordan, K. (1994). Comparison of density functional and MP2 calculations on the water monomer and
dimer. Journal of Physical Chemistry, 98(40), 10089–10094.
Kmiecik, S., Gront, D., Kolinski, M., Wieteska, L., Dawid, A. E., & Kolinski, A. (2016). Coarse-grained protein models
and their applications. Chemical Reviews, 116(14), 7898–7936.
Kohn, W., & Sham, L. J. (1965). Self-consistent equations including exchange and correlation effects. Physical Review,
140(4A), A1133.
References 263

Kremer, K., & Binder, K. (1988). Monte Carlo simulation of lattice models for macromolecules. Computer Physics
Reports, 7(6), 259–310.
Kremer, K., & Grest, G. S. (1990). Dynamics of entangled linear polymer melts: A molecular-dynamics simulation. The
Journal of Chemical Physics, 92(8), 5057–5086.
Ladd, A. J. (1993). Short-time motion of colloidal particles: Numerical simulation via a fluctuating lattice-Boltzmann
equation. Physical Review Letters, 70(9), 1339.
Larson, M. G., & Bengzon, F. (2013). The finite element method: theory, implementation, and applications (Vol. 10).
Berlin Heidelberg: Springer Science & Business Media.
Leach, A. R. (2001). Molecular modelling: principles and applications (2nd ed.). Harlow: Pearson Education.
Lee, T. S., York, D. M., & Yang, W. (1996). Linear-scaling semiempirical quantum calculations for macromolecules.
The Journal of Chemical Physics, 105(7), 2744–2750.
Leibniz, G. (1697). Communicatio suae pariter, duarumque alienarum ad adendum sibi primum a Dn. Jo. Bernoullio,
deinde a Dn. Marchione Hospitalio communicatarum solutionum problematis curvae celerrimi descensus a Dn. Jo.
Bernoullio geometris publice propositi, una cum solutione sua problematis alterius ab eodem postea propositi. Acta
Eruditorum, 16, 201–205.
Lemarchand, C., Bousquet, D., Schnell, B., & Pineau, N. (2019). A parallel algorithm to produce long polymer chains in
molecular dynamics. The Journal of Chemical Physics, 150(22), 224902.
Lennard-Jones, J. E. (1924). On the determination of molecular fields. Proceedings of the Royal Society of London A,
106, 463–477.
Lennard-Jones, J. E. (1929). The electronic structure of some diatomic molecules. Transactions of the Faraday Society,
25, 668–686.
LeVeque, R. J. (2002). Finite volume methods for hyperbolic problems. Cambridge: Cambridge University Press.
Levitt, M., & Warshel, A. (1975). Computer simulation of protein folding. Nature, 253(5494), 694–698.
Liu, Y., Zhao, T., Ju, W., & Shi, S. (2017). Materials discovery and design using machine learning. Journal of
Materiomics, 3(3), 159–177.
Liu, Y., Niu, C., Wang, Z., Gan, Y., Zhu, Y., Sun, S., & Shen, T. (2020). Machine learning in materials genome
initiative: A review. Journal of Materials Science and Technology, 57, 113–122.
Lyubartsev, A. P. (2017). Inverse Monte Carlo methods. In G. Papoian (Ed.), Coarse-grained modeling of biomolecules
(pp. 29–54). Boca Raton: CRC Press.
Maday, Y., & Patera, A. T. (1989). Spectral element methods for the incompressible Navier-Stokes equations. In
A. Noor (Ed.), State-of-the-art surveys on computational mechanics (pp. 71–143). New York: ASME.
Malevanets, A., & Kapral, R. (2000). Solute molecular dynamics in a mesoscale solvent. The Journal of Chemical
Physics, 112(16), 7260–7269.
Martin, M. G., & Siepmann, J. I. (1999). Novel configurational-bias Monte Carlo method for branched molecules.
Transferable potentials for phase equilibria. 2. United-atom description of branched alkanes. The Journal of
Physical Chemistry B, 103(21), 4508–4517.
Masubuchi, Y., & Uneyama, T. (2018). Comparison among multi-chain models for entangled polymer dynamics. Soft
Matter, 14(29), 5986–5994.
Mattice, W. L., & Suter, U. W. (1994). Conformational theory of large molecules: the rotational isomeric state model in
macromolecular systems. New York: Wiley-Interscience.
McCammon, J. A., Gelin, B. R., & Karplus, M. (1977). Dynamics of folded proteins. Nature, 267(5612), 585–590.
Meirovitch, H. (1988). Statistical properties of the scanning simulation method for polymer chains. The Journal of
Chemical Physics, 89(4), 2514–2522.
Metropolis, N. (1987). The beginning of the Monte Carlo method. Los Alamos Science, 15(15 Special Issue, Stanislaw
Ulam 1909–1984), 125–130.
Metropolis, N., Rosenbluth, A. W., Rosenbluth, M. N., Teller, A. H., & Teller, E. (1953). Equation of state calculations
by fast computing machines. The Journal of Chemical Physics, 21(6), 1087–1092.
Moore, T. C., Iacovella, C. R., & McCabe, C. (2014). Derivation of coarse-grained potentials via multistate iterative
Boltzmann inversion. The Journal of Chemical Physics, 140(22), 06B606_601.
Morse, P. M. (1929). Diatomic molecules according to the wave mechanics. II. Vibrational levels. Physical Review, 34
(1), 57.
Mountain, R. D., & Thirumalai, D. (1994). Quantative measure of efficiency of Monte Carlo simulations. Physica A:
Statistical Mechanics and its Applications, 210(3–4), 453–460.
Moyassari, A., Unge, M., Hedenqvist, M. S., Gedde, U. W., & Nilsson, F. (2017). First-principle simulations of
electronic structure in semicrystalline polyethylene. The Journal of Chemical Physics, 146(20), 204901.
Müller-Plathe, F. (2002). Coarse-graining in polymer simulation: from the atomistic to the mesoscopic scale and back.
ChemPhysChem, 3(9), 754–769.
Nairn, J. (2003). NairnFEAMPM A Macintosh application. Salt Lake City: Department of materials science and
engineering, University of Utah.
264 5 Simulation and Modelling of Polymers

Nakamura, K., Ankyu, S., Nilsson, F., Kanno, T., Niwano, Y., von Steyern, P. V., & Örtengren, U. (2018). Critical
considerations on load-to-failure test for monolithic zirconia molar crowns. Journal of the Mechanical Behavior of
Biomedical Materials, 87, 180–189.
Nanda, D., & Jug, K. (1980). SINDO1. A semiempirical SCF MO method for molecular binding energy and geometry
I. Approximations and parametrization. Theoretica Chimica Acta, 57(2), 95–106.
Newton I (1687). Philosophiae naturalis principia mathematica. Reg Soc Praeses, London, 2, 1–4.
Nilsson, F., Gedde, U. W., & Hedenqvist, M. S. (2009). Penetrant diffusion in polyethylene spherulites assessed by a
novel off-lattice Monte-Carlo technique. European Polymer Journal, 45(12), 3409–3417.
Nilsson, F., Krueckel, J., Schubert, D. W., Chen, F., Unge, M., Gedde, U. W., & Hedenqvist, M. S. (2016). Simulating
the effective electric conductivity of polymer composites with high aspect ratio fillers. Composites Science and
Technology, 132, 16–23.
Nosé, S. (1984). A unified formulation of the constant temperature molecular dynamics methods. The Journal of
Chemical Physics, 81(1), 511–519.
Orszag, S. A. (1972). Comparison of pseudospectral and spectral approximation. Studies in Applied Mathematics, 51
(3), 253–259.
Páll, S., Abraham, M. J., Kutzner, C., Hess, B., & Lindahl, E. (2014). Tackling exascale software challenges in
molecular dynamics simulations with GROMACS. In International conference on exascale applications and
software (pp. 3–27). Cham: Springer.
Pallon, L., Hoang, A., Pourrahimi, A., Hedenqvist, M. S., Nilsson, F., Gubanski, S., Gedde, U., & Olsson, R. T. (2016).
The impact of MgO nanoparticle interface in ultra-insulating polyethylene nanocomposites for high voltage DC
cables. Journal of Materials Chemistry A, 4(22), 8590–8601.
Pallon, L. K., Nilsson, F., Yu, S., Liu, D., Diaz, A., Holler, M., Chen, X. R., Gubanski, S., Hedenqvist, M. S., & Olsson,
R. T. (2017). Three-dimensional nanometer features of direct current electrical trees in low-density polyethylene.
Nano Letters, 17(3), 1402–1408.
Pangali, C., Rao, M., & Berne, B. (1978). On a novel Monte Carlo scheme for simulating water and aqueous solutions.
Chemical Physics Letters, 55(3), 413–417.
Park, J., & Paul, D. (1997). Correlation and prediction of gas permeability in glassy polymer membrane materials via a
modified free volume based group contribution method. Journal of Membrane Science, 125(1), 23–39.
Pople, J., & Beveridge, D. (1970). Approximate molecular orbital theory. New York: McGraw-Hill.
Pople, J. A., Santry, D. P., & Segal, G. A. (1965). Approximate self-consistent molecular orbital theory. I. Invariant
procedures. The Journal of Chemical Physics, 43(10), S129–S135.
Pople, J., Beveridge, D., & Dobosh, P. (1967). Approximate self-consistent molecular-orbital theory. V. Intermediate
neglect of differential overlap. The Journal of Chemical Physics, 47(6), 2026–2033.
Pyzer-Knapp, E. O., Li, K., & Aspuru-Guzik, A. (2015). Learning from the harvard clean energy project: The use of
neural networks to accelerate materials discovery. Advanced Functional Materials, 25(41), 6495–6502.
Quateroni, A., Sacco, R., & Saleri, F. (2000). Numerical mathematics. In Applied mathematics (Vol. 37). New York:
Springer.
Rahman, A. (1964). Correlations in the motion of atoms in liquid argon. Physical Review, 136(2A), A405.
Rao, M., & Berne, B. (1979). On the force bias Monte Carlo simulation of simple liquids. The Journal of Chemical
Physics, 71(1), 129–132.
Rapaport, D. C. (1996). The art of molecular dynamics simulation. Cambridge: Cambridge University Press.
Rayleigh, J. (1877). The theory of sound, (edition of 1945). London: Macmillan and co.
Reddy, J. N. (1993). An introduction to the finite element method. New York: McGraw-Hill.
Reith, D., Pütz, M., & Müller-Plathe, F. (2003). Deriving effective mesoscale potentials from atomistic simulations.
Journal of Computational Chemistry, 24(13), 1624–1636.
€
Ritz, W. (1909). Uber eine neue Methode zur L€osung gewisser Variationsprobleme der mathematischen Physik.
Journal für die reine und angewandte Mathematik (Crelles Journal), 1909(135), 1–61.
Roothaan, C. C. J. (1951). New developments in molecular orbital theory. Reviews of Modern Physics, 23(2), 69.
Rosenbluth, M. N., & Rosenbluth, A. W. (1955). Monte Carlo calculation of the average extension of molecular chains.
The Journal of Chemical Physics, 23(2), 356–359.
Rossky, P. J., Doll, J., & Friedman, H. (1978). Brownian dynamics as smart Monte Carlo simulation. The Journal of
Chemical Physics, 69(10), 4628–4633.
Schellbach, K. (1851). Probleme der Variationsrechnung. Journal fur die Reine und Angewandte Mathematik, 1851
(41), 293–363.
Schr€ odinger, E. (1926). An undulatory theory of the mechanics of atoms and molecules. Physical Review, 28(6),
1049–1070.
Seifert, G., Eschrig, H., & Bieger, W. (1986). An approximation variant of LCAO-X-ALPHA methods. Zeitschrift Fur
Physikalische Chemie-Leipzig, 267(3), 529–539.
References 265

Siepmann, J. I., & Frenkel, D. (1992). Configurational bias Monte Carlo: A new sampling scheme for flexible chains.
Molecular Physics, 75(1), 59–70.
Slater, J. C. (1929). The theory of complex spectra. Physical Review, 34(10), 1293.
Sliozberg, Y. R., Kr€ oger, M., & Chantawansri, T. L. (2016). Fast equilibration protocol for million atom systems of
highly entangled linear polyethylene chains. The Journal of Chemical Physics, 144(15), 154901.
Stein, E. (2014). History of the finite element method–mathematics meets mechanics–part I: Engineering
developments. In The history of theoretical, material and computational mechanics-mathematics meets mechanics
and engineering (pp. 399–442). Berlin, Heidelberg: Springer.
Stillinger, F. H., & Rahman, A. (1974). Improved simulation of liquid water by molecular dynamics. The Journal of
Chemical Physics, 60(4), 1545–1557.
Strang, G. (1986). Introduction to applied mathematics. Wellesley: Wellesley-Cambridge Press.
Szabo, A., & Ostlund, N. S. (1996). Modern quantum chemistry: Introduction to advanced electronic structure theory.
Mineola: Dover Science Books.
Thomas, J. W. (1999). Numerical partial differential equations: conservation laws and elliptic equations – Conservation
Laws and Elliptic Equations, Vol 33. Springer Science & Business Media
Urey, H. C., & Bradley, C. A., Jr. (1931). The vibrations of pentatonic tetrahedral molecules. Physical Review, 38(11),
1969.
van der Giessen, E., Schultz, P. A., Bertin, N., Bulatov, V. V., Cai, W., Csányi, G., Foiles, S. M., Geers, M. G.,
González, C., & Hütter, M. (2020). Roadmap on multiscale materials modeling. Modelling and Simulation in
Materials Science and Engineering, 28(4), 043001.
Van Gunsteren, W. F., & Berendsen, H. J. (1990). Computer simulation of molecular dynamics: methodology,
applications, and perspectives in chemistry. Angewandte Chemie International Edition in English, 29(9), 992–1023.
Van Krevelen, D. W., & Te Nijenhuis, K. (2009). Properties of polymers: their correlation with chemical structure;
their numerical estimation and prediction from additive group contributions (4th ed.). Amsterdam: Elsevier.
Verdier, P. H., & Stockmayer, W. (1962). Monte Carlo calculations on the dynamics of polymers in dilute solution. The
Journal of Chemical Physics, 36(1), 227–235.
Verlet, L. (1967). Computer “experiments” on classical fluids. I. Thermodynamical properties of Lennard-Jones
molecules. Physical Review, 159(1), 98.
Versteeg, H. K., & Malalasekera, W. (2007). An introduction to computational fluid dynamics: the finite volume method
(2nd ed.). Harlow: Pearson Education.
Warshel, A., & Levitt, M. (1976). Theoretical studies of enzymic reactions: dielectric, electrostatic and steric stabiliza-
tion of the carbonium ion in the reaction of lysozyme. Journal of Molecular Biology, 103(2), 227–249.
Widom, B. (1963). Some topics in the theory of fluids. The Journal of Chemical Physics, 39(11), 2808–2812.
Young, W., & Elcock, E. (1966). Monte Carlo studies of vacancy migration in binary ordered alloys: I. Proceedings of
the Physical Society, 89(3), 735.
Chapter 6
Mechanical Properties

6.1 Introduction

Perhaps the most characteristic property of materials is their ability to withstand external stresses or
imposed strains. For polymers, the response to stress or strain varies with time of loading, temperature
and external environment to a greater extent compared to other materials, i.e. metals and ceramics.
While metals at ambient conditions are elastic materials showing Hookean response, polymers are
viscoelastic. The term viscoelastic means that the material has both elastic and viscous properties,
i.e. a combination of these. This ‘dual nature’ makes the prediction of stress response and service
lifetime of polymer products more complicated. Therefore, it is important, especially for product
designers and those involved in determining if a plastic component can be used under a given stress
state, that the mechanisms of stress response, yielding and fracture of polymers are understood. The
factors that influence the mechanical properties of a polymer may be of external or internal nature.
External factors are temperature, pressure, strain rate, type of loading, time of loading and environ-
ment. Internal factors include intramolecular structure, secondary bonds, molar mass and molar mass
distribution, chain branching, copolymerisation, chemical or physical crosslinking, crystallinity and
semicrystalline morphology, chain orientation, plasticisers, polymer blends and physical or chemical
aging.
The purpose of this chapter is to explain the correlation between these factors and the mechanical
properties. To be able to properly interpret the results, the chapter starts with explaining stress and
strain and how these interact and can be calculated in different stress situations. The chapter includes
a presentation of the important parameters that define the mechanical properties of a polymer, and the
corresponding methods to reveal these parameters are also discussed. Further, typical mechanical
data are given for several polymers as well as a few examples of the correlation between molecular
structure and mechanical response. Also included are subchapters that deal with linear viscoelasticity,
time–temperature shifting and examples on how different parameters affect creep, stress relaxation
and dynamic mechanical properties. The behaviour of polymers at larger strains is also covered,
including plastic deformation and fracture.
Plastic components are often expected to deform only a little during use. A plastic bucket, a gas
pipe and the coolant fan in a car are a few good examples. Other parts should deform extensively, for
example, a rubber o-ring and a rubber tyre. The human body provides good examples from both the
classes of material.
Another aspect of deformation is whether it is recoverable or permanent. Permanent deformation
can occur if the material is exposed to a stress above its yield stress, but also below that stress during,
e.g. a long-term continuous exposure to stress (creep deformation) or strain (stress relaxation). These
latter cases are due to the viscoelastic nature of some materials. It should be mentioned that if a

# Springer Nature Switzerland AG 2021 267

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_6
268 6 Mechanical Properties

viscoelastic material does not deform permanently after, e.g. a stress cycle, it is referred to as being
anelastic.
The deformation behaviour can be very different for different types of materials. A steel rod
deforms in a recoverable fashion only within 0.1% strain. The shape of the component will be
changed permanently at higher strains, i.e. it will not recover its original shape after unloading. The
unfortunate bending of a nail is a familiar example of permanent deformation, and it demonstrates
plastic deformation of the metal. A rubber band can, on the other hand, be stretched several hundred
percent, and it will still recover its original shape after unloading. Hence, different materials show
great differences in the maximum degree of deformation after which the original shape is preserved
after unloading.
You need only a steel plate, four balls of different materials – steel, rubber, wood and glass – and a
50 cm cylinder of transparent acrylic plastic to conduct the following illustrative experiment. Put the
cylinder on the steel plate, and drop the steel ball from the edge of the cylinder (50 cm above the steel
plate). The ball will fall down and hit the plate. The velocity (vmax) of the ball as it hits the plate is
given by:

mv2max pffiffiffiffiffiffiffiffi
mgh ¼ ) vmax ¼ 2gh ð6:1Þ
2

where m is the mass of the ball, g is the gravitational acceleration (9.81 m s2) and h is the height
(0.5 m). The velocity of the ball just before it hits the plate is ~3 m s1. When the ball hits the plate, it
will stop: v ¼ 0 m s1. Energy is conserved, which means that the kinetic energy (mv2max/2) is
transformed into deformation energy. Both the ball and the plate are deformed for a very short period
of time. Just a moment later, the deformation energy is transformed into kinetic energy, and the ball
starts moving upwards at a velocity of ~vmax. If deformation was completely elastic, the velocity of
the bouncing ball would be precisely vmax at the very moment the ball leaves the steel plate. The steel
ball will bounce a great many times before it stops. We may assume a figure for the number of
bounces: 1000. This number means that 99.9% of the deformation energy is elastic and that 0.1% of
the energy is converted to heat in each bounce. If we had a sensitive thermometer, we would detect a
small temperature rise in the ball and in the plate.
The rubber ball will also bounce a great many times, demonstrating its elastic deformation
behaviour. The wood ball stops bouncing much earlier, indicating that wood is a less elastic material.
The glass ball may behave in two different ways: (1) the ball will bounce a great many times (elastic
behaviour) or (2) it may crack after only a few bounces. In the latter case, part of the kinetic energy
will be transformed into new surface energy (∑σ iAi, where σ i is the specific surface energy of surface
i and Ai is the corresponding new surface area). What is learnt from these experiments is that elastic
materials may be very stiff like steel or very soft like rubber. A part of the deformation energy, in
some cases only a very tiny bit, is transformed into heat.
What is happening in these different materials during deformation on an atomistic level?
Steel is a highly crystalline material although not one giant single crystal. The atoms are
positioned in a lattice and their potentials are minimised. Deformation of the lattice will move the
atoms from their equilibrium positions, and the internal energy will increase. Each atom pair will
behave like a spring at small displacements of the atoms (Fig. 6.1). The potential function (U ) near
the equilibrium state of a given atom pair may be approximated by:

U ¼ C ðr  r 0 Þ2 ð6:2Þ

where C is a ‘spring’ constant (dimension: J m2), r is the actual distance between the atoms and r0 is
the equilibrium distance between the atoms. The force f, applied to keep the atoms at a distance r, is
obtained by taking the derivative of the potential function with respect to the distance:
6.2 Stress 269

Fig. 6.1 Deformation of a

polycrystalline metal with
springs between each
atom pair

dU d 2 
f ¼ ¼C r  2rr 0  r 20 ¼ 2Cðr  r 0 Þ ð6:3Þ
dr dr

The force is thus proportional to the displacement (r-r0) from the equilibrium position (r0). We
may define strain (ε) according to:
r  r0
ε¼ ð6:4Þ
r0

and stress (σ) as the force per unit area (A) of the cross section (unit: N m2 or Pascal, Pa):
f
σ¼ ð6:5Þ
A

Inserting Eqs. (6.4) and (6.5) in Eq. (6.3) yields well-known Hooke’s law:
2Cr 0
σ¼  ε ¼ Eε ð6:6Þ
A

where E is the elastic modulus (unit: N m2 or Pa). A stiff material like steel has a modulus of
ca. 200 GPa, whereas aluminium is softer, with a modulus of ca. 70 GPa (Bowman 2004). What can
be the reason for the difference? You may speculate on the basis of Eq. (6.6).
Glass is amorphous; the arrangement of the atoms lacks long-range order. Their positions are,
however, relatively fixed (frozen liquid). Deformation of glass displaces the atoms from a low energy
state (not the lowest possible internal energy) to a higher energy state. The elastic character of glasses,
inorganic and organic (at low temperatures), can be explained by the small displacement of the atoms
in the frozen-in disordered structure.
The rubber ball is highly elastic, i.e. only a little of the kinetic energy is lost in each bounce.
Rubbers are much more extensible than metals and the modulus is significantly lower, typically a few
MPa. The elasticity of rubbers is due to changes in entropy of the material. You can again speculate
about how that is possible (help is provided in Fig. 6.2). The stiffness of the rubber is not controlled by
its atomic potentials, which is very different from the behaviour of metals. The oddity in the group is
wood with its damping capacity. Why? A liquid like water can store no deformation energy except
pressure–volume work (compression). Stretching of such a liquid only results in flow and complete
conversion of deformation energy to heat. Wood is evidently neither fully elastic nor fully viscous. It
is rather a mixture of the two; it is viscoelastic.

6.2 Stress

6.2.1 Normal stress and Shear Stress

When a force (f) is acting on a body, the body will, according to Newton’s second law of motion
(Pinto 2007), accelerate according to a ¼ f/m, where m is the mass of the body. The body may be
270 6 Mechanical Properties

Fig. 6.2 The multitude of

shapes of polymers and
how they are related to the
degree of extension

Fig. 6.3 Forces acting on a rod: (a) contact force; (b) gravitational body forces

subjected to more than one force, and in such cases the force that relates to the acceleration vector is
the vector sum of the applied forces: a ¼ Σf/m. A body at rest must then obey the relationship:
X X
a¼ f i =m ¼ 0 ) fi ¼ 0 ð6:7Þ
i i

Also, the moment must be zero at equilibrium: ΣM ¼ 0. These two equations are central to the
analysis of stresses. The following practical experiment illustrates the mathematical principles. Take
a piece of rubber band and hold it between your two hands. Stretch the rubber band by separating your
two hands and stop. The rubber band is at rest (a ¼ 0) and yet you need to apply a constant force with
your two hands. The forces are obviously equal and directed from each other: f (left hand) + f (right
hand) ¼ 0. In addition, the moment has to be zero; otherwise, the rubber band would spin about the
axis of the moment. Where does the stress appear and what is it? The rubber band in this particular
example is subjected to a uniaxial stress. Consider that you divide the rubber band into two pieces by
a cut perpendicular to the force vectors. The left-hand piece is at rest which means that a force (f)
must act on the right-hand side of this part. This internal force must act on every section throughout
the rubber band. The stress is defined as this internal force divided by the cross-sectional area. The SI
unit for stress is N m2 or Pascal (Pa). Stress is denoted (σ) and it is a vector. The forces acting on a
body may be of two different kinds (Fig. 6.3): (1) Contact forces are applied to the body by direct
6.2 Stress 271

contact, usually on the surface part. (2) Body forces act on the body through action at a distance, for
example, gravity or magnetism.
The rod in the examples displayed in Fig. 6.3 is attached to a firm body, and, because the rod is at
rest, a counterforce (f) must act on the upper surface in case (a). The counteracting force will be the
same at any position along the rod, f, independent of x. The normal stress (σ) is defined as the force
(f) acting along the normal of a certain cross section divided by the area (A) of the cross section:
f
σ¼ ð6:8Þ
A

The normal stress acting on the cross section is positive in case (a). It is then referred to as a tensile
stress. When the forces have opposite directions, inwards from the surfaces, the normal stress is
negative and referred to as a compressive stress. It is obvious that the stress causes deformation of the
rod and the cross-sectional area changes from the unstressed value (A0) to A. Nominal stress (σN)
refers to the force divided by the original cross-sectional area (A0), whereas true stress (σ) is the force
divided by the actual cross-sectional area (A). At low strains, one per cent or less, the difference
between nominal and true stress is small (<1%). However, the difference between nominal and true
stress in a rubber band stretched to twice its original length is very significant. For a rubber, a class of
materials that show very little change in volume on deformation, it can be shown that the cross-
sectional area is reduced from A0 to A0/2. Hence, the true stress becomes in this particular case:
f f
σ¼ ¼ ¼ 2σN ð6:9Þ
A A0 =2

Case b in Fig. 6.3 shows a rod subjected to gravity. The normal stress along x is also a positive
tensile stress in this case. However, the normal stress changes with x and will be highest at x ¼ 0:
A0 hρg
σ ð x ¼ 0Þ ¼ ¼ hρg ð6:10Þ
A0

where h is the length of the rod, g is the gravitational acceleration and ρ is the density. The normal
stress is zero at x ¼ h (bottom of the rod). Hence, the full description of the normal stress along the bar
is given by the equation:
σ ¼ ðh  xÞρg ð6:11Þ

Figure 6.4 displays a rod subjected to a force acting along the long axis of the rod. Assume a cut
made along a plane with a normal vector at an angle θ to the applied force (f). The cross-sectional area
0 0 0
of this cut will be A ¼ A/ cos θ, and the normal force ( f ; the absolute value of the vector f ) acting on
this plane is:

f 0 ¼ f cos θ ð6:12Þ

Fig. 6.4 Stress

components in an oblique
plane
272 6 Mechanical Properties

Fig. 6.5 Normalised shear stress as a function of the angle between the external force and the normal to the plane at
which the shear stress acts (Eq. 6.15)

where f is the absolute value of f. The corresponding normal stress (σ’; the absolute value of the
vector) is:
f0 f cos θ f
σ0 ¼ ¼ ¼  cos 2 θ ð6:13Þ
A0 A= cos θ A
00 00
The other force component ( f ; the absolute value of the vector f ) is perpendicular to the normal to
the cut plane:
f 0 ¼ f sin θ ð6:14Þ

The corresponding shear stress (τ0 ; the absolute value of the vector) is given by:
f 00 f sin θ f
τ0 ¼ ¼ ¼  sin θ cos θ ð6:15Þ
A0 A= cos θ A

A shear stress is a stress with the force vector in the plane. Figure 6.5 shows that the shear stress
has a maximum at θ ¼ 45 . A uniaxially stressed rod has no shear stresses acting on the plane normal
to f or in any plane with normals perpendicular to f. A pair of scissors is a well-known application of
shear forces.
The normal stress in the case presented in Fig. 6.5 shows a monotonic decrease with increasing
angle θ between θ ¼ 0 and 90 , above which it increases to a maximum value at θ ¼ 180 .
Shear stresses play a key role when trying to solve the problem of ice formation on windmill blades
(ice formation makes the blades heavier). When developing ‘icephobic’ surfaces, i.e. surfaces which
the ice will have problems sticking to, the key is to reduce the critical shear stress needed for the ice to
move away (slip) from the blade surface when the blade is rotating. Icephobic surfaces are defined as
surfaces where the critical shear stress (ice adhesion strength) is below 100 kPa. One way of creating
‘passive’ surfaces with a minimal ice adhesion strength is to use a soft surface (using, e.g. an
elastomer) in combination with a lubricant (Golovin et al. 2016).
6.2 Stress 273

6.2.2 The Stress Tensor

Consider again the body at rest. The conditions that have to be fulfilled are: ∑fi ¼ 0 and ∑Mi ¼ 0.
The body may (hypothetically) be cut into two pieces. These parts must meet the same conditions as
the whole body, viz.:
X
f i ðextÞ þ f ∗ ¼ 0 ð6:16Þ

where the internal net force (f*) is exactly balancing the external forces acting on the cut surface of
the particular part of the body (lower part, for instance). On the cut surface of the other part (upper
part), the internal force must be –f*. An analogous equation holds for the momentum:
X
Mi ðextÞ þ M∗ ¼ 0 ð6:17Þ

where M* is the internal moment acting on the cut surface.

It is convenient to introduce three orthogonal unity vectors on the cut surface, n, s and t (Fig. 6.6),
and the internal net force acting on the cut surface is given by:

f ∗ ¼ Fn þ V s s þ V t t ð6:18Þ

where F is the normal force component and Vs and Vt are the orthogonal shear force components.
Three moments about the three axes (n, s and t) act on the cut surface to counteract the external
moments. These are denoted, respectively, Mn, Ms and Mt.
A bent (90 angle) cylinder subjected to a 250 N vertical force at one of the ends serves as a good
example (Fig. 6.7). The first task is to calculate the internal force and moment that act on the cut: Only

Fig. 6.6 Illustration of

directions in a cut plane

Fig. 6.7 Illustration of

calculation of forces and
moments acting on cut
surfaces
274 6 Mechanical Properties

one external force acts on the outer part (in N): (0, 100, 0). The internal force must then be (0, 100,
0). The moment about x is Mx ¼ 0.80 (i.e. 0.15–0.7 m) • -100 ¼ 80 Nm. The second task is to
determine internal force and moment in the cut along b: The internal force (in N) is (0, 100, 0). The
moment about x is Mx ¼ 1.50 • 100 ¼ 150 Nm. The moment about z is Mz ¼ 0.80 • 100 ¼ 80 Nm.
There is no moment about y, since the force is parallel to y.
It is convenient to express the stress (σ) acting on any plane in one normal stress (σ n) and two
orthogonal shear stress components (τns and τnt):
σ ¼ σ n n þ τns s þ τnt t ð6:19Þ

The full description of the state of stress at a point requires that the stress vectors are known for
three orthogonal planes (Bowman 2004) (Fig. 6.7):
0 1
σ xx σ xy σ xz
B C
σij ¼ @ σ yx σ yy σ yz A ð6:20Þ
σ zx σ zy σ zz

This array of nine stress components (σij) is referred to as the stress tensor. It is a second-rank
Cartesian tensor. The number of components in an n-rank Cartesian tensor is 3n. Hence, a scalar is of
zero-rank and a vector is of first-rank. The diagonal components (i ¼ j) are the normal stresses, and
the other six components (i 6¼ j) are shear stresses. Are the nine components independent? Analysis of
a small cube (illustration of a point) with the edges Δx, Δy and Δz with the prerequisite ∑M ¼ 0
yields (Fig. 6.8):
Δx Δy Δx Δy
σ xy ΔxΔz   σ yx ΔxΔz  þ σ xy ΔyΔz   σ yx ΔxΔz  ¼0
2 2 2 2 ð6:21Þ
∴σ xy ¼ σ yx

The same type of analysis of the other two pairs of shear stresses yields: σ xz ¼ σ zx and σ yz ¼ σ zy.
The stress tensor is thus symmetric and has only six independent stress components:

Fig. 6.8 Components in the stress tensor describing the state of stress at a point based on illustrations by (Saada (1993)
and Ward (1985)
6.2 Stress 275

0 1
σ xx σ xy σ xz
B C
σij ¼ @ σ xy σ yy σ yz A ð6:22Þ
σ xz σ yz σ zz

The stress vector (σn) referring to any cut plane with normal vector n ¼ (nx, ny, nz) through the
point at which the stress tensor (σij) is known can be calculated according to:
σ nx  σ xx nx  σ xy ny  σ xz nz ¼ 0
σ ny  σ xy nx  σ yy ny  σ yz nz ¼ 0 ð6:23Þ
σ nx  σ xz nx  σ yz ny  σ zz nz ¼ 0

or in more compact form:

0 1 0 10 1
σnx σxx σxy σxz nx
B C B CB C
B σny C ¼ B σxy σyy σyz C B C
@ A @ A @ ny A ð6:24Þ
σnz σxz σyz σzz nz
σn ¼ σij n

The shear stress is zero in all planes in which the resulting stress vector in the cut is parallel to the
normal of the plane. It can be shown that three planes exist with orthogonal normal vectors for which
the shear stresses are zero. The normal stresses in these planes are referred to as principal stresses.
The corresponding directions of the normal vectors are referred to as principal stress directions. The
stress tensor in a coordinate system parallel to principal stress directions is thus given by:
0 1
σ1 0 0
B C
σij ¼ @ 0 σ 2 0 A ð6:25Þ
0 0 σ3

where σ 1, σ 2 and σ 3 are the principal stresses (according to convention: σ 1 > σ 2 > σ 3). Uniaxial stress
state means that σ 2 and σ 3 are zero. If only one of principal stresses is zero, the stress state is referred
to as plane. The resulting stress (σn) in any cut plane (with normal n) is according to Eq. (6.24) given
by:
0 1 0 10 1 0 1
σ nx σ1 0 0 nx σ 1 nx
B C B CB C B C
@ σ ny A ¼ @ 0 σ 2 0 A@ ny A ¼ @ σ 2 ny A ð6:26Þ
σ nz 0 0 σ3 nz σ 3 nz

The normal stress (σ) in this plane is obtained by multiplying the normal vector and the resultant
stress vector:
0 1
σ 1 nx
 B C
σ ¼ nx n y n z @ σ 2 n y A ¼ σ 1 n x 2 þ σ 2 n y 2 þ σ 3 n z 2 ð6:27Þ
σ 3 nz

The shear stress (τ) in the cut plane can be calculated according to Pythagoras’ theorem:

j σn j 2 ¼ σ 2 þ τ 2 ð6:28Þ

and by combining Eqs. (6.27) and (6.28):

276 6 Mechanical Properties

Fig. 6.9 The maximum

shear stress appears in the
plane, which is at an angle
of 45 to the plane’s normal
to the highest (σ 1) and
lowest (σ 3) principal
stresses

τ2 ¼ ðσ 1  σ 2 Þ2 n2x n2y þ ðσ 2  σ 3 Þ2 n2y n2z þ ðσ 1  σ 3 Þ2 n2x n2z ð6:29Þ

The maximum shear stresses appear in the planes with normal directions at a 45 angle to the
principal stress directions:
 pffiffiffi pffiffiffi
σ  σ3
τ1 ¼ 2 0, 1= 2, 1= 2
2  pffiffiffi
σ1  σ3 pffiffiffi
τ2 ¼ 1= 2, 0, 1= 2 ð6:30Þ
2  pffiffiffi pffiffiffi 
σ  σ2
τ3 ¼ 1 1= 2, 1= 2, 0
2

The highest shear stress is τ2, which is in the bisector plane of the maximum and minimum
principal stresses (σ 1 and σ 3), as shown in Fig. 6.9.

6.2.3 The Mohr Stress Circle

The state of stress at a point can be graphically represented by the Mohr circle, which is drawn in the
normal stress (σ)–shear stress (τ) plane (Bowman 2004; Saada 1993). The third dimension must, in
this particular case, be along one of the principal stresses.
Force equilibrium in the element displayed in Fig. 6.10 yields the following equations:

σ φ  σ xx cos 2 φ  τxy sin φ cos φ  σ yy sin 2 φ ¼ 0 ð6:31Þ

τφ þ σ xx cos φ sin φ  τxy cos 2 φ þ τxy sin 2 φ  σ yy sin φ cos φ ¼ 0 ð6:32Þ

Simplification, considering that:

cos 2 φ þ sin 2 φ ¼ 1
2 sin φ cos φ ¼ sin 2φ ð6:33Þ
cos φ  sin φ ¼ cos 2φ
2 2

and rearrangement, give:

6.2 Stress 277

Fig. 6.10 Stresses acting on a small volume element

   2 hσ  σ i2
σ xx þ σ yy xx yy
σφ  þ τ2φ ¼ þ τ2xy ð6:34Þ
2 2

Equation (6.34) shows that the point (σ φ,τφ) is on a circle in the σ-τ plane. The circle has the radius
[(σ xx-σ yy)/2)2 + τxy2]0.5 and the centre at the point [(σ xx + σ yy)/2,0]. This circle, illustrated in Fig. 6.11,
is referred to as the Mohr circle in a plane stress condition (or in a plane perpendicular to a principal
stress), and it is used as a graphical tool to reveal the stress state at a point. The normal stresses are
obtained as the σ values at the centre  the radius (Fig. 6.11):
rhffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
σ xx þ σ yy σ xx  σ yy i2
σ1 ¼ þ þ τxy 2 ð6:35Þ
2 2
rhffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
σ xx þ σ yy σ xx  σ yy i2
σ2 ¼  þ τxy 2 ð6:36Þ
2 2

It is seen in Fig. 6.11 that the maximum shear stress is (σ 1  σ 2)/2. The directions of the principal
stresses in the plane are obtained using the condition that τφ ¼ 0 according to:
τxy 2τxy
tan 2φ ¼ ¼ ð6:37Þ
ðσ xx þσ yy Þ σ xx  σ yy
σ xx  2

In order to determine the Mohr circle, three quantities need to be known: either three points on the
circle or two points on the circle and the centre of the circle. When the circle is drawn, you can readily
determine the principal stresses in the xy-plane and their directions and graphically determine the
normal and shear stresses acting on any cut at a known angle (φ) to x. You can also use the analytical
expressions displayed in this section, i.e. Eqs. (6.35, 6.36 and 6.37). The normal stress and shear stress
in the cut at an angle φ to x are obtained by clockwise rotation from (σ xx, τxy) by an angle
2φ (Fig. 6.11).
278 6 Mechanical Properties

Fig. 6.11 The Mohr stress circle for a plane stress state

In systems deviating from the above conditions, for example, in a three-dimensional stress
condition, the system can be represented by three Mohr strain circles, relating σ 1 to σ 2, σ 1 to σ 3
and σ 2 to σ 3.

6.3 Strain

6.3.1 Introduction: Uniaxially Loaded Specimens

The shape of a body is changed by the action of stresses. A tensile stress will stretch a specimen along
the direction of the applied stress, and simultaneously the specimen contracts in the two perpendicular
directions. Take a band of rubber, apply three ink dots as shown in Fig. 6.12 and measure with a ruler
the displacements (Δx and Δy) of the dots after loading. You may also measure the change in the
specimen thickness (Δz) on loading.
The displacements along the three orthogonal axes are commonly expressed as strains defined
according to:
Δx
εxx ¼
x0
Δy
εyy ¼ ð6:38Þ
y0
Δz
εzz ¼
z0

In the particular case shown in Fig. 6.12, the strain along x (εxx) is positive (Δx > 0), whereas both
εyy and εzz are negative since both Δy and Δz are negative (contraction). There are several ways of
expressing strain, as will be discussed in this chapter (Bowman 2004). The above type is referred to as
engineering strain or nominal strain or occasionally linear strain. Any cubic volume element with
one edge parallel to the applied stress keeps its 90 angles during deformation. This means that the
volume of the deformed specimens is readily calculated on the basis of the three orthogonal linear
strain values. The three strains expressing the change in shape of the cube are so-called normal
6.3 Strain 279

Fig. 6.12 Uniaxial loading along x and deformation along x, y and z

strains. A synonym for this is extensional strain. Careful measurement of the three normal strains εxx,
εyy and εzz permits the assessment of change in volume on deformation:
 
ΔV ðx0 þ ΔxÞðy0 þ ΔyÞðz0 þ ΔzÞ  x0 y0 z0 ðx0 þ x0 εxx Þ y0 þ y0 εyy ðz0 þ z0 εzz Þ  x0 y0 z0
¼ ¼ ¼
V0 x0 y0 z0 x0 y 0 z0
 
¼ð1 þ εxx Þ 1 þ εyy ð1 þ εzz Þ  1 ¼ 1 þ εxx þ εyy þ εzz þ εxx εyy þ εxx εzz þ εyy εzz þ εxx εyy εzz  1
ð6:39Þ

The following simple expression is obtained by neglecting the products of two or more strains
(these are very small at low strains):
ΔV
ffi εxx þ εyy þ εzz ð6:40Þ
V0

A very useful material parameter relating the normal strain along the uniaxial stress (εxx) and the
normal strain perpendicular to the applied uniaxial stress (εyy or εzz) is Poisson’s ratio (υ) (Nielsen
and Landel 1994):
εyy ε
υ¼ ¼  zz ð6:41Þ
εxx εxx

This equation assumes that the strains in the two orthogonal directions (along y and z) are the
same, which is only true for an isotropic material. The term ‘isotropic’ refers to a material with the
same properties (e.g. stiffness) in all possible directions. By combining Eqs. (6.40) and (6.41), the
following equation is obtained:
ΔV
¼ εxx  2υεxx ¼ εxx ð1  2υÞ ð6:42Þ
V

Equation (6.42) implies that materials exhibiting a Poisson’s ratio smaller than 0.5 show positive
volume strains, i.e. they increase their volume on uniaxial loading. Figure 6.13 presents Poisson’s
ratio data for a collection of different materials. Very stiff materials like diamond and SiC have a very
low Poisson’s ratio. The polycrystalline metals show higher values: υ ¼ 0.2–0.4. Materials with
υ ¼ 0.5 deform without volume change. Unfilled rubbers and mercury have Poisson’s ratios close to
0.5.
Fibres, due to their high degree of chain packing perpendicular to the tensile direction, have a low
Poisson’s ratio in the fibre direction. Most glassy polymers have a Poisson’s ratio in the range 0.3–0.4.
280 6 Mechanical Properties

Fig. 6.13 The volume change as a function of unidirectional strain for different materials. Numbers refer to Poisson’s
ratio. Drawn after data of Van Krevelen and Te Nijeuhuis (2009)

If υ is not known for a specific glassy polymer of interest, a good approximation at ambient
temperature is 0.35 (Nielsen and Landel 1994; van Krevelen 1990). Above the heat deflection
temperature, i.e. when the glassy material is softened, it is better to use a Poisson’s ratio closer to
0.5. Rubbery semicrystalline polymers, e.g. polyethylene, have values in the range 0.45–0.49.
Poisson’s ratio is approximated as a constant in the above examples. When a large strain and long
loading time are considered, υ is not a constant but rather a function of both strain and time. For
flexible polymers, Poisson’s ratio, in general, decreases with increasing strain due to the polymer
chains orienting in the draw direction, and the properties of the polymer (including υ) approach those
of fibres. Also, internal voids/cracks may develop (volume increase), which adds to the decrease in
Poisson’s ratio. Stiff fillers counteract the compression in the direction perpendicular to the draw
direction, and this lowers Poisson’s ratio (refer to filled and unfilled rubbers). Moreover, υ can vary
with the type of loading.
For a glassy polymer, as illustrated in Fig. 6.14, Poisson’s ratio increases with time of mechanical
load. The glassy polymer, which has a relatively low Poisson’s ratio, experiences a phase transfor-
mation and approaches the rubbery/viscous state, which is associated with a larger υ value. There are
several applications where it would be of interest to have a negative Poisson’s ratio. This would imply
that the cross section of a rod expands when it is tensile strained in the rod direction and retracts when
it is compressed. If this would be the case for a polymer gasket, it would be easy to ‘hammer’ it into a
circular hole but not easy to remove it by pulling it out. Is there a polymer that has a negative value?
No. However, there are ‘constructions’ which exhibit a negative Poisson’s ratio (Lakes 1987; Lakes
and Witt 2002). If a polymer foam is heated to slightly below the softening temperature and exposed
to a hydrostatic pressure, the cell walls collapse inwards. As observed in Fig. 6.15, the cell structure
will always expand in the direction perpendicular to any orthogonal draw direction – consider that the
pulling is performed from the middle section of the sides.
Many materials undergo plastic deformation at larger strains, and this deformation is largely
without volume change (υ  0.5). The unusual behaviour of decreasing the volume on uniaxial
loading may only occur in materials that show a phase transition (e.g. crystallisation) induced by
deformation. Vulcanised natural rubber is known to crystallise when stretched to 300–400%.
6.3 Strain 281

Fig. 6.14 Poisson’s ratio for PMMA as a function of loading time. Drawn from data (averaged/smoothed) collected
from creep experiments at 40–100  C (Lu et al. 1997)

Fig. 6.15 A collapsed cell

with a negative Poisson’s
ratio. Drawn from Lakes
(1987)

Fig. 6.16 Deformation of an oblique cube. The quadratic shape of the unstressed element is changed into a
parallelogram by the shear stresses

Let us return to the uniaxially loaded specimen shown in Fig. 6.12 and the cube in Fig. 6.16. The
orientation of the cubic volume element in consideration may be rotated with respect to the direction
of the applied load (45 rotation in the example shown; Fig. 6.16). The angles between the edges will
not be 90 in the stressed element. This kind of deformation is referred to as shear strain (Bowman
282 6 Mechanical Properties

Fig. 6.17 Definition of

shear strain by an example

2004). Note that both normal stresses and shear stresses act on the surfaces (only the shear stresses are
indicated in Fig. 6.16).
The shear strain (γ xy) is defined as the change of angle AB–AD (in radians) from the undeformed
90 (Fig. 6.17). The exact calculation of the shear strain is obtained by the angular change, according
to:

γ xy ¼ tan 1 ð0:12L=LÞ þ tan 1 ð0:12L=LÞ ¼ 0:2388 ð6:43Þ

However, because the angular changes are very small, the following approximation can be made:
tan θ  sin θ  θ. Hence,
γ xy ¼ ð0:12L=LÞ þ ð0:12L=LÞ ¼ 0:24 ð6:44Þ

The difference between the exact value, 0.2388, and the approximate value, 0.24, is very small
(0.5%).

6.3.2 The Strain Tensor and Other Strain Concepts

The strict definition of strain is based on the displacement vector u, which is a function of the space
coordinates x, y and z (Ward 1985; Saada 1993). This vector expresses the displacement of every
point in the continuum. For each and every point, it has three components along the three orthogonal
axes: ux, uy and uz. For the sake of simplicity, the normal and shear strains are defined on the basis of
the displacement vectors in two-dimensional space (Fig. 6.18). The normal strains along x and y are
defined according to:
lim ux ðx þ dx, yÞ  ux ðx, yÞ ∂ux
εxx ¼ ¼ ð6:45Þ
dx ! 0 dx ∂x

lim uy ðx, y þ dyÞ  uy ðx, yÞ ∂uy

εyy ¼ ¼ ð6:46Þ
dy ! 0 dy ∂y

The double subscripts ‘xx’ and ‘yy’ indicate simply that the strain is the displacement vector
component along the same direction, for example, the gradient of the x-component of u along x.
6.3 Strain 283

Fig. 6.18 Deformation of an element in xy space expressed as displacement vector u

The shear strain (γ xy) is the change in the 90 angle of the left-hand lower corner of the quadrate:
 
lim uy ðx þ dx, yÞ  uy ðx, yÞ ux ðx, y þ dyÞ  ux ðx, yÞ ∂uy ∂ux
γ xy ¼ þ ¼ þ ð6:47Þ
dx, dy ! 0 dx dy ∂x ∂y

The change in angle of the left-hand lower corner is thus the consequence of the x-gradient of the
y-component of the displacement vector and the y-gradient of the x-component of the displacement
vector. These two strains are denoted εxy and εyx, and they are given by:
γ xy
εxy ¼ εyx ¼ ð6:48Þ
2

The state of strain at a point in three-dimensional space can be expressed in analogous manner to
the stress, namely, as a second-rank tensor:
0 1
εxx εxy εxz
B C
εij ¼ @ εxy εyy εyz A ð6:49Þ
εxz εyz εzz

The strain tensor has many similarities with the stress tensor. It is also symmetric about i ¼ j.
Hence, the strain tensor has six independent components. The following very useful expression
makes it possible to calculate the normal strain (εn) along a vector n ¼ (nx, ny, nz) from the strain
tensor:

εn ¼ n2x εxx þ n2y εyy þ n2z εzz þ 2nx ny εxy þ 2nx nz εxz þ 2ny nz εyz ð6:50Þ

This equation can be compactly written (note the analogy to the stress tensor):
0 10 1
εxx εxy εxz nx
B CB C
εn ¼ ð nx ny nz Þ@ εxy εyy εyz A@ ny A ð6:51Þ
εxz εyz εzz nz
284 6 Mechanical Properties

Another similarity between strain and stress is that three orthogonal directions exist which define a
volume element with no shear strains. The normal strains in these directions are referred to as
principal strains (ε1, ε2 and ε3; ε1 being largest). If two of the principal strains are zero, the strain
state is referred to as uniaxial. Plane strain is achieved when one of the principal strains is zero. The
volume strain can be calculated according to Eq. (6.40), viz.:
ΔV
ffi ε1 þ ε2 þ ε3 ð6:52Þ
V0

Several other ways exist of denoting strain. In plasticity theory, when multiple step forming is the
case and when large deformations are considered, true strain/logarithmic strain (ε) is commonly used
(Bowman 2004). It is obtained from the expression below, where l is the actual length:
ðL
∂l
ε¼ ð6:53Þ
l
L0

which is integrated with the starting condition ε ¼ 0 when L ¼ L0:

L L0 þ ΔL
εL ¼ ln ¼ ln ¼ ln ð1 þ εÞ ð6:54Þ
L0 L0

At small strains, ε  εL. However, for larger (plastic) strains, the logarithmic strain is a better
descriptor. In rubber technology, λi (i ¼ 1,2 or 3) is commonly used:
x
λ1 ¼
x0
y
λ2 ¼ ð6:55Þ
y0
z
λ3 ¼
z0

where x0, y0 and z0 are the dimensions of the unstressed specimen and x, y and z are the specimen
dimensions after the application of the stress. Calculations are greatly simplified provided that the
coordinate system coincides with the principal strain directions.

6.3.3 The Mohr Strain Circle

In analogy with stress, the strain in a plane perpendicular to a principal strain, or in plane strain
conditions, is represented by a circle, the Mohr strain circle (Case and Chilver 1959). The principal
strains (ε1 and ε2) are given by (a graphical explanation is provided in Fig. 6.19):
  sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  2  
εxx þ εyy εxx  εyy γ xy 2
ε1 ¼ þ þ
2 2 2
  sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  2   ð6:56Þ
εxx þ εyy εxx  εyy γ xy 2
ε2 ¼  þ
2 2 2
6.4 Assessment of the Mechanical Properties 285

Fig. 6.19 The Mohr strain circle

6.4 Assessment of the Mechanical Properties

6.4.1 Introduction

Structures and machines, or their components, fail because of fracture or excessive deformation. In
attempting to prevent such failure, the designer estimates how much stress can be anticipated and
specifies materials that can withstand the expected stresses. A stress analysis, accomplished either
experimentally or by means of a mathematical model, indicates expected areas of high stress in a
machine or structure. Mechanical property tests, carried out experimentally, indicate which materials
may safely be employed.
A wide range of test methods are available. Traditional methods use a relatively large volume of
material, [cm3] amounts or more. Accuracy and reproducibility of the measurements increase,
generally speaking, with increasing volume of test specimen. Chemists and material scientists that
produce precious new materials in very small quantities would like to assess the mechanical
properties of the synthesised materials with reasonable accuracy. Fortunately, new methods using
very small amounts of sample have been developed.
Figure 6.20 shows a few commonly used structures in mechanical testing. Most tests performed
are based on uniaxially stressed bars/strip or dumbbell-shaped specimens (Fig. 6.21). Three-point
bending and torsion are other relatively common tests. Hydrostatic pressure testing of pipes and
containers is also performed.
The results of tests are sensitively dependent on the test pieces. Rough edges may cause early
fracture. The stress distribution should be uniform in the specimen. Statistical analysis of test results
is always preferred, particularly in the study of fracture. A Swedish engineer, Waloddi Weibull
(1887–1979), gave name to one of the most important statistical functions (Weibull distribution) that
describe the distribution of lifetime or fracture stress.
286 6 Mechanical Properties

Fig. 6.20 Different specimens and loadings: (1) uniaxial testing, (2) three-point bending, (3) torsion of cylinder and
(4) hydrostatic pressure testing. Arrows show direction of force

Fig. 6.21 A dumbbell

‘dogbone’ specimen

6.4.2 Tensile Testing

In tensile testing, it is common to use a constant displacement rate and obtain the load (stress) versus
the displacement (strain). Commonly, dumbbell specimens are used (Fig. 6.21). These can be directly
moulded into the required shape by injection or compression moulding. They can also be produced by
cutting the specimen out from, e.g. a larger sheet, through stamping, water-jet and laser techniques.
Care should be taken when selecting the preparation method. Injection moulding introduces molecu-
lar orientation that can be different from the product, for which the mechanical properties are
assessed. Stamping requires a sufficiently weak sample. Water-jet, laser cutting and machining can
6.4 Assessment of the Mechanical Properties 287

produce thermal degradation of the specimen at the edges. The reason for using a dumbbell specimen
is to have the main mechanical deformation located far from the clamps holding the sample. The
stress situation at the clamps can be complicated, and if straight specimens are used, it is possible that
the specimen breaks at the clamps. Nevertheless, when testing thin plastic sheets/films (specimens
below 1 mm in thickness), straight specimens can also be used. In such case, specimens breaking at
the clamps should be discarded. The strain along the specimen can be measured by an extensometer
mechanically clamped onto the specimen. Nowadays, however, the use of video extensometers is
common practice, where the strain is captured between gauge marks on the specimen. Thus, no
physical contact is required here. It also makes it easier to use extensometers also on thin films. The
strain can also be measured as the relative displacement of the two clamps, which is then referred to as
nominal strain. The risk with this strain measurement is that the specimen may glide between the
clamps, a risk that increases the stronger the sample is. Apart from specimens having a ‘sheet’
geometry, these can also be tensile tested in the form of tubes and rods. Pins, glue and threading can,
apart from clamps, also be used to hold the specimen. There exist different load cells used in the
tensile tester. It is important to choose one that covers the size of the load needed to stretch the
specimen, but is also sensitive enough to track small variations in the obtained stress values. Since
polymers are viscoelastic, the choice of strain/displacement rate can affect the determined properties
substantially. Hence, also strain rate has to be chosen with care.
In creep tests the specimen is subjected to a constant load, and the strain is measured as a function
of time. In stress relaxation tests, the specimen is stretched to a constant displacement, and the force/
load is recorded as a function of time. The decrease in force with time (stress relaxation) is due to the
viscoelastic nature of the polymers, and the test is referred to as the stress relaxation test. The
compression test is the opposite to the tensile test. It is commonly performed on, e.g. polymer
foams, which during service normally are subjected to compressive forces, rather than tensile forces.
In compression, usually there is no need for regular clamps; the sample is compressed between two
flat metal platens. Buckling can be a problem, which can be minimised by using short specimens.
Friction between the specimen and the platen affects the results, which, if needed, can be avoided by
lubricating the platens.

6.4.3 Dynamic Mechanical Analysis, Dilatometry and Thermal Mechanical

Analysis

Dynamic mechanical thermal analysis (DMTA, or DMA) is described in detail in Chap. 1, and it will
here only be described briefly. Besides generating pure mechanical data under dynamic conditions, it
also gives information on processes in the polymer. It can, e.g. be used to determine the glass
transition temperature and its frequency dependence. If the glass transition is difficult to detect
with calorimetric measurements, DMTA is often a good alternative.
As mentioned in Chap. 1, the real (G0 ) and imaginary (G00 ) parts of the shear modulus for a
cylindrical specimen with a length l and a radius r in a torsional pendulum measurement (free
vibration) are given by:

2lMω2
G0 ¼ ð6:57Þ
πr 4

and
288 6 Mechanical Properties

2l M 2 Λ
G00 ¼ ω  ð6:58Þ
πr 4 π

M is the moment of inertia of the rod connected to the specimen, and the angular vibrational
frequency is denoted ω. For a viscoelastic material in this free vibration condition, damping occurs
which means that the amplitude of the vibration/swing decays with time. This is described by the
logarithmic decrement Λ:

1 An
Λ¼ ln ð6:59Þ
k Anþk

where the amplitude of the nth swing is An and k is the number of swings. In the forced oscillation
technique, a larger frequency range can be used than in the torsion pendulum technique. In commer-
cially available instruments, the measurement can be performed under also bending, shear and
uniaxial tension conditions.
Dilatometry and thermal mechanical analysis (TMA) can be used to obtain mechanical properties
of polymers. The methods are described in detail in Chap. 1 and will only be briefly mentioned here.
Volume dilatometry is useful for determining the mechanical response under a hydrostatic condition,
e.g. in subsea applications. A sample is immersed in a liquid, and it is important that the liquid does
not swell the sample (mercury, water, alcohols and silicon fluids have been used). The sample volume
is determined by monitoring the level of the fluid as a function of pressure/force and temperature. In
linear dilatometry (TMA), the compression modulus (short-term or in creep) can be determined
(in one direction) by the use of a penetrating probe as a function of temperature.

6.4.4 Fracture Testing

Fracture stress and strain can be obtained in a conventional constant elongation-rate instrument. The
result will be very strongly dependent on the specimen tested. Scratches on the specimen reduce both
the stress and strain at break, particularly for brittle materials. Defects on the specimens are in most
practical cases unavoidable and are commonly dealt with by running multiple tests (perhaps ten) and
evaluating the results by statistical methods. Notched specimens are used in some cases in order to
control the stress concentration (Fig. 6.22). A notch is a ‘scratch’ deliberately made in a controlled
way. A crack will start to grow from the base of the notch, and by using a travelling microscope, it is
possible to follow crack growth as a function of the applied stress. This type of experiment is
performed in fracture mechanics, and for linear elastic materials, it is possible to use the data
obtained from one experiment (a certain notch geometry) to predict the behaviour of other specimens
with different geometries. Fracture mechanics can be applied to viscoelastic materials but with much
more difficulty (refer to sections below on fracture mechanics). The stress–strain curves (also fracture
stress and strain) obtained for viscoelastic materials depend on the applied strain rate. This is not the
case for elastic materials.
Un-notched or notched specimens may also be tested at constant load (creep test), and in this case
the time to fracture is determined as a function of the applied stress (refer to section on creep failure
below). The way the notch is made may affect the fracture mechanics parameters obtained. Salazar
et al. (2013) have shown that the fracture parameters when the notch was made with a steel razor
blade were different from when the notch was made by femtosecond laser ablation. This was ascribed
to the damage zone in front of the notch being different, yielding different local stress and strain
states.
6.4 Assessment of the Mechanical Properties 289

Fig. 6.22 A sample bar

with a notch

Many materials, sensitive to the presence of flaws, cracks and notches, fail suddenly under impact.
The impact tests employ either a pendulum load or a falling weight to obtain impact data. These are
described in the impact failure section below.
Materials that survive a single application of stress frequently fail when stressed repeatedly. This
phenomenon, known as fatigue, is measured by mechanical tests that involve repeated application of
different stresses varying in a regular cycle from maximum to minimum values. Most fatigue testing
machines produce this cyclically varying load hydraulically. A material is generally considered to
suffer from low-cycle fatigue if it fails in 10,000 cycles or less. The stresses acting on a material in the
real world are usually random in nature rather than cyclic. Consequently, several cumulative fatigue
damage theories have been developed to enable investigators to extrapolate from cyclic test data to
prediction of material behaviour under random stresses. Because these theories are not applicable to
most materials, a technique that involves mechanical application of random fatigue stresses, statisti-
cally matched to real-life conditions, is now employed in test laboratories. More on fracture
mechanics, fatigue and test methods are given in the fracture mechanics part later in this chapter.

6.4.5 Surface Mechanics Methods

There exist several types of methods to assess a material’s resistance to indentation, yielding hardness
(which is not a true material property) and stiffness (Broitman 2017). They can be divided into
macroindentation tests (including Brinell, Vickers, Rockwell, Shore, IRHD and Barcol),
microindentation tests (including micro IRHD, micro Vickers and Knoop) and nanoindentation
290 6 Mechanical Properties

tests. For polymers Shore, Rockwell, Barcol and IRHD hardness tests are often used. In the Shore
(durometer) test, a spring-loaded needle-shaped indentor is applied onto the material. Depending on
the stiffness, two different Shore scales are used. Shore A is applied on soft polymers and Shore D on
stiff polymers. The difference is that a thinner needle and a larger force are applied in Shore D, than in
Shore A. The scale is from 0 to 100, and the latter means no penetration of the needle. In the Rockwell
hardness test, the indentor has the form of a ball or diamond. The method consists of several load
stages where first a load of lower size is applied, followed by a larger load, the removal of the larger
load and the reapplication of the lower load. After removal of the indenter, the height of the indent
(difference between the final indent and the position of the indentor after the first low load applica-
tion) is used for assessing the Rockwell hardness value, which is indicated by a number following an
HRX notation, where X indicates the actual scale A–K. The different scales imply different loads and
indentor geometries. The IRHD test (International Rubber Hardness Degree, ball indentor) is similar
to the Rockwell test in that two different forces are used. The penetration depth is measured as the
difference in indent caused by the two forces. In the Barcol test, which is applied mainly to rigid
plastics, an indent is obtained by a uniform force increased to a maximum reading. The size of the
indent is converted to a Barcol value.
Besides hardness values, an indenter modulus can also be obtained from an indenter measurement.
By, e.g. combining a tensile tester and an indenter set-up, the force as a function of penetration depth
is determined. By taking the slope of the force–penetration curve, an indenter modulus is obtained
(unit: N/m).
In contrast to the above methods, nanoindentation uses a nanosized indenter tip (commonly made
of diamond). The technique allows for the determination of hardness, modulus, fracture toughness,
etc. of the tested polymeric material. The loading and unloading generate a hysteresis loop, and from
the initial slope of the unloading curve, a reduced/effective modulus representing the contact between
the tip and the sample is obtained, which is consequently a combination of the sample and tip moduli.
The sample modulus can, however, be calculated from the known tip modulus and the known or
estimated Poisson’s ratios of the sample and the tip (Broitman 2017). An in-depth description of the
nanoindentation theory is available in Das and Bhattacharyya (2017) and Broitman (2017). One of the
most common instruments for nanoindentation is a Triboindentor (Hysitron), which can be coupled
with a scanning probe microscope.
Atomic force microscopy (AFM, fully described in Chap. 2) can be used to determine a number of
mechanical properties, such as the elastic modulus (Krieg et al. 2019; Roa et al. 2011). In essence, the
AFM probe is used as an indentor by which a force between μN and tens of pN is applied. As the
cantilever is moved towards the sample, its deflection/position and the applied force are recorded. The
effective modulus is obtained from the slope of the force versus the distance between the sample and
the cantilever (he3/2, the exponent due to the application of the Hertz model for calculation of the
cross-sectional contact area) and the combined curvature radius of the tip and the sample in contact.
The sample modulus is obtained from the reduced effective modulus in the same way as in the
nanoindentation technique.

6.5 Definition of Mechanical Parameters from the Tensile Test

6.5.1 The Stress–Strain Curve

The stress–strain (tensile) test is probably the most common method to determine mechanical
properties, since the test procedure is relatively simple and fast. The data may, however, be of limited
value for a product developer and is mostly sufficient only for comparing materials. The reason is that
6.5 Definition of Mechanical Parameters from the Tensile Test 291

Fig. 6.23 Stress–strain

curves for brittle, ductile
and elastomeric materials.
Y and F refer to the yield
and fracture points,
respectively

Fig. 6.24 Definitions of

stiffness, strength and
ductility from a
stress–strain curve

the type of loading does not resemble any practical situation. How often is the material/product
exposed to a constant strain rate? In addition, mechanical parameters determined by stress–strain
measurements are classified as short-term performance parameters and may not be used to predict
long-term performance in, for example, creep. At large strains, the polymer undergoes molecular and
morphological changes, including chain and crystal reorientation and crystallinity increase, which
substantially changes its mechanical behaviour. Typical stress–strain curves are shown below for
brittle, ductile and elastomeric materials. The yield and fracture stresses and strains are defined in the
figure (Fig. 6.23).
From a stress–strain curve, it is possible to obtain stiffness, strength and ductility (Fig. 6.24)
(Nielsen and Landel 1994). A stiff material shows a steep initial stress–strain curve (high modulus).
The strength is normally the maximum stress, i.e. the yield or fracture stress of a ductile material or
fracture stress of a brittle material. Ductility is the ability to deform plastically without fracturing. It
can be measured as the strain at break. Toughness, in turn, is the ability to absorb energy before
fracture and can be obtained as the area under the stress–strain curve. A material may, in principal,
have any combination of stiffness/softness, strength/weakness and ductility/brittleness. In most cases,
however, the strength is proportional to the stiffness (Hedenqvist et al. 2003). However, for oriented
systems, this is not necessarily the case (Ito et al. 2002; Smith and Lemstra 1980).

6.5.2 The Stiffness and the Tensile Modulus

Many solid materials, for example, polycrystalline metals and inorganic and organic glasses, show
elastic deformation at small strains. In these cases, stress and strain are related without time as a
292 6 Mechanical Properties

parameter of importance. The strain response to a sudden change in stress is instantaneous. Another
important feature of elastic materials is that they recover fully and instantaneously on unloading. The
Englishman Robert Hooke was the first to formulate the behaviour of elastic materials (Bowman
2004). He stated in 1660 that the stretching of a solid body is proportional to the force applied to
it. The concepts of stress and strain were not known at the time. Thomas Young (English physician,
physicist and Egyptologist) formulated the mathematical expression known as Hooke’s law around
1800 (Bowman 2004):
σ ¼ Eε ð6:60Þ

where σ is the applied uniaxial stress, ε is the normal strain along the direction of the applied stress
and E is the so-called elasticity modulus or, as it is also called, Young’s modulus. The unit for
elasticity modulus is Pascal (N m2). E is a material constant, and it expresses the stiffness of the
material under the experimental conditions (temperature, degree of humidity, pressure). Diamond is
the stiffest known material with an elasticity modulus of approximately 1000 GPa (1012 Pa) (Klein
and Cardinale 1993). Glassy polymers are elastic at low temperatures with E values in the range of a
few GPa. Very soft rubbery materials are also elastic, and their Young’s moduli are of the order of
1 MPa (106 Pa). The origin of rubber elasticity is entropy change, whereas for the stiffer materials,
Young’s modulus is controlled by the bond potential functions. It is interesting to note that Young’s
modulus varies by a factor of 1000 for this group of materials.
Rubbers are elastic, but their stress–strain equation is different from Hooke’s law (Gedde and
Hedenqvist 2019a):
 
1
σ ¼ E λ2 þ ð6:61Þ
λ

In addition, the elasticity modulus of rubbers increases with increasing temperature.

6.6 The Three-Dimensional View for Modulus and Strain

Both stress and strain are second-ranked Cartesian tensors, each with six independent components.
According to Hooke’s law, the components of the stress tensor (σij) should be strictly proportional to
the corresponding components of the strain tensor (εkl), i.e. (Bowman 2004):
σij ¼ Cijkl εkl ð6:62Þ

where Cijkl is the elastic modulus tensor with 81 elastic constants (rank four Cartesian tensor). The
stress and strain tensors are both symmetric, i.e. the subscripts i and j can change place and so can
k and l:
Cijkl ¼ Cjikl ; Cijkl ¼ Cijlk ð6:63Þ

The number of independent elastic constants is reduced to a more convenient 21. An isotropic
material (stiffness is the same in all possible directions) requires only the determination of two
independent elastic constants. Hooke’s generalised law is applicable to isotropic elastic solids,
provided that the coordinate system is oriented so that x, y and z are parallel to the principal
directions:
6.6 The Three-Dimensional View for Modulus and Strain 293

Fig. 6.25 The body is subjected to pure shear; 1 and 3 refer to the principal directions in the xy-plane. The
corresponding Mohr stress and strain circles are shown at the bottom

1  
εx ¼ σx  υ σy þ σz
E
1
εy ¼ σ y  υðσ x þ σ z Þ ð6:64Þ
E
1  
εz ¼ σ z  υ σ x þ σ y
E

where E is Young’s modulus and υ is Poisson’s ratio. Both these quantities can be assessed by a
suitable uniaxial stress–strain experiment. The equations relating shear strain and shear stress are,
according to Hooke’s law and Fig. 6.25, given by:
τxy
εxy ¼
2G
τyz
εyz ¼ ð6:65Þ
2G
τ
εzx ¼ zx
2G

Note that the shear modulus (G) is given by the ratio between shear stress and shear strain
(γ ij ¼ 2εij). An important expression relating shear modulus and Young’s modulus can be derived
by considering the Mohr stress and strain circles for a specimen subjected to plane stress (principal
294 6 Mechanical Properties

stress along z is zero) and pure shear along x and y (Fig. 6.25). It is realised from the Mohr stress
circle that:
σ 1 ¼ σ 3 ¼ τxy ð6:66Þ

and from the strain circle:

ε1 ¼ ε3 ¼ εxy ð6:67Þ

The following equation relating strain and stress is obtained by applying Hooke’s generalised law
(σ 2 ¼ 0):
1
ε1 ¼ ðσ  υσ 3 Þ ð6:68Þ
E 1

Insertion of Eqs. (6.66) and (6.67) in Eq. (6.68) yields:

 
1þυ
εxy ¼ τxy ð6:69Þ
E

By comparing Eqs. (6.65) and (6.69), we get:

1 1þυ E
¼ )G¼ ð6:70Þ
2G E 2ð1 þ υÞ

Hence, of the three parameters (E, G and υ) that describe the elastic properties of the isotropic
solid, only two of them are independent. For a Hookean solid with a typical Poisson’s ratio (υ ¼ 0.4),
the shear modulus is E/2.8.
Another frequently used quantity describing the volumetric changes can be derived from the
generalised Hooke’s law (Eq. 6.64):
1  
εx þ εy þ εz ¼ ð1  2υÞ σ x þ σ y þ σ z ð6:71Þ
E

The average normal stress (σ) at a point is defined as:

1 
σ¼ σx þ σy þ σz ð6:72Þ
3

As shown above, the volume strain (ΔV/V0) is:

ΔV=V 0 ¼ εx þ εy þ εz ð6:73Þ

The bulk modulus (κ) is defined according to (Bowman 2004):

σ
κ¼ ð6:74Þ
ΔV=V

By inserting Eqs. (6.72) and (6.73) in Eq. (6.71):

1 σ E
ΔV=V 0 ¼ ð1  2υÞ3σ ) ¼ ð6:75Þ
E ΔV=V 0 3ð1  2υÞ

Hence,
6.6 The Three-Dimensional View for Modulus and Strain 295

E
κ¼ ð6:76Þ
3ð1  2υÞ

The bulk modulus expresses the volume strain at a point in terms of an average normal stress
existing at that point. Materials with Poisson’s ratio equal to 0.5 have infinite bulk modulus. The bulk
modulus permits determination of the dilatation (change in volume) for a hydrostatic state. It is,
therefore, extensively used in fluid mechanics. It is also used in subsea applications where the
hydrostatic pressure from water plays a role.
The bulk modulus is larger than the tensile modulus for materials with intermediate or large
Poisson’s ratios, as shown in Eq. (6.76), and especially for rubbers the difference in bulk and tensile
moduli is several orders of magnitude (Table 6.1). This means that even though rubbers have very low
tensile and shear moduli, they are almost incompressible when exposed to hydrostatic pressures. As
observed in Table 6.1, and shown above, the shear modulus is the lowest modulus.
The tensile and shear moduli are determined from the slope of the stress–strain curve. Figure 6.26
shows ways of determining the tensile modulus from a stress–strain curve. Because there is normally
a curvature of the stress–strain curve, the secant and tangent moduli must be specified at a given
strain. Figure 6.26 illustrates that the initial modulus is higher than the secant modulus, which is
higher than the tangent modulus at the position of the secant.
There are ways of obtaining the modulus other than from stress–strain data (van Krevelen 1990).
The shear modulus is related to the velocity of elastic transverse waves in the material. Longitudinal
sound waves are associated with a volume change in the material and hence are related to the bulk
modulus. The bulk modulus can also be obtained from the polymer surface tension, the specific heat
of crystals and the heat of sublimation of crystals. The bulk modulus is the inverse of the compress-
ibility and can be obtained from isothermal volume (V)–pressure (p) data:

Table 6.1 Mechanical Material υ E (GPa) G (GPa) κ (GPa)

properties of selected
Water 0.5 0 0 2.0
materials at room
temperature (Van Krevelen Natural rubber 0.49 0.001 0.0003 2.0
and Te Nijeuhuis 2009) LDPE 0.49 0.2 0.07 3.3
Nylon 0.44 1.9 0.7 5.0
PMMA 0.4 3.2 1.1 5.1
PS 0.38 3.4 1.2 5
Epoxy 0.4 >2.5 0.9 6.4
Steel 0.28 220 80 160
Glass 0.23 60 25 37
Graphite whiskers 0 1000 500 333

Fig. 6.26 The definition of

the initial (Ei), secant (Es)
and (Et) tangent moduli
296 6 Mechanical Properties

1
κ¼   ð6:77Þ
1 ∂V
V ∂p T

The bulk modulus increases slightly with increasing crystallinity (Van Krevelen and Te Nijeuhuis
2009). High crosslink density, stiff chains and strong intermolecular forces increase the bulk
modulus, the latter evidenced by the fact that PVC has a higher bulk modulus than polyethylene,
even though the latter has high crystallinity (Van Krevelen and Te Nijeuhuis 2009).

6.7 Energy During Deformation

6.7.1 Energy Stored During Elastic Deformation

Let us now consider the energy conservation during loading and unloading for a Hookean elastic
solid. The work done on a specimen in stretching from l0 to l1 is given by:
ðl1 ðl1 ðl1
W¼ f dl ¼ σAdl ¼ EεAdl ð6:78Þ
l0 l0 l0

where A is the cross-sectional area. The strain (ε) is defined as (l  l0)/l0. The following expression is
obtained after insertion of the definition of strain into Eq. (6.78):
ðl1 2
AE AE Al E ðl1  l0 Þ
W¼ ðl  l0 Þdl ¼ ðl1  l0 Þ2 ¼ 0 2
ð6:79Þ
l0 2l0 2 l0
l0

By considering that the volume of the specimen is given by Al0 and by applying the definition of
strain, the following equation is obtained:

W Eε21
¼ ð6:80Þ
V 2

where ε1 is the strain corresponding to the specimen length l1. Eq. (6.80) may be rewritten in another
form by considering Hooke’s law:
W σε1
¼ ð6:81Þ
V 2

A graphical interpretation of Eq. (6.81) is provided in Fig. 6.27. The area under the stress–strain
curve is equal to work per unit volume of stressed specimen. The work done on the system is stored
elastically and is equal to the increase in internal energy. The atoms are displaced from their
‘equilibrium’ positions, and this is the principal reason for the increase in internal energy. When
the specimen is unloaded, the specimen is doing work on the surroundings. No heat is evolved during
loading or unloading. The mechanical energy is thus fully conserved in the loading–unloading cycle.
6.7 Energy During Deformation 297

Fig. 6.27 Stress–strain curve of Hookean solid. The shaded area is equal to the work done on the specimen at strain ε1

6.7.2 Energy Transformed During Viscous Flow and Viscoelastic Deformation

Let us now calculate the work done on a viscous liquid (Newtonian liquid). It is assumed that a
constant stress is acting on the specimen:
ðl1 ðt1 ðt1
dl dε
W ¼ f dl ¼ f dt ¼ σAl0 ε_ dt ð6:82Þ
dε dt
l0 0 0

The stress and strain rate are both constant at small strains, and the integral in Eq. (6.82) is readily
solved:
W
¼ σ ε_ t1 ¼ η_ε2 t1 ð6:83Þ
V

The work done on the specimen is fully transformed into heat. The temperature rise of the
specimen under adiabatic conditions can be calculated if specific heat and material density are
known. Purely viscous liquids are typically based on low molar mass species. Polymeric melts
show a combination of viscous and elastic behaviour. The viscous component originates from the
friction between ‘sliding’ chains, whereas the elastic part is due to the entropy-elastic stretching of the
chain segments between entanglements (Fig. 6.28).
The findings about energy conservation may be summarised as follows: elastic materials conserve
the mechanical energy, whereas viscous materials show complete and irreversible conversion of the
work to heat. The work, which is either stored elastically or released in the form of heat, can be
expressed in the following way:
ðl ðl ð
εmax
dl
W ¼ f dl ¼ A σdl ¼ A σ dε ð6:84Þ
dε
l0 l0 0

The derivate (dl/dε) ¼ l0, after insertion in Eq. (6.84), yields:

ð
εmax ð
εmax

W ¼ Al0 σdε ) W=V ¼ σdε ð6:85Þ

0 0
298 6 Mechanical Properties

Fig. 6.28 (a) Origin of the

viscous component by
friction between two
polymer chains; the upper
chain is sliding past the
lower chain. (b) Origin of
elastic component by the
trapping of the chain (thick
line) by two chain
entanglements (marked
with E in the figure). The
piece of the chain between
the two entanglement
points is extended with an
associated entropy-elastic
force

Fig. 6.29 General method for the assessment of work performed on a stressed specimen

The total work performed on the system per unit volume is simply the area under the stress–strain
curve (Fig. 6.29).
It must be realised that the method illustrated in Fig. 6.29 yields the total work done on the
specimen. A part of this work is recoverable, but a significant part may be lost as heat by the viscous
processes.

6.8 Oriented Polymers and Multiaxial Stresses

For a multiaxial stress and strain state, the modulus or compliance tensor is used to correlate stress
and strain (Ward 1985). Eq. (6.86) defines the compliance tensor for a uniaxially oriented polymer
(orientation direction is ‘33’), and from this the tensile and shear strains (εi) in each of the orthogonal
principal directions can be calculated at a given set of stresses (σ j ).
6.8 Oriented Polymers and Multiaxial Stresses 299

S11 S12 S13 0 0 0

S12 S22 S23 0 0 0
S13 S23 S33 0 0 0
εi ¼ σj ð6:86Þ
0 0 0 S44 0 0
0 0 0 0 S44 0
0 0 0 0 0 S66

The tensile strains in the principal directions are:

ε1 ¼ S11 σ 1 þ S12 σ 2 þ S13 σ 3 ð6:87Þ

ε2 ¼ S12 σ 1 þ S22 σ 2 þ S13 σ 3 ð6:88Þ

ε3 ¼ S13 σ 1 þ S23 σ 2 þ S33 σ 3 ð6:89Þ

and the shear strains are:

ε13 ¼ ε23 ¼ S44 σ 13 ð6:90Þ

ε12 ¼ S66 σ 12 ð6:91Þ

The moduli in the orientation direction (L), the transverse direction (T) and the associated shear
moduli are obtained from the compliances:
1 1 1 1
EL ¼ , ET ¼ , GLT ¼ , GTT ¼ ð6:92Þ
S33 S11 S44 S66

Equations (6.87, 6.88 and 6.89) show the fact that provided Poisson’s ratio is not zero, stresses
perpendicular to the direction of interest must be considered. In fact, if Poisson’s ratio is 0.5, i.e. no
volume change occurs during stretching, the equations show that the strain in the direction of interest
will be zero if all principal stresses are equal.
It is tempting to ask: ‘What are the properties of a perfectly aligned polymer?’ Large samples
showing perfect orientation have not been prepared. The properties are either measured on a very
small piece, for example, a single crystal (crystallite), or are calculated from other measurements. The
highest possible axial elastic modulus of a polymer can be experimentally determined from X-ray
diffraction data by measuring the change in the distance between crystal planes in accurately stressed
crystallites. It can also be calculated from data obtained by vibrational spectroscopy. The latter
provides data on the elastic constants of the individual bonds deformed by stretching and the bond
angle deformation, and a theoretical modulus can be calculated. The maximum elastic moduli (along
the chain axis) for a few polymers are given in Gedde and Hedenqvist (2019b).
A simple but very useful principle is to consider that the elastic deformation of a single molecule
occurs by three possible mechanisms: bond stretching, bond angle deformation and torsion about a σ
bond. The elastic constants for these three deformation mechanisms are very different. The following
comparative values may be used: 100 (bond stretching), 10 (bond angle deformation) and 1 (torsion
about a σ bond).
300 6 Mechanical Properties

Fig. 6.30 Young’s modulus of cold-drawn polyethylene as a function of the post-necking molecular draw ratio. The
molecular draw ratio immediately after necking is close to 8. Drawn after data from Sadler and Barham (1990)

The high axial elastic modulus of polyethylene and polyamide 6 is due to the fact that these
polymers have a preferred conformation that is fully extended, i.e. all-trans. The elastic deformation
is caused by the deformation of bond angles and by bond stretching, both showing high elastic
constants. Isotactic polypropylene and polyoxymethylene crystallise in helical conformations and,
therefore, exhibit a maximum stiffness that is only 20% of the maximum stiffness of the all-trans
polymers. The elastic deformation of a helical chain involves, in addition to the deformation of bond
angles and bond stretching, deformation by torsion about the σ bonds. The latter shows a very low
elastic constant. The relatively low maximum stiffness of the fully extended (all-trans) poly(ethylene
terephthalate) is due to the low molecular packing of this polymer. The three-dimensional covalent
bond structure of diamond leads to its extraordinarily high stiffness by largely inhibiting bond angle
deformation (Fig. 6.30).
The transverse modulus is determined by the weak secondary bonds and for this reason is only a
small fraction of the axial (along chain) elastic modulus. X-ray diffraction indicates the following
values for linear polyethylene: Ec  300 GPa and Ea  Eb  3 GPa. Polar polymers with stronger
intermolecular bonds show higher transverse moduli, 3–12 GPa.
Let us discuss in more detail the elastic properties of highly (ultra) oriented polymers. It is known
that the elastic modulus shows good correlation with the draw ratio. For a given polymer, this
relationship is specific. The correlation between elastic modulus and the Herman’s orientation
functions f, fc (for crystal) or fa (for amorphous phase) is not unique in that sense. It may be argued
on the basis of the data of Sadler and Barham (1990) that the relevant correlation is between Young’s
modulus and molecular draw ratio (Fig. 6.30). A high c-axis orientation is already achieved at
relatively low draw ratios, and it increases only moderately on further drawing. The molecular
draw ratio assesses the extension of the end-to-end vectors and the axial molecular continuity of
the microfibrils. It is believed that long and thin molecular connections, amorphous or crystalline,
constitute a ‘third’ strongly reinforcing component in an oriented semicrystalline polymer.
6.9 Linear Viscoelasticity 301

6.9 Linear Viscoelasticity

6.9.1 The Spring and the Dashpot

At low strains (typically below 1%), the mechanical behaviour of a polymer can be described by a
combination of one or several springs and dashpots. The polymer is said to be linear viscoelastic. The
term linear refers to the stress being linearly dependent on strain history (Banks et al. 2011), and
linear viscoelasticity will be the focus in the following couple of sections (refer also to Ward 1985;
Banks et al. 2011). The spring is characterised by a linear relationship between stress and strain where
modulus is the proportionality constant (Fig. 6.31):
σ
ε¼ ð6:93Þ
E

For the dashpot, there is a linear relationship between stress and strain rate where the size of the
strain rate is determined by the viscosity (η, Fig. 6.31):
dε σ
¼ ð6:94Þ
dt η

The spring alone can be used for describing the mechanical response of a glassy polymer at low
temperature, i.e. where the material is essentially elastic. The dashpot alone can describe a low molar
mass polymer melt.

6.9.2 The Maxwell (iso-Stress) Element

The Maxwell element consists of the two elements placed in series (Fig. 6.32) (Ward 1985; Banks
et al. 2011). The equation describing the time-dependent strain under a constant stress (σ o), i.e. under
creep, is:
σ0 σ0
ε ¼ εs þ εd ¼ þ t ð6:95Þ
E η

where indices s and d refer to the spring and the dashpot, respectively.
The creep behaviour described by the Maxwell element is not likely to occur for polymers since it
predicts a constant strain rate and an infinite strain at infinite time (Fig. 6.32). The Maxwell element
is, however, more suitable for predicting stress relaxation of polymers (Fig. 6.33). At constant strain:

Fig. 6.31 The spring (left) and the dashpot (right) and their stress–strain behaviour
302 6 Mechanical Properties

Fig. 6.32 The Maxwell

element and its creep
behaviour

Fig. 6.33 The Maxwell

element and its stress
relaxation behaviour

dε dεs dεd 1 dσ σ
¼ þ ¼ þ ¼0 ð6:96Þ
dt dt dt E dt η

which may be rewritten as:

ðσ ðt
dσ E
¼ dt ð6:97Þ
σ η
σo 0

For solids, it is better to use a relaxation time instead of viscosity. By inserting the relaxation time
in Eq. (6.97), followed by integration, the stress as a function of time is:
 
t
σ ¼ σ 0 exp  ð6:98Þ
τ

The ability for a chain segment to relax an imposed stress by reorientation depends on its size.
Since there is a distribution of chain segment lengths between physical or chemical entanglements,
there is also a distribution of ‘effective stiffness’ (Ei) and load-bearing segments with relaxation times
(τi), rather than a single modulus and a single relaxation time:

t t
σ ¼ σ 1 exp  þ σ 2 exp  þ :... ¼
τ1 τ2
X
1 ð6:99Þ
t t t
ε E1 exp  þ E2 exp  þ :... ¼ε Ei exp 
τ1 τ1 i¼1
τi
6.9 Linear Viscoelasticity 303

Fig. 6.34 The fit of stress/force relaxation data of a hyperbranched polymer using three terms in Eq. (6.99) (solid line).
As a comparison, the fit using a single term is also displayed (broken line). Drawn after Hedenqvist et al. (2000)

Equation (6.99) represents a series of Maxwell elements coupled in parallel. It is assumed here that
all chain segments experience the same strain. In Fig. 6.34, the stress relaxation data of a
hyperbranched polymer is fitted by using the sum of three terms in Eq. (6.99). If the modulus and
relaxation time distributions are continuous, Eq. (6.99) can be represented by an integral and by
dividing the right and left sides by the strain, the Prony series for tensile modulus is obtained (Nielsen
and Landel 1994):
ð
1
 
t
Eð t Þ ¼ EðτÞ exp  dτ ð6:100Þ
τ
0

The stretched exponential, also referred to as the Kohlrausch–Williams–Watts (KWW) function,

which is a modification of Eq. (6.98), is often used to describe stress relaxation data:
   
t β
σ ¼ σ 0 exp  ð6:101Þ
τ

where ß is the width parameter (0 < ß 1) (Kohlrausch 1847 and Williams and Watts 1970). The
lower the ß value, the broader is the stress relaxation spectrum and, for ß ¼ 1, the stress relaxation
conforms to the Maxwell equation (Eq. (6.98)).
The dynamic mechanical response is now derived for a Maxwell element. Figure 6.35 shows the
sinusoidal strain response to an applied sinusoidal stress. The phase difference between stress and
strain is the damping (δ) which is related to the physical state and the molecular mobility of the
polymer. The physical meaning of damping may be understood by considering the dashpot containing
a viscous liquid. In order to lift the shaft of the dashpot, a force is required. The higher the liquid
viscosity, i.e. the higher the damping, the longer it takes for the force/stress to initiate a movement of
the shaft, i.e. the larger is the phase shift between stress and strain.
304 6 Mechanical Properties

Fig. 6.35 Dynamic mechanical stress and strain. The phase shift (δ) is 22.5 . The stress and strain amplitudes are
arbitrarily chosen

A positive damping (δ) means that the stress is ahead of the strain (Fig. 6.35). Hence, the cyclic
stress and strain can be written as (Ward 1985):
ε ¼ ε0 sin ðωtÞ ð6:102Þ

σ ¼ σ 0 sin ðωt þ δÞ ð6:103Þ

where index o refers to the amplitude and t and ω are time and angular frequency, respectively. The
dynamic mechanical characteristics of the Maxwell element are obtained by writing the Maxwell
constitutive equation in complex form:

dε∗ 1 dσ ∗ σ ∗
¼ þ ð6:104Þ
dt E dt Eτ

where the complex strain and stress are defined as (Ward 1985):

ε∗ ¼ ε0 ½ cos ðωtÞ þ i sin ðωtÞ ¼ ε0 exp ðiωtÞ ð6:105Þ

σ ∗ ¼ σ 0 ½ cos ðωt þ δÞ þ i sin ðωt þ δÞ ¼ σ 0 exp ½i ðωt þ δÞ ð6:106Þ

The derivatives of these are:

dε∗
¼ iω ε0 exp ðiωtÞ ð6:107Þ
dt

and

dσ ∗
¼ iω σ 0 exp ½iðωt þ δÞ ¼ iω σ 0 0 exp ðiωtÞ ð6:108Þ
dt

Inserting Eqs. (6.106), (6.107) and (6.108) in Eq. (6.104) yields

 
iωτ
σ00 ¼ Eε0 ð6:109Þ
1 þ iωτ

which, by eliminating the complex number in the denominator, can be rewritten as:
6.9 Linear Viscoelasticity 305

Fig. 6.36 Complex

modulus in the complex
plane

ω2 τ 2 ωτ
σ00 ¼ þi Eε0 ð6:110Þ
1 þ ω2 τ 2 1 þ ω2 τ 2

which, inserted in Eq. (6.108) followed by integration, becomes:

ω2 τ 2 ωτ
σ∗ ¼ þi Eε0 exp ðiωtÞ ð6:111Þ
1 þ ω2 τ 2 1 þ ω2 τ 2

The modulus determined from dynamic mechanical data, i.e. the complex modulus, is:
σ∗ σ0
E∗ ¼ ¼ exp ði δÞ ð6:112Þ
ε∗ ε0

which is obtained from Eqs. (6.105) and (6.106). The storage tensile modulus (E0 ) which is in phase
with the strain and the loss tensile modulus (E00 ) which is 90 out of phase with the strain are defined
as (Sharma et al. 2018):
σ0
E∗ ¼ ½ cos δ þ i sin δ ¼ E0 þ i E00 ð6:113Þ
ε0

From Eq. (6.113) and Fig. 6.36, it is observed that the ratio of E00 and E0 is:
E00
tan δ ¼ ð6:114Þ
E0

that is referred to as the loss tangent, loss factor or damping. The ratio of the loss and storage tensile
moduli is also related to the resonant frequency ( fR) of a specimen (Fig. 6.36):
1 E00
/ 0 ð6:115Þ
fR E

The complex modulus for the Maxwell element may now be derived from Eqs. (6.105) and
(6.111):

ω2 τ 2 ωτ
E∗ ¼ Eþi E ¼ E0 þ i E00 ð6:116Þ
1 þ ω2 τ 2 1 þ ω2 τ 2

E is here a model constant, and it is the storage modulus and the loss modulus that are extracted
from the experiment. Finally, the loss tangent is obtained for the Maxwell element as:
306 6 Mechanical Properties

Fig. 6.37 The storage and

loss moduli and the loss
tangent as a function of
angular velocity for a
single Maxwell element

E00 1
tan δ ¼ 0 ¼ ωτ ð6:117Þ
E

Hence, the loss tangent decreases monotonically with increasing angular velocity for a spring and
a dashpot mounted in series, which is due to the motion of the dashpot being ‘frozen in’ at high
frequency. The resulting moduli and loss tangent are shown in Fig. 6.37. As in the case of stress
relaxation, the dynamic mechanical data of the real polymer has to be fitted with a series of Maxwell
elements mounted in parallel, which yields the following storage and loss moduli:

X
1 ð
1
0 ω2 τ i 2 ω2 τ 2
E ðωÞ ¼ Ei ¼ EðτÞ dτ ð6:118Þ
i¼1
1 þ ω2 τ i 2 1 þ ω2 τ 2
0

X
1 ð
1
00 ωτi ωτ
E ð ωÞ ¼ Ei ¼ Eð τ Þ dτ ð6:119Þ
i¼1
1 þ ω2 τ i 2 1 þ ω2 τ 2
0

6.9.3 The Voigt–Kelvin (iso-Strain) Element

If the spring and the dashpot are placed in parallel, the Voigt–Kelvin element is obtained (Fig. 6.38)
(Ward 1985).
This element cannot describe stress relaxation because the dashpot counteracts an initial instanta-
neous strain. The spring and the dashpot experience the same strain, and the total stress is the stress
over both elements:
dε
σ ¼ σ s þ σ d ¼ Eε þ η ¼ σ0 ð6:120Þ
dt

In order to derive the creep behaviour, Eq. (6.120) is first rewritten as:
6.9 Linear Viscoelasticity 307

Fig. 6.38 The

Voigt–Kelvin element and
its creep behaviour

dε E σ E σ
þ ε ¼ 0 , ε_ þ ε ¼ 0 ð6:121Þ
dt η η η η

which is solved by applying the solution in the form:

ε ¼ A þ B exp ðCtÞ ð6:122Þ

where A, B and C are constants. By implementing Eq. (6.122) and its derivative into Eq. (6.121), the
following relationship is obtained:

σ0 E
ε¼ þ B exp  t ð6:123Þ
E η

which, by applying the initial condition: ε(0) ¼ 0, finally yields:

h  i
σ t
ε ¼ 0 1  exp  ð6:124Þ
E τ

where τ ¼ η/E is the retardation time. It should be mentioned that viscoelastic materials show, in
contrast to a Hookean elastic material, hysteresis during loading and unloading (Banks et al. 2011).
For a Voigt–Kelvin material, the stress during loading is higher than during unloading. The dynamic
mechanical response of a Voigt–Kelvin element can be derived by starting from the complex form of
Eq. (6.125):

dε∗
Eε∗ þ τ E ¼ σ∗ ð6:125Þ
dt

where the complex stress and strain are defined in Eqs. (6.105) and (6.106), and the strain rate is given
in Eq. (6.107). The stress is defined as:

σ ∗ ¼ σ 0 exp ½iðωt þ δÞ ¼ σ 0 0 exp ðiωtÞ ð6:126Þ

The implementation of Eqs. (6.105), (6.106), (6.107) and (6.126) in Eq. (6.125) yields:
σ 0 0 ¼ E ε0 ð1 þ iωtÞ ð6:127Þ

Equation (6.127) in Eq. (6.126) yields:

σ ∗ ¼ ð1 þ iωτÞ Eε0 exp ðiωtÞ ð6:128Þ

The complex compliance (D*) is the reciprocal of the complex modulus (Ward 1985):
308 6 Mechanical Properties

ε∗ ε0
D∗ ¼ ¼ exp ði δÞ ð6:129Þ
σ∗ σ0

which can be written as:

ε0
D∗ ¼ ð cos δ  i sin δÞ ¼ D0  i D00 ð6:130Þ
σ0

where D0 and D00 are the storage and loss tensile compliances. The complex compliance for the
Voigt–Kelvin element is derived by using Eqs. (6.105), (6.106) and (6.129):
ε∗ 1
D∗ ¼ ¼ ð6:131Þ
σ ∗ 1 þ iωτ

which, by elimination of the complex number in the denominator, becomes:

1 ωτ
D∗ ¼ Di D ¼ D0  i D00 ð6:132Þ
1 þ ω2 τ 2 1 þ ω2 τ 2

D is a model constant, and it is the storage tensile compliance (Eq. (6.133)) and the loss tensile
compliance (Eq. (6.134)) that are obtained experimentally:

X
1 ð
1
0 1 1
D ð ωÞ ¼ Di ¼ D ðτ Þ dτ ð6:133Þ
i¼1
1 þ ω2 τ i 2 1 þ ω2 τ 2
0

X
1 ð
1
00 ωτi ωτ
D ð ωÞ ¼ Di ¼ D ðτ Þ dτ ð6:134Þ
i¼1
1 þ ω2 τ i 2 1 þ ω2 τ 2
0

From Eqs. (6.133) and (6.134), the loss tangent for the tensile case is determined as:
D00
tan δ ¼ ¼ ωτ ð6:135Þ
D0

In a Voigt–Kelvin configuration, the damping due to the dashpot increases with increasing angular
frequency, which is also predicted by Eq. (6.135).

6.9.4 The Burgers Element and Other Linear Viscoelastic Models

If the Maxwell and Voigt–Kelvin elements are placed in series, the Burgers element is obtained
(Fig. 6.39).
The strain during creep is the sum of the strains of the Maxwell (M) and the Voigt–Kelvin
(V) elements:
 
σ σ σ t
ε ¼ 0 þ 0 t þ 0 1  exp  ð6:136Þ
EM EM τ M EV τV
6.9 Linear Viscoelasticity 309

Fig. 6.39 The Burgers

element and its creep
behaviour

Fig. 6.40 The SLS model

(with two moduli)

It is possible to predict the creep behaviour of several polymers well above the glass transition with
the Burgers element (Qausar 1989). Creep according to this element involves an instantaneous strain
on loading and a non-recoverable strain on unloading.
In the standard linear solid (SLS) model, a spring is coupled in parallel with a Maxwell element
(Fig. 6.40) (Banks et al. 2011). With this element, it is possible to predict stress relaxation and
dynamic mechanical response for a system having a finite relaxed modulus.
For polymers, the configuration in Fig. 6.40 is usually expanded with an infinite/large number of
Maxwell elements coupled in parallel. The modulus frequency dependence of a loosely crosslinked
polymer, where viscous flow is negligible, is accurately described by a spring coupled in series with a
310 6 Mechanical Properties

Fig. 6.41 The frequency

dependence of dynamic
mechanical properties of a
loosely crosslinked
elastomer exposed to a
cyclic stress

Fig. 6.42 The damping as

a function of temperature
of a crosslinked elastomer
exposed to a cyclic stress

number of Voigt–Kelvin elements (Ward 1985). The loss tangent (damping) frequency dependence
is, for this material, described by a combination of Eqs. (6.117) and (6.135). At low frequencies, the
molecule segments are able to move in phase with the cyclic stress (Fig. 6.41). The heat generated due
to intermolecular friction is low, which in turn results in low damping and a low loss modulus. The
material behaves like a rubber with a low but finite storage modulus. As the frequency increases, the
friction and damping increase. This continues until the molecule segments can no longer move with
the stress frequency. In this region, the storage modulus increases rapidly with increasing frequency.
At very high frequencies, the molecule segments are unable to move with the cyclic stress, and the
system ‘freezes in’, which again results in low damping and a low loss modulus. The polymer
behaves like a glass with a high storage modulus.
It is important to know the tan δ (damping) behaviour of the rubber when designing tires. The
magic triangle of a tire is the relationship between the three properties: abrasion resistance, wet skid
resistance and rolling resistance (Weng et al. 2020). An ideal tire has a tread with high abrasion
resistance, high wet skid resistance (good wet traction) and low rolling resistance. High abrasion
resistance increases the lifetime of the tire and minimises the generation of polluting microparticles,
and low rolling resistance decreases the fuel consumption. Low damping at ca. 40–60  C (broken
arrow in Fig. 6.42) decreases the rolling resistance (Xu et al. 2017a, b), and high damping at
ca. 0–20  C (solid arrow) increases the wet skid (increasing wet grip/decreasing braking distance)
(Rogers and Waddell 2013). High wet grip often means a high rolling resistance since a high hysteric
loss (heat release) increases ‘friction’ (Aldhufairi and Olatunbosun 2018). The abrasion resistance
increases with a decrease in the glass transition temperature (less fatigue and cutting/frictional wear)
(Klüppel 2014). The situation is actually more complex since, for a viscoelastic material, the damping
is both temperature and stress frequency (strain rate) dependent. An increase in temperature at
constant frequency has the same effect as decreasing the frequency at constant temperature. The
6.10 Correlations Between Stress Relaxation and Dynamic Mechanical Data 311

frequency effect on tire properties varies; a lower damping at a frequency of 10–100 Hz leads to
reduced rolling resistance, and an increase in damping at 1000–100000 Hz favours improved wet grip
and reduced braking distance.
The low damping (low friction) between a metal wheel and metal rail was the reason for
introducing the first railways. It was discovered that horses could pull a significantly higher load if
the wagons were running on rails than without them.

6.10 Correlations Between Stress Relaxation and Dynamic Mechanical Data

It can be of interest to transform data from one specific mechanical test to another. This may be due to
the laboratory having only one type of instrument or that time can be saved by using a specific test
relative to another. Data from fast dynamic mechanical tests may be transformed into long-term stress
relaxation or creep data (Schwarzl and Struik 1968 and Schwarzl 1969, 1970, 1971, 1975). In the
following section, an example is given where experimental stress relaxation data of semicrystalline
polycarbonate (which is not a common polymer) is transformed to dynamic mechanical modulus
data. This can further be transformed into dynamic compliance data and then static creep data (not
shown). The derivations are made in shear terms but are of course valid for tensile properties.

6.10.1 From Stress Relaxation Modulus to Dynamic Mechanical Modulus

Consider that the polymer can be described by an infinite number of Maxwell elements and a spring,
all coupled in parallel. The shear modulus can be written as (compare with Eq. (6.100)):
ð
1
 
t
GðtÞ ¼ G1 þ gðτÞ exp  dτ ð6:137Þ
τ
0

where G1 is the shear modulus at infinite time, i.e. the relaxed modulus. g(τ) is the so-called
relaxation spectrum, and the exponential term is the intensity function. The complex shear modulus
is defined as:

G∗ ðiωÞ ¼ G0 ðωÞ þ i G00 ðωÞ ð6:138Þ

where the first and second terms on the right side are the storage and loss moduli, respectively, and
they can be written as (compare with Eqs. (6.118) and (6.119)):
ð
1
0 ω2 τ 2
G ðωÞ ¼ G1 þ gð τ Þ dτ ð6:139Þ
1 þ ω2 τ 2
0

and
ð
1
00 ωτ
G ð ωÞ ¼ gðτÞ dτ ð6:140Þ
1 þ ω2 τ 2
0
312 6 Mechanical Properties

Fig. 6.43 Shear relaxation

modulus of semicrystalline
polycarbonate as a function
of logarithmic time at
155  C. Drawn from data of
Mercier and Groeninckx
(1969)

Fig. 6.44 Normalised

shear relaxation modulus of
semicrystalline
polycarbonate as a function
of logarithmic time at τ ¼ t.
Also shown are the step
function (dotted line) and
the relaxation spectrum
(dashed line) (Eq. (6.146))

Hence, the dynamic mechanical parameters can be obtained from stress relaxation if the relaxation
spectrum is known. The relaxation spectrum may be extracted from the stress relaxation curve by first
plotting the relaxation modulus as a function of logarithmic time (Fig. 6.43). The stress relaxation
strength is defined as the difference between the initial modulus and the relaxed modulus. For
semicrystalline polycarbonate at 155  C, it is ca. 800 MPa (Fig. 6.43).
For simplicity, it is worthwhile to eliminate the exponential term. This is obtained by
approximating the exponential term of the intensity function as a step function (first approximation)
or a straight line (second approximation) (Fig. 6.44) (Schwarzl and Staverman 1952). The step
function takes the value 1 for τ > t and zero for τ < t. After having normalised the modulus to be
between 0 and 1, Eq. (6.137) is now written as:
ð
1

GðtÞ ¼ gðτÞ dτ ð6:141Þ

τ¼t

The first approximation is obtained as the negative slope of the relaxation modulus at time t:
 1
∂GðtÞ ∂GðtÞ
gð τ Þ ¼ ! gð τ ¼ t Þ ¼ ð6:142Þ
∂t t ∂t

The second approximation involves approximating the exponential with a polynomial in t/τ that
has the same area under the curve as the exponential and that decays to zero over a finite range of τ/
t (Schwarzl and Staverman 1952) (Fig. 6.44):
6.10 Correlations Between Stress Relaxation and Dynamic Mechanical Data 313

Fig. 6.45 Storage (●) and

loss (○) shear moduli
calculated from
Eqs. (6.139) and (6.140)
using the relaxation
spectrum (Eq. 6.146). The
integral is approximated
here with a finite series
of sums

ð
1
 
1 t
GðtÞ ¼ gð τ Þ 1  dτ ð6:143Þ
2τ
τ¼2t

Differentiating twice with respect to t and inserting the boundaries yield:

  2
∂ GðtÞ
t
g τ¼ ¼ 2t ð6:144Þ
2 ∂t2

Since the relaxation spectrum is generally broad, it is more convenient to express Eq. (6.141) with
a logarithmic relaxation time. In doing so, the logarithmic relaxation spectrum is first defined:
HðτÞ ¼ τ gðτÞ ð6:145Þ

Consequently, the first and second approximations become:

∂GðtÞ
H ðτ Þ ¼  ð6:146Þ
∂ ln ðtÞ

  ∂GðtÞ
2
∂ GðtÞ
t
H τ¼ ¼ þ ð6:147Þ
2 ∂ ln ðtÞ ∂ ln ðtÞ2

Figure 6.45 shows G0 (ω) and G00 (ω) as a function of ln(ω) calculated from H(τ ¼ t/2) and
Eqs. (6.139) and (6.140).

6.10.2 From Dynamic Mechanical Modulus to Dynamic Mechanical

Compliance

Since the complex modulus is the ratio between stress and strain during a periodically varying strain
and the complex compliance is the ratio of strain and stress during a periodically varying stress, they
are inversely related (Schwarzl and Struik 1968):
314 6 Mechanical Properties

1
J ∗ ðiωÞ ¼ ð6:148Þ
G∗ ðiωÞ

Since J* and G* are inversely related, their arguments are of opposite signs and their sizes are
inversely related (G ¼ 1/J ); hence, it follows that:
 
2 0:5
G∗ ¼ G0 þ G00 exp ðiδÞ ¼ G exp ðiδÞ ¼ G cos δ þ i G sin δ ¼ G0 þ i G00
2
ð6:149Þ

and
 
2 0:5
J ∗ ¼ J 0 þ J 00 exp ðiδÞ ¼ J exp ðiδÞ ¼ J cos δ  i J sin δ ¼ J 0  i J 00
2
ð6:150Þ

Thus
G0 J 0 G00 J 00
cos δ ¼ ¼ , sin δ ¼ ¼ ð6:151Þ
G J G J

and
G0 G0 G00 G00
J0 ¼ J ¼ , J 00 ¼ J ¼ 2 ð6:152Þ
G G2 G G

By replacing G with G0 and G00 (see Eq. (6.149)), the storage and loss compliances can be extracted
from the storage and loss moduli:
G0
J0 ¼ ð6:153Þ
G þ G00
02 2

and
G00
J 00 ¼ ð6:154Þ
G þ G00
02 2

6.11 Boltzmann Superposition Principle

Within the linear viscoelastic strain region, it is possible to apply the Boltzmann superposition
principle (BSP) to, for example, a specimen exposed to a series of different stress levels (Ward
1985). This means that if a specimen is exposed to two separate stresses, σ1 and σ2, the total strain is a
simple sum of the strains originating from the separate stresses, i.e. ε(t)1 and ε(t)2:
εðtÞ ¼ ε1 ðtÞ þ ε2 ðtÞ ð6:155Þ

Hence, the stresses give independent and additive contributions to strain. The creep compliance
can thus be expressed as:
6.11 Boltzmann Superposition Principle 315

Fig. 6.46 Three stress

contributions at t1, t2 and t3
yield a total strain ε(t).
Drawn from McCrum et al.
(1997)

ε1 ðtÞ ε2 ðtÞ
DðtÞ ¼ ¼ ð6:156Þ
σ1 σ2

If a constant stress Δσ o is applied at t ¼ 0, then:

ε0 ðtÞ ¼ Δσ 0 DðtÞ ð6:157Þ

If an additional stress Δσ 1 is applied at t1, it gives rise to an additional strain at time t:

ε1 ðtÞ ¼ Δσ 1 Dðt  t1 Þ ð6:158Þ

The total strain becomes:

εðtÞ ¼ Δσ 0 DðtÞ þ Δσ 1 Dðt  t1 Þ ð6:159Þ

If there is a sequence of stresses (Fig. 6.46), the total strain according to BSP is:
εðtÞ ¼ ε1 ðtÞ þ ε2 ðtÞ þ ε3 ðtÞ þ . . . ¼ Δσ 1 Dðt  t1 Þ þ Δσ 2 Dðt  t2 Þ þ Δσ 3 Dðt  t3 Þ þ . . . ð6:160Þ

Hence, the strain after n tensile stress steps is:

X
n
εð t Þ ¼ Δσ i Dðt  ti Þ, t > tn ð6:161Þ
i¼1

If the stress is changed continuously with time under creep conditions, then Eq. (6.161) may be
written in integral form:
ðt
∂σ
εðtÞ ¼ Dðt  t0 Þ 0 dt0 ð6:162Þ
0 ∂t
316 6 Mechanical Properties

Fig. 6.47 Tensile 50

stress–strain data at 20  C
for an ethylene–butene 40

True stress (MPa)

copolymer at a strain rate of
0.05 mm min1 (●) and 30
5 mm min1 (○). Drawn
after Gaucher-Miri et al.
20
(1997)
10

0
0 0.5 1 1.5 2 2.5
True strain (%)

where t’ is the time when the stress σ(t’) is applied. In a similar way, the stress relaxation where the
strain changes continuously with time is given by:
ðt
∂ε
σ ðtÞ ¼ Eðt  t0 Þ 0 dt0 ð6:163Þ
0 ∂t

where E(t-t’) is the relaxation modulus in the time interval t-t’. ε(t’) is the strain applied at time
t’ (0 < t’ < t).

6.12 The Influence of Strain Rate on the Viscoelastic Properties

The mechanical response of a polymer is, because of viscoelastic reasons, dependent on the rate of
deformation or the time of loading, i.e. the time scale of the experiment (Fan et al. 2017). If a loosely
crosslinked polymer is stretched at infinite rate, it will behave like a glass. The higher the rate of
deformation, the less is the time allowed for stress relief through chain segment reorientation and
disentanglement. This is illustrated in Fig. 6.47 where an ethylene–butene copolymer is tensile tested
at two different strain rates. Note that the initial modulus (initial slope of the curve), as well as the
yield stress, is higher at the higher strain rate.

6.13 The Influence of Temperature on the Viscoelastic Properties:

Time–Temperature Shifting and Superposition

The mechanical properties of a viscoelastic material are generally strongly temperature dependent.
Thus, it is necessary to measure the properties at the temperature of interest or to use extrapolation
methods. It is necessary, for practical reasons, to determine long-term properties by short-term
experiments (at elevated temperature). Figure 6.48 displays how the time–temperature superposition
principle works to extrapolate short-term creep compliance data to very long times. There is no real
mystery to this: most processes are accelerated by elevation of temperature. The creep compliance
taken at a temperature T1 is shifted by log aT along the log time axis to the reference temperature
(Tref). This horizontal shift is commonly referred to as the horizontal shift and can be written as:
6.13 The Influence of Temperature on the Viscoelastic Properties. . . 317

Fig. 6.48 Creep

compliance (D) at three
different temperatures
(T1 > T2 > T3). The high
temperature data (T2 and
T3) are first shifted
vertically and then
horizontally, yielding the
master curve at T3

Fig. 6.49 Creep

compliance as a function of
time at constant
temperature (log–log
diagram). The unrelaxed
compliance (Du) refers to
the glassy state, and the
relaxed compliance (Dr)
refers to the rubbery state

τ2
aT ¼ ð6:164Þ
τ1

where τ1 and τ2 are relaxation or retardation times at, respectively, T1 and T2 (Ward 1985). If dynamic
mechanical data is considered, the ratio in angular frequency can be written as (note that indices 1 and
2 appear at opposite positions compared to in Eq. (6.164); frequency is the reciprocal of time):
ω1
aT ¼ ð6:165Þ
ω2

Before executing the horizontal shift, a vertical shift should be conducted. The vertical shift is
performed by multiplication by a factor of Tref ρref/(T1 ρ1). The vertical shift is given an explanation
by the illustration displayed in Fig. 6.48 (note that compliance is the opposite to modulus). The
relaxed rubber modulus is, according to the rubber elasticity theory, proportional to the product ρT
(Gedde and Hedenqvist 2019a). This is the theoretical foundation behind the vertical shift factor.
The horizontal shift shows, in the case of the glass transition, very special temperature depen-
dence. The curve displayed in Fig. 6.49 is obtained by first vertical shifting (just a little bit) and then
horizontal shifting to maximise the overlap to obtain the shift factor (log aT).
The temperature dependence of the shift factor displayed in Fig. 6.50 is typical for a polymer near
its glass transition temperature (Tg). This was discovered a long time ago by Vogel and Fulcher (Liu
et al. 2006) who suggested the following equation:
318 6 Mechanical Properties

Fig. 6.50 The horizontal

shift factor as a function of
the reciprocal temperature
(often denoted as
‘Arrhenius diagram’). The
shift factor at the reference
temperature (Tref) is
obviously 0. The curve
approaches asymptotically
1/T0; this kind of curve is
typical for the glass
transition

B
aT ¼ exp ð6:166Þ
T  T0

where B is a constant and T0 is a temperature at which all segmental mobility stops. The latter is, for
many polymers, approximately 50 K below the glass transition temperature. Later, comprehensive
work by William, Landel and Ferry yielded the WLF equation (named after the authors) (Eq. 6.167)
(Liu et al. 2006). This expression turns out to be consonant with the Vogel–Fulcher equation but,
provided that the glass transition temperature is used as the reference temperature for the shifting,
contains ‘universal’ constants:
  
17:44 T  T g
aT ¼ exp ð6:167Þ
51:6 þ T  T g

The temperature dependence of the shift factor (log aT) for the sub-glass processes (well below the
glass transition temperature) is different and simpler:
 
ΔE
aT ¼ exp ð6:168Þ
RT

where ΔE is the activation energy of the relaxation process and R is the gas constant. Hence, the sub-
glass processes have a temperature dependence according to the Arrhenius law. Explicitly, the shift
factor may now be obtained as:

τ2 ΔE 1 1
ln ðaT Þ ¼ ln ¼  ð6:169Þ
τ1 R T2 T1

6.14 Non-linear Viscoelasticity as Illustrated with Creep Behaviour

There is not usually a sharp boundary between linear (LVE) and non-linear viscoelastic behaviour
(Starkova and Aniskevitch 2007). The non-linear behaviour is usually observed as a deviation from a
linear stress–strain curve when using creep isochrones (constant time data) or a deviation from a
linear relationship between the logarithm of stress over strain rate and logarithmic time using constant
strain rate data. Using a defined deviation from linearity (e.g. 3%), it is possible to define the stress
6.14 Non-linear Viscoelasticity as Illustrated with Creep Behaviour 319

Fig. 6.51 A creep curve at

constant stress and I II III
temperature including the
primary (I), secondary
(II) and tertiary (III) stages.

Strain
Also shown is the initial
elastic strain (a)

a
Time

and strain above which non-linear behaviour occurs. The non-linear viscoelastic behaviour is not
simply determined by a certain stress (σ LVE) and strain (εLVE) limit; it also depends on time. In
addition, stress and strain are interrelated. One way of defining a time-independent threshold for the
onset of non-linearity is to use a threshold deformation energy: WLVE ¼ C • (σ LVE • εLVE)/2, where C
defines the deviation from pure elastic deformation (Starkova and Aniskevitch 2007). In this case, the
time-dependent stress and strain limits cancel out. In the remainder of this section, we will focus on
creep under non-linear viscoelastic behaviour.
Most materials experience creep at elevated temperature or stress. Polymers, however, are
especially creep-sensitive due to the relatively low degree of molecular packing and weak intermo-
lecular forces. If a polymer component is loaded with a constant stress and the deformation is
monitored as a function of time until fracture occurs, a three-stage curve is often observed
(Fig. 6.51) (Pennings et al. 1994).
After an initial immediate elastic deformation, the primary creep stage occurs. The initial part of
this stage is characterised by linear viscoelastic behaviour that gradually turns into non-linear
viscoelastic behaviour. The secondary creep stage is characterised by a constant creep rate. The
material is here mainly viscous, and deformation is plastic, i.e. if the material is unloaded, deforma-
tion is mainly non-recoverable. In the tertiary stage, creep accelerates, and the material is damaged
through, for example, the formation of a large number of small voids (crazing). Finally, during
uniaxial tension, local necking or fracture occurs. The number of creep stages that occur depends on
the stress level (Landel and Nielsen 1993). At low stresses, only the primary stage is observed,
whereas at higher stress levels, the secondary and the tertiary stages occur. In the following part of the
chapter, only the primary stage will be considered since it is the region of most practical interest. If
the secondary or tertiary stages occur during service of a product, the designer has clearly failed with
the material selection. The primary creep stage is usually displayed with a logarithmic time scale
since it occurs over several decades (Fig. 6.52), which makes the initial stage more visible. The major
part of the primary stage is characterised by complete recovery.
In polymer products exposed to a constant stress, it is, as mentioned above, necessary to ensure
that the secondary and tertiary stages never occur, since it would mean that the product is permanently
deformed and, in the worst case, will fail. A temporary temperature increase will increase the risk that
these stages occur and, therefore, it is necessary to understand the effects of temperature on creep
kinetics.
Non-linear viscoelastic models are normally based on the idea that the strain is composed of an
elastic part, a viscoelastic part and a plastic part (Ward 1985). The elastic contribution is simply σ/E,
and the viscoelastic strain is obtained by integrating the following equation:
320 6 Mechanical Properties

Fig. 6.52 Creep compliance of PMMA at 84  C and a stress of 0.015 MPa as a function of time (a) and logarithmic
time (b). The arrow indicates the point above which creep is both elastic and plastic. Drawn from Chèriére et al. (1997)

∂ε
¼ f ðσ, εÞ ð6:170Þ
∂t

The plastic strain is obtained by considering that its strain rate is a unique function of stress:
∂ε
¼ f ðσ Þ ð6:171Þ
∂t

Usually, the stress dependence of the viscoelastic and plastic strains is considered through simple
power laws. The total strain under creep (constant stress) may thus be written in general form as:
σ
ε¼ þ aσ n ½1  exp ðbtÞ þ cσ n t ð6:172Þ
E

where a, b, c and n are material-dependent constants. To describe the complete primary stage,
i.e. both the linear and non-linear viscoelastic parts of the creep event, several empirical potential
laws have been used. They include:
1
εðtÞ ¼ ε0 þ a t3 þ b t ð6:173Þ

which is the Andrade creep law for metals (Plazek 1960). Findley and Khosla (Findley and Khosla
1955) have suggested two creep relationships:
εðtÞ ¼ ε0 þ a tn ð6:174Þ

εðσ, tÞ ¼ ε0 sinh ðaσ Þ þ a tn sinh ðbσ Þ ð6:175Þ

and Van Holde (1957) proposed the following relationship:

1
εðσ, tÞ ¼ a þ b t3 sin ðcσ Þ ð6:176Þ

which reduces to Andrade’s equation (Eq. (6.173)) if a constant stress is used.

Schapery (1969, 1997) and Muliana and Haj-Ali (2004) used irreversible thermodynamics to
obtain a general non-linear viscoelastic model. For uniaxial isothermal conditions, excluding the
viscoplastic part, the model becomes:
6.14 Non-linear Viscoelasticity as Illustrated with Creep Behaviour 321

ðt
dg2 σ
εðtÞ ¼ g0 D0 σ þ g1 Dt ðψ  ψ 0 Þ dτ ð6:177Þ
dτ
0

where D0 and Dt are the initial and transient parts of the creep compliance.
ðt ðτ
dt0 dτ0
ψ¼ and ψ 0 ¼ ð6:178Þ
aσ aσ
0 0

are reduced times. aσ is the shift factor for the reduced times, and g0, g1 and g2 are stress-dependent
parameters. At small strains, i.e. within the linear viscoelastic region, these parameters are equal to
unity, and Eq. (6.177) turns into the Boltzmann integral (Eq. (6.162)).
By considering that the different stress contributions do not give independent contributions to
strain, the Boltzmann superposition principle can, according to the Gibbs–Rivlin theory, be extended
to the non-linear viscoelastic region (Yannas and Haskell 1971 and Ward 1985). This leads to a
multiple integral representation using the Gibbs–Rivlin theory:
ðt ðt ðt
∂σ ðτÞ ∂σ ðτ1 Þ ∂σ ðτ2 Þ
εð t Þ ¼ D1 ðt  τÞ dτ þ D 2 ðt  τ 1 , t  τ 2 Þ dτ1 dτ2 þ . . . . . . ð6:179Þ
∂τ ∂τ1 ∂τ2
1 1 1

The first term is the linear viscoelastic Boltzmann integral, and the following terms are due to joint
stress contributions (two, three, four, etc.. . . stress terms). Even though the model is generic, it finds
only limited practical use because of the great number of parameters that have to be experimentally
determined.
An internal variable approach that is less computationally difficult is mentioned by Banks
et al. (2011).
Creep data is occasionally described with a stretched exponential of the form:
   
t β
εðtÞ ¼ εo þ Δε 1  exp  ð6:180Þ
τ

where Δε is the creep intensity and τ is the retardation time. β takes values between 0 and 1 and may
vary with loading time (Chèriére et al. 1997). Additionally, logarithmic creep laws have been applied
to the first part of the primary stage (Fig. 6.52) since a power law would be too slow here (Chèriére
et al. 1997). The drawback with the semi-empirical or empirical equations is the fact that they do not
predict strain in complicated stress cases and that they do not allow for transformation of data
between creep, stress relaxation and dynamic mechanical measurements. The Boltzmann superposi-
tion principle is not applicable in the major part of the primary creep stage due to the non-linear
viscoelasticity that originates from either large strains or long creep times. This implies that a test
with the exact stress program must be performed in order to determine the creep behaviour after a
specific stress history. A possible way to overcome this relatively time-consuming experiment is to
collect isochronic stress–strain creep data. This is achieved by first collecting strain data as a function
of time at different stress levels (Fig. 6.53). From the data in Fig. 6.53, stress is plotted as a function of
strain at specific times (Fig. 6.54). Consequently, by selecting the appropriate isochronous curve, the
stiffness, obtained as, for example, the secant modulus at a given load, can be obtained at a specific
loading time. This is valuable information for a product developer.
322 6 Mechanical Properties

Fig. 6.53 Creep strain as a

function of time at different
stress levels

Fig. 6.54 Stress as a

function of strain at
different times
(isochronous curves). The
curves are created from
creep curves (Fig. 6.53)

Observe that the stress–strain curve in Fig. 6.54 is not the same as the stress–strain curves
determined from constant strain rate measurements (tensile stress–strain tests).

6.15 Short-Term Mechanical Properties of Selected Polymers

The short-term mechanical properties, i.e. properties determined from, for example, constant strain
rate (tensile) tests, of a number of selected polymers are given in Table 6.2. Included are glassy
amorphous polymers, rubbery amorphous crosslinked polymers, rubbery semicrystalline polymers
and a thermoset. They all show very different behaviour, depending on their chemistry and structure.
As stated earlier, it is possible to classify the polymers in terms of three mechanical properties. The
polymer is stiff or soft, strong or weak and tough (and/or ductile) or brittle. These properties are
quantified by their modulus, tensile strength (yield or fracture), fracture energy and fracture strain.
From Table 6.2 it is clear that, among the materials presented, PMMA, PET, PF and PS are the stiffest
materials. PMMA, PET, PC, PA6,6, POM, PF, PS and PVC are relatively strong polymers, and PET,
PC, PA6,6, PP, HDPE, LDPE, PTFE and natural rubber are ductile materials. Note that fracture
energy is not given here.
In the following section, the mechanical properties of some of the selected polymers are discussed
from a molecular and morphological point of view. The arguments are only general, and large
variations do exist due to variations in, for example, molar mass, crystallinity, humidity and degree
of orientation. Natural rubber is soft, weak and ductile (Table 6.2). It has a low modulus due to the
flexible main chain and low intermolecular forces. For the same reason, the yield or fracture stresses
6.15 Short-Term Mechanical Properties of Selected Polymers 323

Table 6.2 Short-term mechanical properties of selected polymers (Van Krevelen and Te Nijeuhuis 1990)
Polymer Yield stress (MPa) Fracture stress (MPa) Fracture strain (%) E (GPa)
Natural rubber <8 <8 >800 0.001
PTFE 13 25 200 1
LDPE 8 10 800 0.2
HDPE 30 30 600 1
PP 32 33 400 1.4
PVC 48 50 30 2.6
PC 65 6 125 2.5
PETa 50 54 275 3
PMMA 65 10 3.2
POM 65 40 2.7
PA6.6b 57 8 200 2.0
PS 50 2.5 3.4
PF 55 1 3.4
a
Amorphous polymer
b
Dry polymer

are low, and the fracture strain is high. The large ductility is a consequence of an imposed load being
rapidly relaxed by molecular rearrangements of stressed chains. Due to the fluorine atoms in PTFE
being placed symmetrically along the carbon chain, the polymer is nonpolar and the intermolecular
forces are small. The presence of the fluorine atoms that repel each other electrostatically results in a
stiff chain that crystallises in extended form in a slightly turning helix (13 repeating units in a
complete helix shift) (Corradini et al. 1987). Despite the crystallinity, the weak secondary forces
result in a material that has a high intermolecular mobility and, therefore, is soft, weak and ductile.
LDPE has similar mechanical properties to PTFE. As for PTFE, polyethylene possesses only small
intermolecular interactions. The intramolecular mobility, on the other hand, is higher than for PTFE,
probably due to the smaller hydrogen and the larger fluorine atoms. Due to the weak secondary forces
and the low crystallinity, the polymer is soft, weak and ductile. HDPE is stiffer and stronger than
LDPE simply because of the higher degree of crystallinity. The crystalline component has a higher
modulus than the amorphous component, and the thicker crystals in HDPE are more difficult to
deform (yield) than the thinner crystals in LDPE (Brooks et al. 1999).
The bulky methyl side group of PP decreases chain mobility, and compared to polyethylene, the
glass transition is approximately 50  C higher (Boyd 1984). The polymer has similar mechanical
properties to HDPE. PVC has bulky and negatively charged chlorine atoms that repel each other. The
dipolar nature of the polymer and the bulky side group reduce chain mobility, and hence the modulus
and the yield stress are higher and the ductility is lower compared to polypropylene.
Let us now consider intrinsic toughness and brittleness. Why does a plate of PMMA or polystyrene
snap into two pieces or more when bent, while polycarbonate or amorphous PET do not. The
important parameters here are molecular flexibility and amount of chain entanglements (Wu 1992).
The former can be expressed by the characteristic ratio C1. The entanglement density (ve) is
obtained as the ratio of polymer amorphous density (25  C density is considered in the ve values
below) divided by the molar mass between entanglement junctions. A low characteristic ratio
combined with a high chain entanglement density is synonymous with toughness and ductility.
Although polycarbonate is glassy at room temperature, it is among the toughest polymers. The
polymer becomes tough already at the ß-transition at ca. –200 to 150  C, the exact position
depending on the measuring method (Ward 1985; Ramsteiner 1983). Polymers with stiff main
chain units joined by flexible groups/bonds, as in the case of polycarbonate, have low characteristic
324 6 Mechanical Properties

Fig. 6.55 Molecular structure of PC. Carbon, oxygen and hydrogen atoms are grey, striped and white, respectively.
The chain repeating unit can be considered as having two rotational stiff ‘virtual’ subunits separated by the carbonate
carbon and the skeletal carbon between the two aromatic units, or alternatively by the rotational subunits ((C¼O)-O-
and -O-C6H6-C(CH3)- (Wu 1992)

Fig. 6.56 Molecular

structure of PET. Atom
types the same as for PC

Fig. 6.57 Molecular

structure of PMMA. Atom
types are the same as
for PC

ratios (Fig. 6.55). For PC, C1 ¼ 2.4 (close to a tetrahedral skeletal bond chain with freely rotating
joints). Its chain entanglement density is 0.672 mmol cm3.
The planar stiff terephthalic units make the PET a stiff polymer (Fig. 6.56). As in the case of PC,
the stiff units are joined by flexible groups. The rotational freedom around the main chain oxygen is
high (Hedenqvist et al. 1998). This results in amorphous PET being an intrinsically tough
(pseudoductile) polymer: C1 ¼ 4.2 and ve ¼ 0.815 mmol cm3 (Wu 1992).
The large steric side group in PMMA increases the characteristic ratio to C1 ¼ 8.2. This, together
with an ‘intermediate’ chain entanglement density (0.127 mmol cm3), shows that PMMA is
intrinsically brittle (Fig. 6.57).
The bulky aromatic side group of polystyrene (Fig. 6.58) increases the characteristic ratio even
further compared to PMMA, C1 ¼ 10.8. This, combined with its low chain entanglement density
(0.0561 mmol cm3), indicates that PS is even more brittle than PMMA. It should be noted that
confined space situations, such as in ultra-thin films, change the brittle/tough behaviour. Both the
6.15 Short-Term Mechanical Properties of Selected Polymers 325

Fig. 6.58 Molecular

structure of PS. Atom types
the same as for PC

H O H

N N
N

O H O

H O H

N N
N

O H O

Fig. 6.59 Two adjacent PA6,6 molecules. The hydrogen bonds are illustrated by broken lines

fracture (craze) and yield stress decrease with confinement (Vogt 2018). The chain entanglement
density decreases in ultra-thin films, and the radius of gyration and characteristic ratio become
anisotropic.
The flexible oxygen–methylene sequence of POM yields a flexible chain and a high crystallinity.
The glass transition is in the vicinity of that of polyethylene (85  C) (Van Krevelen and Te
Nijeuhuis 2009). The stronger intermolecular forces in POM, due to the negatively charged oxygen,
lead to a stiffer, stronger but more brittle polymer compared to HDPE.
The hydrogen bonds present in PA6,6 decrease the chain mobility (Fig. 6.59). In addition, the
peptide bond is stiff. Consequently, the glass transition is significantly higher for PA6,6 than for
polyethylene. However, it normally contains some water, which ‘lubricates’ the chains and leads to a
tough and ductile polymer. The onset of chain gliding requires a higher force in the case of PA6,6
compared to HDPE, since it involves the disruption of strong intermolecular hydrogen bonds. The
326 6 Mechanical Properties

crystals are also stronger and stiffer than for PE. The yield stress is thus higher for the former
(Fig. 6.59).
The high degree of crosslinks of thermosets results in low molecular mobility. Thermosets, here
illustrated by phenol formaldehyde, PF, are, therefore, normally strong and stiff, but may be brittle
(Table 6.2).

6.16 Rubber Elasticity

Conventional rubbers are crosslinked amorphous polymers well above their glass transition tempera-
ture. The elasticity of rubbers is predominantly entropy-driven, which leads to a number of spectacu-
lar phenomena (Gedde and Hedenqvist 2019). The stiffness increases with increasing temperature.
Heat is reversibly generated on deformation.
A more detailed analysis shows that the elastic force originates from both changes in conforma-
tional entropy and changes in the internal energy. The latter are normally small, and at constant
volume, they relate to changes in conformational energy. Polymers like polyethylene, with an
extended conformation as the low energy state, exhibit a negative energetic force contribution,
whereas rubbers like natural rubber show positive energetic forces.
Statistical mechanical models have proven useful in describing the stress–strain behaviour of
rubbers (Gedde and Hedenqvist 2019). The affine network model assumes that the network consists
of phantom Gaussian chains and that the positions of the junctions are prescribed by the macroscopic
deformation. The phantom network model assumes that the positions of the junctions fluctuate about
their mean positions prescribed by the macroscopic deformation ratio. The change in Helmholtz free
energy (ΔA) on deformation at constant temperature (T ) is due to a decrease in the number of possible
conformations of the chains of the network:
 
ΔA ¼ KNkT λ21 þ λ22 þ λ23  3 ð6:181Þ

where K is a dimensionless number different for the different models, N is the number of chains in the
network, k is the Boltzmann constant, T is the temperature (in K) and λi are the extension ratios along
the axes of an orthogonal coordinate system. Eq. (6.181) can be used to obtain the stress–strain
equation for different stress states. For the case of a rubber specimen subjected to a constant uniaxial
stress, the following true stress (σ)–strain (λ) expression holds:
 
1
σ ¼ KN 0e RT λ2  ð6:182Þ
λ

where N 0e is the number of moles of Gaussian chains in the network per unit volume of rubber and R is
the gas constant. The statistical mechanical theories thus predict that the stiffness of a rubber
increases with the crosslink density and temperature. The theoretical equation captures the trend in
experimental stress–strain data of unfilled rubbers in compression (λ < 1) and at low extension ratios
(λ < 1.2). At higher extension ratios, the experimental data fall below the theoretical curve. The
Mooney equation, which was derived from continuum mechanics, is, however, capable of describing
the nominal stress–strain data at λ between 1 and 2:

1 C2
σN ¼ 2 λ  C1 þ ð6:183Þ
λ2 λ

where C1 and C2 are empirical constants. It may be noted that the Mooney equation is identical to the
equations derived from statistical mechanics when C2 ¼ 0. Experiments have shown that for a range
6.17 Some Examples of Parameters Affecting the Mechanical Properties 327

of different rubbers, C1 depends on the crosslink density, basically as the modulus does according to
the statistical mechanical theory, whereas C2 is approximately constant. Stress–strain data of rubbers
swollen in organic solvents analysed by the Mooney equation show interesting results: C1 is
practically independent of the degree of swelling, whereas C2 decreases with increasing degree of
swelling approaching C2 ¼ 0 at v2  0.2. Hence, highly swollen rubbers (v2 0.2) behave according
to the statistical mechanical theories. Stress–strain data described by the Mooney equation suggests
that the networks at low extension (λ ¼ 1.0–1.2) deform nearly affinely, whereas at higher extensions
(λ ¼ 1.5–2), the fluctuations of the junctions increase (phantom network behaviour).
None of the Gaussian statistical mechanical theories is adequate to describe the stress–strain
behaviour at large strains (λ > 3 to 4). The pronounced upturn in the stress–strain curve can be
accounted for by the finite extensibility of the network not accounted for by the Gaussian distribution
that describes the statistics of the phantom chains. In addition, some rubbers crystallise during
extension, and a smaller part of the increase in stiffness at high extension ratios is due to the presence
of crystallites in the material.
Loose chain ends, temporary and permanent chain entanglements and intramolecular crosslinks
are complications not directly addressed by the classical statistical mechanics theory. Chain ends and
intramolecular crosslinks do not contribute to the elastic force, whereas chain entanglements add
more junction points to the covalent network.

6.17 Some Examples of Parameters Affecting the Mechanical Properties

It is important to understand the modulus–temperature relationship for a polymer and its variation
with molar mass, crystallinity, chain branching, etc. to be able to select the right material for a specific
product. In Fig. 6.60, a schematic modulus–temperature curve is shown for an uncrosslinked
amorphous polymer (Takahashi et al. 1964).
The relationship between modulus and temperature is closely related to the mobility of the
polymer chain segments. Most polymers show four to five different states: the glassy, the leathery
(main transition), the rubbery and the viscous state ( flow region) (Van Krevelen and Te Nijeuhuis
2009). The viscous state may further be divided into a low-temperature elasto-viscous (rubbery flow)
state, where the melt shows a combination of both elastic and viscous behaviour, and a high-
temperature more or less viscous (liquid flow) state. In the glassy state, at low temperature, atomic
vibrations occur through bond stretching, bond bending and torsional vibrations around specific
bonds. Additionally, local chain segment or side group rotations may occur. If the glassy polymer
is exposed to a small or moderate stress/strain, only local displacements of atoms and chain segments

Fig. 6.60 A schematic

modulus–temperature
curve for an uncrosslinked
polymer. The curve
encompasses the glassy (a),
the main transition (b), the
rubbery (c) and the
terminal, flow or viscous
region (d)
328 6 Mechanical Properties

occur, again through bond stretching, bending and torsional motions. These displacements are
generally reversible yielding an elastic material response. The amplitude and frequency of the atomic
vibrations increase with increasing temperature. Thus, the free volume and the interatomic distances
increase, which in turn lowers the attractive forces between adjacent molecule segments. A small
step-like decrease in modulus may be observed in the glassy region if a ‘strong’ sub-glass process is
activated (step in the a-region in the above figure) (R€ osler et al. 2007). As temperature increases
further, the amplitude and frequency of the vibrations increase, and when the glass transition is
approached, the glassy state turns into the leathery state, the latter characterised by chain segment
short-range diffusional motion (Van Krevelen and Te Nijeuhuis 2009).
As the modulus–temperature curve levels out, the rubber region (also referred to as the rubber
plateau region) is approached. The rubber region is characterised by a high molecular mobility, and
the molecular motions characteristic of the leathery region are faster (van Krevelen 2009). Slow long-
range motions are also activated here. The width of the rubber plateau region increases with
increasing molar mass (chain entanglements) (Aguilar-Vega 2013).
The elasto-viscous state is characterised by slippage of long-range entanglements (van Krevelen
2009). In the purely viscous state, whole chains are diffusing, which is accomplished through gliding
and rotation (reptation).
Examples of materials that are in the respective states at room temperature are polystyrene (a),
plasticised poly(vinyl chloride) (depending on the plasticiser content) (b), natural rubber (c) and a
polymer melt (d).
Chemical crosslinks act like physical entanglements with the difference that disentanglement and
gliding of entanglement points are not possible (Nielsen and Landel 1994). Hence, rubber materials
(chemically crosslinked elastomers) show little relaxation or creep due to segment gliding even at
high temperature or long loading time. The following example of styrene–butadiene rubber (SBR)
illustrates what happens when crosslink density increases (Fig. 6.61). The uncrosslinked polymer is
viscous at low frequency (large compliance), and with increasing frequency, it reaches the rubber
plateau region, the main transition region and almost the glassy region. When crosslinked, the viscous
region is absent and replaced with a broad rubber plateau (Aguilar-Vega 2013). The polymer cannot
flow because of the crosslinks. In addition, the compliance in the rubber plateau region decreases with
increasing crosslink density, i.e. decreasing molar mass between crosslinks (Mc).

-4

-5
log (J') (Pa )
-1

-6

-7

-8

-9
-15 -10 -5 0 5

log (f) (Hz)

Fig. 6.61 Storage shear compliance for SBR. The uppermost curve represents uncrosslinked material, and the crosslink
density increases from the middle to the lower curve. Drawn after data of Mohammady et al. (2002)
6.17 Some Examples of Parameters Affecting the Mechanical Properties 329

The glass transition is shifted towards lower frequency when the crosslink density increases. This
implies that at constant frequency, the glass transition temperature increases with increasing crosslink
density. As observed in Fig. 6.61, the width of the glass transition region increases. This is due to a
broad distribution in Mc because of a, usually, inhomogeneous distribution of crosslinks in sulphur-
vulcanised material. Chain segments in regions with a higher content of crosslinks are less mobile
than in regions with lower crosslink density. The former chain segments need a high activation energy
to move, and their motions are thus activated at a lower frequency (or higher temperature).
Thermosets, i.e. materials with a high crosslink density, show a less pronounced change in mechani-
cal properties in the glass transition region and a low ‘rubber’ compliance. Natural rubber, for
example, turns into ebonite at a high crosslink density (high sulphur content), and at the same time,
the glass transition increases from ca. –80 to 110  C (Eremeeva 1965). Both hockey pucks and
bowling balls are made of this material.
The effect of molar mass is mainly to change the chain entanglement density. Below Tg, larger
chain segment displacements are essentially absent, and only local molecular displacements occur.
These local activities are not affected significantly by entanglement density. Hence, the modulus of
the glassy material is essentially unaffected by molar mass variations (Nielsen and Landel 1994). An
increase in molar mass results in greater resistance to chain gliding, which results in an increase of the
temperature of the onset of viscous flow. Hence, the width of the rubber plateau region increases with
increasing molar mass (chain entanglement density) (DeGennes 1979).
The main effect of plasticisers is to lower the glass transition temperature of the polymer. The
reason for the unique combination of stiffness and flexibility of plasticised poly(vinyl chloride) is due
to the presence of a small amount of crystallites (Nielsen and Landel 1994). Without the crystallites,
plasticised PVC with a high content of plasticiser would be a viscous liquid unable to carry any load.
A broadening of the glass transition region is observed in the presence of a plasticiser with limited
solubility in the polymer due to incomplete mixing of the polymer and plasticiser on a molecular level
(Schmieder and Wolf 1952). Most protein and polysaccharide films need plasticisers to be ductile and
tough (because of a rigid hydrogen bond network and/or bulky side groups/branches). This is
illustrated in Fig. 6.62 by the stress–strain behaviour of wheat gluten films plasticised with 30 wt.
% glycerol. Newly prepared films are elastomeric. However, during ageing the films lose plasticiser
and become stiffer and stronger, but brittle.
With the addition of a small amount of plasticiser, the sub-glass (β) transition in several polymers
has been observed to reduce in strength or even disappear (Nielsen and Landel 1994). This means that
the molecular mobility is reduced between the β and the glass transition (α) temperatures, leading
‘unexpectedly’ to a stiff and brittle polymer in the presence of a plasticiser. This phenomenon is often
referred to as antiplasticisation, a misleading term according to Nielsen (Nielsen and Landel 1994).
At higher concentrations of plasticiser, expected plasticisation occurs, leading to a decrease in

Fig. 6.62 Sketch of the

stress–strain curves of a
glycerol-plasticised wheat
gluten film cast at pH 4.
Drawn after Olabarrieta
et al. (2006)
330 6 Mechanical Properties

stiffness and increase in ductility and creep/stress relaxation rates, since the polymer chains are forced
apart. Hydrogen bond-forming polymers, e.g. polyamide or poly(vinyl alcohol), can absorb water
vapour to such an extent that the polymer is plasticised and creep is facilitated. Plasticised PVC which
contains 30% di-ethyl-hexyl-phthalate has a glass transition temperature slightly above 0  C (Wolf
1951). In spite of that, the polymer creeps very slowly at room temperature. The reason for the very
slow creep is the existence of the aforementioned small number of crystallites.
The modulus in the rubber plateau region increases markedly with increasing degree of crystallin-
ity (Mercier and Groeninckx 1969). This is due partly to the higher modulus of the crystals compared
to the rubbery amorphous material. The increase is also partly due to the constraining effects of the
crystals on the amorphous phase. Since the amorphous chain segments are anchored to the crystals,
the presence of the crystals leads to a decrease of their mobility and hence an increase in the modulus
of the amorphous component. The impact of crystals in the glassy state is smaller than in the rubbery
state since the glassy chain segments are already constrained, and the glassy amorphous component
has a modulus closer to the crystal modulus, as compared to the rubbery component. The impact of
the crystals decreases at the onset of melting, where the modulus drops drastically. The decrease in
modulus above the melting point is also due to material softening initiated by thermal expansion.
Although the fall in modulus at Tg is less pronounced for high-crystallinity materials, the position of
the glass transition temperature is normally not affected significantly by variations in the degree of
crystallinity (Nielsen and Landel 1994).
The mechanical behaviour of random and block copolymers differs. Block copolymers of the
‘stiff’ styrene and the ‘flexible’ butadiene monomers yield two phases, and the resulting tanδ-T curve
shows two maxima representing the glass transition of butadiene (~  70  C) and of styrene
(~100  C) (Fig. 6.63). In contrast, the random copolymer shows only a single maximum between
those of the block copolymer (a single glass transition).
Polymer blends are commonly not molecularly miscible. Therefore, most binary polymer blends
yield two-phase structures similar to block or graft copolymers. The two-phase system yields two
glass transitions that originate from each specific phase (Nielsen and Landel 1994). The modulus thus
falls abruptly at the two glass transition temperatures, and these temperatures do not change with the
composition of the blend. An example of an immiscible system is the blend of acrylonitrile styrene
and butadiene. Two glass transitions are observed which originate from the two components. High-
impact polystyrene (HIPS) containing approximately 10% cellular butadiene rubber particles

Fig. 6.63 DMA data of random- (bold line) and block- (thin line) type solution styrene–butadiene copolymers. Drawn
after data of Wetton et al. (1991)
6.17 Some Examples of Parameters Affecting the Mechanical Properties 331

Fig. 6.64 Storage shear

modulus (●) and loss
tangent (○) as a function of
temperature for high-
impact polystyrene. Drawn
after Molan and Keskkula
(1968)

Fig. 6.65 Storage (left) and loss (right) moduli as a function of temperature for PVC blended with NBR containing
28% acrylonitrile. The ratio in PVC/NBR is (●) 100/0, (○) 83/17, (■) 67/33, (□) 50/50, (◆) 20/80 and (Δ) 0/100.
Drawn from Takayanaga (1983) and Takayanaga and Goto (1985)

possesses, as well as the polystyrene Tg peak, a small Tg peak at 80  C due to the rubber component
(Fig. 6.64). The size of the butadiene damping peak is primarily a function of the amount of butadiene
present, but it is also a function of the relative sizes of the relaxation strengths of the two components.
PVC and nitrile rubber (NBR, containing more than 25% acrylonitrile) is one of the few miscible
polymer blends, and its mechanical properties are similar to plasticised homopolymers. Note that only
a single glass transition is observed (Fig. 6.65).
Polyoxymethylene (POM) and polyvinyl phenol (PVP) are also miscible (Fig. 6.66). The γ-
transition in POM is local and not dependent on the blend composition. The α-transition is no longer
visible in the blends. Tg (i.e. ß in Fig. 6.66) increases from ~20 to 70  C (90/10) and 80  C (80/20) due
to poly(vinyl phenol) having a Tg of 153  C.
Several partially compatible blends exist, i.e. where the two components are partially miscible.
PVC and ethylene-vinyl acetate (EVAc, containing 63% of vinyl acetate) is an example of a partially
miscible blend. The system consists of pure PVC particles embedded in a matrix consisting of EVAc
and PVC. Therefore, the damping peak (loss tangent) of PVC remains at the glass transition of PVC
(high-temperature peak), while the EVAc peak (low-temperature peak) is shifted towards the PVC
peak with increasing PVC content in the blend (Fig. 6.67).
332 6 Mechanical Properties

Fig. 6.66 Miscible blends of POM and PVP. The curves correspond to POM/PVP 100/0 (bold curve), 90/10 (broken
curve) and 80/20 (thin curve). Drawn after Keating et al. (1997)

Fig. 6.67 Loss tangent as a

function of temperature for
PVC blended with EVAc
(63% vinyl acetate) at a
mixing temperature of
160  C. The PVC/EVAc
ratio is (●) 10/90, (○)
25/75, (■) 50/50, (□)
70/30, (◆) 85/15, (Δ) 92.5/
7.5 and (♦) 100/0. Drawn
after Shur and Rånby
(1975)

6.18 Yielding of Polymers

6.18.1 Introduction

With increasing strain, the stress–strain curve becomes non-linear. This generally occurs between
0.1% and 1% strain, and at strains as high as 1–10%, the nominal stress (force divided by initial
cross section) reaches a plateau (McCrum et al. 1997). The point where dσ/dε is zero is defined as
the yield point. It is often difficult to distinguish between elastic recoverable strain and plastic
non-recoverable strain for polymers since the degree of recovery is both temperature and time
dependent, i.e. the degree of recovery depends on for how long a time the material is allowed to
recover before stress or strain sets in again. Polymers are usually also viscoplastic, meaning that the
plastic deformation is time dependent under load. The degree of recovery depends strongly on the
molar mass of the polymer and is also affected by the presence of chemical crosslinks. Polymers with
very high molar mass or degree of crosslinking show almost complete recovery due to the high
content of physical crosslinks (chain entanglements) and chemical crosslinks, which ensure that the
6.18 Yielding of Polymers 333

Fig. 6.68 Neck (middle)

and craze (right) formation
in a dumbbell specimen
under load. τ and σ are
shear and normal stresses

t
s

Fig. 6.69 A true

stress–strain curve with
(solid line) and without
(dashed line) strain
stress

softening. In both cases

strain hardening occurs

strain

original chain conformations are basically restored on unloading. The driving force for recovery is
entropic.
A material that yields (exhibiting a yield point) is ductile, whereas a material that does not yield is
usually brittle. Whether a polymer is ductile or brittle depends on its resistance to yielding and to
crazing or to subsequent crack propagation. These properties are in turn dependent on temperature,
type of loading, component geometry, strain rate and environment (e.g. the presence of plasticising
gases or liquids) and, of course, on the internal polymer properties. Crazing refers to the formation of
a large number of voids surrounded by elongated fibrils, and these are formed in front of a growing
crack.
Yielding occurs by slip at angles of 45  with respect to the draw direction, whereas crazing occurs
by elongation of microscopic fibrils parallel to the macroscopic draw direction, surrounded by
developing voids (Fig. 6.68). The latter occurs only under a hydrostatic tensile stress condition
(refer to, e.g. the failure envelope in plane stress conditions (R€
osler et al. 2007)), leading to a volume
change, whereas yielding occurs under constant volume. Yielding normally involves a large volume
of material, whereas crazing can be more local (Deblieck et al. 2011). The drop in nominal stress
beyond the yield point is due to local necking of the specimen characterised by a drastic decrease in
the specimen cross section. However, the true stress–strain curve may also show a stress overshoot
under yield (Fig. 6.69). The strain softening, observed mainly in glassy polymers beyond the
overshoot, has been explained as being due to at least two effects. (1) In glassy polymers a
mechanically induced molecular disordering is considered to occur at the onset of plastic
334 6 Mechanical Properties

deformation, which reduces the modulus (Chen and Schweizer 2011). The disordered structure
implies a lower activation energy for further molecular displacement (plastic deformation) (Siviour
and Jordan 2016). (2) The work that is dissipated during deformation causes, especially at higher
strain rates under adiabatic conditions, a local temperature increase and thus a decrease in resistance
to plastic deformation (Mullikan and Boyce 2006; Stachurski 1997).
As the deformation proceeds beyond the yield point, the neck propagates and the polymer chains
orient in the deformation direction. This is referred to as cold drawing (Birley et al. 1992). The
increase in stress at larger strains is referred to as strain hardening and is characterised by a strain
hardening modulus. It should be mentioned that necking and stable neck propagation is favoured by a
low yield strength and a high strain hardening modulus. The strain hardening modulus for an
amorphous polymer increases with the number of load bearing (effective) chain entanglements
which, in turn, increases with molar mass. In the presence of crystals, a more complex structure
yields more parameters that affect the stress–strain behaviour. Apart from the effects of the crystals
themselves, the tie chains connecting the crystals also play an important role. This is illustrated by
simulations on linear and branched bimodal (molar mass) polyethylene crystallised systems, which
show that the strain hardening region and consequently fracture strength increase in the presence of
short-chain branches on the longer chains (Moyassari et al. 2018, 2019). These yield a material that
has a higher tie chain and chain entanglement concentration before deformation and that allows for a
greater disentanglement during deformation than in the linear systems of similar molar mass
distribution. The crystals are also more strain-resistant in the branched systems.

6.18.2 Theories of Yielding in Amorphous Polymers

Several molecular theories for yielding exist (refer to, e.g. Stachurski 1997). These include transition
state theories, free volume theories, conformational state theories, dislocation theories and segmental
motion theories. Constitutional models that consider large viscoplastic deformation of glassy amor-
phous polymers include the Boyce–Parks–Argon model, the Oxford-glass–rubber model and the
Eindhoven glassy polymer model (Wang et al. 2019).
The earliest transition state theory was derived by Eyring (McCrum et al. 1997). The Eyring
model for plastic flow of solids provides general correlations between temperature (T ), strain rate (_ε)
and yield strength (σ Y) of polymers. The theory predicts a yield stress that increases linearly with the
logarithm of the strain rate and that decreases with increasing temperature (Eq. (6.184)). Figures 6.70
and 6.71 show that this is also the case experimentally for polycarbonate.
 
2 ε_
σ Y ¼ ∗ ΔH þ 2:303 R T log ð6:184Þ
V ε_ 0

where ΔH is the activation energy of flow, V* is the activation volume and ε_ o is a constant (note that
the logarithm is negative, ε_ o > ε_ ).
The yield strength of amorphous glassy polymers is shown to increase with the cohesive energy
density (stronger intermolecular interactions), decreasing temperature below the glass transition
temperature (Tg-T ) and intra-chain contributions (Deblieck et al. 2011). The last term is essentially
increasing with increasing characteristic ratio (C1). Yield strength is normally also increasing with
hydrostatic pressure, which is observed by the yield strength in compression often being higher
compared to in tension (Siviour and Jordan 2016).
A constitutive model for large deformations involving time/rate-dependent effects has been
developed by Boyce and co-workers (refer to Mullikan and Boyce 2006). It consists of elements
describing the intermolecular resistance (elastic-viscoplastic effects) in parallel with a Langevin
6.18 Yielding of Polymers 335

Fig. 6.70 Yield stress data versus temperature (●) of polycarbonate from Bauwens-Crowlet et al. (1969) fitted (line)
using Eq. 6.184

spring describing the non-linear entropically driven hardening. The elastic-viscoplastic part contains
one or several elements in parallel where the element is a linear spring in series with a viscoplastic
dashpot. Wang et al. (2019) developed a constitutive model based on a finite strain and thermome-
chanical coupling to describe large strain behaviour of glassy amorphous polymers. It is based on the
assumption that elastic, viscoplastic and thermal components can be considered as separate
contributions to the deformation behaviour. Features include hyperelasticity, strain softening and
hardening and positive dissipation due to non-reversible viscoplastic flow.
A different approach to successfully predict even non-linear mechanical behaviour is to use Group
Interaction Modelling (GIM), based on general properties of the ‘mers’. The basis of the model is to
consider that the actual mechanical properties are a consequence of the stored and dissipated energy
as a function of deformation. GIM is able to predict relatively well, for example, the non-linear
compressive yield stress behaviour as a function of strain rate and temperature for PMMA and PC,
and the strain softening in PMMA (Porter and Gould 2009), as well as the yield strength and stiffness
dependence on strain rate and temperature in epoxy thermosets (Foreman et al. 2010).

6.18.3 Yielding in Semicrystalline Polymers

It has been shown for single polymer crystals that plastic deformation occurs through screw
dislocations if the force is applied perpendicular to the fold surface, i.e. nearly parallel to the chain
direction (Fig. 6.72). If the force is applied along the polymer chain, plastic deformation occurs
through unfolding of molecules out of the crystals and recrystallisation into a microfibrillar
morphology.
At strains lower than those where necking sets in, for a spherulitic polycrystalline polymer, several
mechanisms of deformation can occur. In the regions where the force is perpendicular to the lamellar
crystal fold surface, microvoids (microcrazes) are formed between crystals, and the crystals are,
therefore, held together by microfibrils (Fig. 6.73). These microfibrils induce a compressive force on
the material that results in reversible elastic stress–strain behaviour. Furthermore, crystals are sheared
plastically by intralamellar molecular slips that occur where the force is exactly along the polymer
chain.
336 6 Mechanical Properties

Fig. 6.71 Yield stress versus strain rate of polycarbonate obtained from data of Bauwens-Crowlet, Bauwens and
Homés (1969)

Fig. 6.72 The yield

mechanism in
semicrystalline polymers.
Drawn after Darras and
Séguéla (1993). Arrows
correspond to shear stresses

Fig. 6.73 A lamellar stack containing three crystals (grey areas) surrounded by amorphous material (left). When a
stress is applied in the direction of the large arrow (right), the crystals are separated and bridged by microfibrils
surrounded by microcrazes (black parts). The small arrow highlights an intralamellar slip. Drawn after Petermann and
Ebener (1999)

In regions where the crystal fold surface is oriented parallel to the force direction, interlamellar
slip occurs. In cases where the crystals form a helix (twisting sheets in Fig. 6.74), microcrazing and
interlamellar slip can occur simultaneously between neighbouring crystals. For an isotropic semi-
crystalline polymer, there are only a few crystals that have their (001) planes in the direction of
6.18 Yielding of Polymers 337

Fig. 6.74 Polyethylene spherulite with twisting lamellae prior to yield deformation. Drawn after Darras and Séguéla
(1993)

Fig. 6.75 Deformation

during necking. The local
neck region contains
disordered material. The
grey areas are crystals.
Drawn after Petermann and
Ebener (1999)
neck region

maximum shear stress (Fig. 6.74). Hence, before and during yielding, it is suggested by Darras and
Séguéla (1993) that crystals which are not bound too tight to neighbouring crystals reorganise in the
direction of maximum shear stress, thereby allowing the slip through the crystal to occur at the lowest
possible energy.
In cases where two yield points are observed, it is suggested that the first is due to a fine chain slip
and a martensitic type of deformation in the crystal, and in the second yield point, a more extensive
chain slip occurs which leads to crystal fragmentation (Schrauwen et al. 2004).
Syndiotactic polypropylene has been studied in order to reveal the mechanisms of deformation
during necking (Petermann and Ebener 1999). Polypropylene, which has undergone necking, always
contains the orthorhombic crystal unit cell with chains in the trans conformation, irrespective of the
type of crystal unit cell present in the undeformed material. This indicates that the necking procedure
involves molecular unfolding and recrystallisation steps via a disordered state (Fig. 6.75). The
material which has been deformed may contain oriented micellar crystals, shish-kebab crystals and
disordered material.
Crystallisation from oriented melts or crystallisation induced by large strains in the solid state may
yield a fringed micellar type of polypropylene morphology where the micelles act as crosslinks in the
amorphous matrix. This type of morphology usually yields elastomeric behaviour. If the micelles
338 6 Mechanical Properties

Fig. 6.76 Needle crystal

morphology. Needle
crystals are grey. Drawn
after Petermann and Ebener
(1999)

Fig. 6.77 The logarithm of yield stress as a function of reciprocal stem length for polyethylene. Drawn after Brooks
et al. (1999)

grow preferentially in the chain direction, they turn into needle crystals (Fig. 6.76). This morphology
is observed in polybutene-1, and it yields mechanical properties similar to that of short glass fibre-
reinforced thermoplastics (Petermann and Ebener 1999). The aspect ratio of the needle crystals
ranges from 1 to more than 100.
Young (1974, 1988) developed a theory based on nucleation-initiated yield for semicrystalline
polymers. The theory assumes that yield occurs through thermal activation of screw dislocations
within the crystal lattice. Since the Burgers vector is considered to be along the chain axis and slip
should then occur through any (hk0)-plane, the yield stress is expected to increase with increasing
crystal thickness. Since the chain axis is normally tilted and different from the crystal surface normal
in polyethylene, it is better to consider the chain stem length rather than the crystal thickness. The
theory predicts the following relationship for the tensile yield stress:
 
K ðT, ε_ Þ 2π ΔGa ðT Þ
σY ¼ exp  þ1 ð6:185Þ
2π l K ðT, ε_ Þ b2

where l and ΔGa(T ) are the stem length and Gibbs free energy for nucleation of dislocation,
respectively. K ðT, ε_ Þ is the geometrical mean of the shear moduli C44 and C55. The theory predicts
a linear relationship between the logarithm of yield stress and the reciprocal stem length. By
examining a great number of polyethylene resins, Brooks et al. (1999) were able to verify this linear
relationship experimentally (Fig. 6.77).
6.19 Fracture of Polymers 339

6.19 Fracture of Polymers

6.19.1 Introduction and the Brittle–Ductile Transition

Failure of polymer products is of great concern and affects everyday life and society in several ways.
Apart from the cost and hassle associated with replacing a product part with a spare part, premature
failure can, in the worst case, result in accidents. In order to be able to predict the lifetime of polymer
products, and to make these last in service as long as it is necessary, it is of importance to understand
how and why polymer products fail/fracture. Fracture normally develops from a crack generated
during service or pre-existing due to, for example, impurities in the material (Devries and Nuismer
1985; Xu et al. 2017a, b). Since cracks can lead to failure, efforts have been put into creating self-
healing polymeric materials that can repair damages/cracks. Both extrinsic and intrinsic strategies
exist. The former includes preloading the material with microcapsules or short hollow glass fibres
containing a healing agent that, when a crack penetrates the capsule or fibre, ends up in the crack and
polymerises there (Awaja et al. 2016; Naebe et al. 2016). In the intrinsic approach, the self-healing is
accomplished, for example, by having a thermoplastic or thermoset polymer blend where one
component is mendable (through, e.g. the presence of reversible Diels–Alder crosslinks).
Glassy polymers, for example, polystyrene and poly(methyl methacrylate), appear to fracture in a
brittle manner without any signs of ductility at room temperature. Close examination, however,
reveals patterns of plastic deformation in a localised region close to the crack tip, as indicated by a
fibrillar structure. This fibrillar structure originates from the craze structure that is formed, as
mentioned previously, under normal stress along the fibrils prior to fracture. Electron microscopy
reveals that the fibrils have a diameter of the order of tens of nm, and macroscopically the craze
region looks like an extension of the crack (Fig. 6.78). It is also referred to as the cohesive region
(Knauss 2015). The centre-to-centre distance between fibrils is 10–50 nm (Michler 2001).
A network of interconnected voids surrounded by fibrils forms the craze, which is observed
visually since the structure scatters light. The void formation relieves the local stress (Knauss
1989). Molecular entanglements ensure the formation of fibrils which, in turn, are capable of bearing
limited load. Therefore, fibrils are generally absent in very low molar mass polymers, which are also
very brittle. Polystyrene and some other glassy polymers and lightly crosslinked thermosets are brittle
in tension but ductile in compression.

Fig. 6.78 Craze fibrils at the tip of a through-thickness crack. Arrow indicates the direction of crack propagation, and
the grey area refers to the craze/cohesive region
340 6 Mechanical Properties

Fig. 6.79 Brittle/ductile

transition temperature
(TBD)

Fig. 6.80 Fracture energy

as a function of molar mass
for PMMA. Drawn after
Kausch and Dettenmaier
(1982)

Whether a polymer is ductile/tough or brittle depends on its resistance to yielding and to crazing
and subsequent crack propagation. This tendency depends on temperature, as indicated by the brittle/
ductile transition (TBD, Fig. 6.79) (Ward 1985). Below TBD, the failure mode is brittle and the
strength is determined by the size of the brittle fracture stress (σ B). Above TBD, the failure is ductile
and the strength is determined by the size of the yield stress (σ Y). TBD increases with velocity of
loading and is also dependent on the type of stress concentration, microstructure of the material,
environment and component design. The transition from a ductile to a brittle mode is associated with
a large drop in the energy-absorbing capacity, and, unfortunately, this transition occurs close to the
service temperature for many plastics. The main molecular processes involved in fracture are chain
conformational changes, chain separation (cavitation), chain gliding and chain scission. Fracture is
favoured by chain scission and, to a certain degree, by chain separation and chain gliding, whereas
conformational changes lead to stress relaxation and lower the tendency to fracture. Whether
yielding, crazing or a combination of these dominates depends on the specific polymer and can be
related to the strength of the β sub-glass transition. It is strong (large tan (δ) value) in polycarbonate,
which dissipates energy through the activation of shear yielding in relatively large local regions. In
contrast, polystyrene has a weak β-transition and local crazing dominates. Polymethyl methacrylate
has a β-transition strength between the two polymers and experiences both shear-yielding and crazing
(Grellman and Lach 2001).
Chain disentanglement and chain gliding are prohibited by a high chain entanglement density,
which increases with molar mass. Thus, the fracture resistance increases with molar mass, as
illustrated for PMMA in Fig. 6.80. In general, high forces favour chain scission and lower forces
favour chain disentanglement (Persson 1999).
6.19 Fracture of Polymers 341

When entangled polymer melts, gels or solutions are exposed to elongational flow, brittle fracture
with scission of polymer chains (carbon–carbon bonds) can occur (Wagner and Narimissa 2018). In
contrast to, for example, a highly crosslinked elastomer, where enthalpic fracture is the main
mechanism, entropic fracture is the main mechanism in entangled systems. In the former case,
fracture occurs when an entangled segment is fully stretched and the force becomes higher than the
rupture force of the carbon–carbon bond (critical strain energy is reached, which corresponds to the
total strain energy when all the carbon–carbon bonds in the segment are stretched to the same critical
extent). In entropic fracture, it is considered that failure occurs before all bonds are critically stretched
and is due to thermal fluctuations, which means that under a very short time period, the strain energy
is localised to one single carbon–carbon bond which then ruptures. Hence, in a purely entangled melt/
solution, brittle fracture under extensional strain occurs before full orientation of the polymer chains
is achieved (entropic fracture) and consequently requires less strain energy than in pure enthalpic
fracture. In elastomers, a progressive damage model has been developed where the free energy
associated with large-scale deformation has both an entropic and an enthalpic contribution (Talamini
et al. 2017). The latter becomes important as the chain segments/Kuhn segments are stretched and is
the driving force for damage and fracture of the elastomer.
The following two subchapters deal with brittle and ductile fracture followed by a description of
the different modes of failure.

6.19.2 Fracture Mechanics

In fracture mechanics, the prediction of the instability of a pre-existing crack is central. Based on
load–load line and load–deflection curves, it is possible to divide the polymer crack behaviour into
different modes (Fig. 6.81). Polymers following linear elastic fracture mechanics (LEFM) are
typically brittle polymers below Tg, particulate- and fibre-filled polymers and/or thick specimens
(plane strain) (Grellman 2001). Otherwise, polymers follow elastic-plastic fracture mechanics
(EPFM). In the case of LEFM, the stress intensity at the crack tip is considered to be the critical
factor determining crack growth. In EPFM, the sizeable plastic deformation in front of the crack tip
leads to a widening of the crack, and crack growth is strain controlled rather than stress controlled. In
this case, it is more meaningful to use crack opening displacement (COD) or crack tip opening

Fig. 6.81 Load displacement for a material following LEFM without (1) and with (2) small-scale yielding and EPFM
with unstable (3) and stable (4) crack growth. Drawn after Grellman (2001). The displacement is either measured
through load line changes, SENB deflection or crack/crack tip opening changes (Fig. 6.82)
342 6 Mechanical Properties

Fig. 6.82 Compact tension specimen (mode I, notch length a, load line displacement (vLL)) (left) and a single edge
notched beam (SENB) (right)

Fig. 6.83 An elliptical

through-thickness crack in
an elastic sheet subject to a
stress (σ) acting in the
y-direction, which causes a
stress distribution along x,
as shown. Drawn after
McCrum et al. (1997)

displacement (CTOD) as a critical factor for crack growth. In both LEFM and EPFM, the crack
growth can be treated using an energy approach. LEFM and EPFM will be described further below.
The term linear elastic fracture mechanics (LEFM) refers to materials that follow Hooke’s law,
i.e. where the relationship between stress and strain is linear (Williams 1978; McCrum et al. 1997;
Grellman 2001). For this to hold, the extent of yielding at the crack tip must be small or absent
(Fig. 6.83).
Consider a wide sheet that contains a crack and which experiences a force (F) (Fig. 6.83). Wide in
this case means that W 2a, where 2a is the length of the crack and W is the width of the sheet. The
stress is then defined as:
F
σ¼ ð6:186Þ
WB

where B is the thickness of the sheet. If a crack is present, the stress will be amplified, and the
maximum stress amplification at the crack tip is (Francois et al. 2013):
6.19 Fracture of Polymers 343

σm 2a
¼1þ ð6:187Þ
σ b

where σ m is the strength at the tip. Hence, for a sphere-shaped crack, the amplification is only 3, but
for an ellipsoidal crack, the amplification may be several orders of magnitude higher. With a few
exceptions, including diamond and some ceramics, materials are able to relax the amplified stress by
plastic flow or other crack blunting mechanisms before it reaches these high values. The crack will
only propagate if it is accompanied by decrease of internal energy. In order to examine the
mechanisms involved in crack inhibition and propagation, a thin sheet is considered. The total energy
of the crack-containing system contains two parts: the elastic strain energy and the work of crack
formation. The elastic strain energy is:

Vσ 2 σ 2 π a2 B
Ue ¼  ð6:188Þ
2E E

where the first term is the elastic strain energy of the sheet of volume V before the crack is introduced
and the second term is the lowering of elastic strain energy associated with the presence of the crack.
E is the elastic modulus of the material. The work of crack formation is associated with the formation
of two opposite crack surfaces. The work performed per unit surface area is Gc. Part of this work is
directly due to the formation of new surface, and part of the work is dissipated as heat. For materials
where no heat is dissipated, Gc is equal to twice the surface energy (γ). For polymers, however, Gc is
much larger than 2γ. The total change in the internal energy due to the presence of a crack is:

σ 2 π a2 B
ΔU ¼  þ 2 a B Gc ð6:189Þ
E

where the first term is the lowering of strain energy (Eq. (6.188)) and the second term is the increase
in energy due to crack surface formation. For small cracks (small a), the crack is stable due to t ΔU
being dominated by the surface formation energy, which increases linearly with a. For large cracks,
the first (negative) term of Eq. (6.189), i.e. the strain energy decrease, dominates ΔU and the crack
becomes unstable. At the point of crack propagation, am, the surface formation energy equals the
strain energy decrease:

∂ σ 2 π a2 B ∂
¼ ð2 a B G c Þ ð6:190Þ
∂a E ∂a

Hence

σ 2 π a ¼ E Gc ð6:191Þ

and the maximum crack size that is stable is:

E Gc
am ¼ ð6:192Þ
π σ2

For a larger than am, ΔU/Δa decreases, which means that the total internal energy decreases with
increasing a; hence, the crack will propagate catastrophically. The stress that is capable of producing
crack propagation is the fracture stress, and it is readily derived from Eq. (6.191):
344 6 Mechanical Properties

0:5
E Gc
σF ¼ ð6:193Þ
πa

Note that this holds for plane stress, i.e. for a thin sheet. A more useful parameter than the fracture
energy Gc when assessing a material’s fracture resistance is the plane stress critical stress intensity
factor Kc (McCrum et al.1997). For a wide sheet, it is:

K c ¼ σ F ðπ aÞ0:5 ð6:194Þ

which, together with Eq. (6.193), yields an expression that relates Kc and Gc:

K c ¼ ðE Gc Þ0:5 ð6:195Þ

Provided the largest crack can be detected in the thin sheet, the stress intensity K at an applied
stress can be calculated:

K ¼ σ ðπ aÞ0:5 ð6:196Þ

If the calculated stress intensity is lower than Kc, the crack will not propagate. Eqs. (6.195) and
(6.196) can be obtained from the stress distribution around the crack tip (Williams 1978). The stresses
can be written in cylindrical polar coordinates (Fig. 6.84):
2
1 ∂ϕ 1 ∂ ϕ
σr ¼ þ ð6:197Þ
r ∂r r 2 ∂θ2

2
∂ ϕ
σθ ¼ ð6:198Þ
∂r 2

∂ 1 ∂ϕ
σ rθ ¼  ð6:199Þ
∂r r ∂θ

where ϕ is the stress function which satisfies the biharmonic equation:

Fig. 6.84 Stresses around a crack tip. Drawn after Williams (1978)
6.19 Fracture of Polymers 345

∇4 ϕ ¼ 0 ð6:200Þ

The stress function is defined as:

ϕ ¼ r n ΘðθÞ ð6:201Þ

where Θ(θ) is determined from Eq. (6.197). If only the region very close to the crack tip is considered,
and considering that the crack surfaces are stress-free (i.e. σ θ ¼ σ rθ ¼ 0 at θ ¼ π), then the hoop
stress becomes:
   
3c 3θ θ 3b 3θ θ
σ θ ¼ 0:5 cos þ 3 cos þ 0:5 sin þ sin ð6:202Þ
4r 2 2 4r 2 2

which was fully derived by Williams (1978). b and c are constants. The first term is associated with a
stress normal to the crack propagation direction, and the second term is associated with a ‘sliding’
stress condition. Since the first case is the most frequently occurring stress condition, the following
treatment will only consider the first term. It can be rewritten as:
K 1 θ
σθ ¼ 0:5 2
cos ½1 þ cos ðθÞ ð6:203Þ
ð2π r Þ 2

For an infinite plate subjected to a uniform stress, σ, and containing a crack of size 2a, the stress
function solution yields the result for σ yy at y ¼ 0 (Fig. 6.83):
σx
σ yy ¼ ð6:204Þ
ðx 2  a2 Þ0:5

Considering that x ¼ a + r, then Eq. (6.204) can be expanded as:

 0:5
a
σ yy ¼ σ þ ... ð6:205Þ
2r

The above expression can be combined with Eq. (6.203) at θ ¼ 0 to yield Eq. (6.196):

K ¼ σ ðπ aÞ0:5 ð6:206Þ

If a thick sheet is considered, the theory has to be modified since plane stress is no longer valid. For
a thin sheet, the thickness of the specimen decreases at the crack tip, but for a thick sheet, the
surrounding material is opposing the contraction (elastic constraint, plane strain), which then leads to
finite stresses in the thickness direction. Hence, a three-dimensional stress state occurs in the vicinity
of the crack tip. Plane strain conditions reduce yielding at the crack tip due to the three principal
stresses being tensile. The fracture stress for plane strain is:
0:5
E GIC
σF ¼ ð6:207Þ
π ð1  υ 2 Þ a

where GIC is the fracture energy for plain strain in a thick sheet and υ is Poisson’s ratio. Index I refers
to mode I where the crack is exposed to tensile stresses (mode II corresponds to in-plane shear sliding
in the direction of the notch, and mode III corresponds to in-plane shear perpendicular to the notch). A
critical stress intensity factor for the case of plane strain is:
346 6 Mechanical Properties

K IC ¼ σ F ðπ aÞ0:5 ð6:208Þ

As for the case of plane stress, the fracture energy and critical stress intensity factor are related by,
in this case, combining Eqs. (6.207) and (6.208):
0:5
E GIC
K IC ¼ ð6:209Þ
ð1  υ2 Þ

Since materials show minimum ductility at plane strain, it is generally enough to consider this
case. If the material does not fracture at a given stress at plane strain, it would also survive the stress at
plane stress conditions. If yielding is prohibited at the crack tip, the possibility to reduce elastic strain
with plastic flow is less, and hence the probability of crack propagation increases. In accordance with
plane stress, the stress intensity factor in plane strain can be determined from the stress and the largest
crack size:

K I ¼ σ ðπ aÞ0:5 ð6:210Þ

The treatments above consider only cases that involve linear fracture mechanics, i.e. when the
extent of plastic deformation close to the crack tip is small. In order for this to hold for ductile
polymers, it is necessary that plane strain prevails. It is observed experimentally that this holds when
the crack size (a), film thickness (B) and film width (W ) obey:

K IC
a, B, ðW  aÞ > 2:5 ð6:211Þ
σY

Since it is difficult to test large specimens, there are methods that allow for a certain degree of
non-linear elastic fracture behaviour when determining KIC (Williams 1978; McCrum et al. 1997).
These methods are also applicable to cases when the material shows non-linear elasticity,
viscoelasticity or stable crack growth. In the case of small-scale yielding, blunting occurs at the
crack tip, and the yield zone becomes more or less colinear in shape (Fig. 6.85, line zone). It should be
mentioned that, theoretically, the yield zones in plane stress and strain have different forms and size,
depending on what yield condition is used (Thamm and Ágoston 2004). It depends on whether only
deviatoric stresses are considered (von Mises yield criterion) or also hydrostatic conditions are
considered. It also depends on the actual Poisson’s ratio (0 for pure plane stress and 0.5 for pure
plane strain).
When extensive plastic deformation occurs at the crack tip, it is not possible to use LEFM
(Williams 1978). In this region the material is better described by elastic-plastic fracture mechanics
(EPFM). At the crack tip, a region of porous material forms fibrils (ligaments) which, as mentioned
earlier, is referred to as the craze zone (Fig. 6.86).

Fig. 6.85 Schematic

representation of regions of
plastic deformation at the
crack tip. Drawn after
Williams (1978) and
Thamm and Ágoston
(2004)
6.19 Fracture of Polymers 347

Fig. 6.86 Craze zone.

Drawn after Williams
(1978)

Fig. 6.87 Definition of the

J-integral

In EPFM, it is more fruitful to consider a crack opening displacement (δ) rather than a crack length
a when relating the size of the crack to the fracture resistance of the polymer. The crack tip opening
displacement method (CTOD) uses a critical displacement (δc), above which crack propagation
occurs, as a measure of the ductility of the material. If there is a linear relationship between stress
and strain or plane strain prevails, the critical displacement is directly correlated with the fracture
energy (McCrum et al. 1997):
GIC ¼ δc σ c ð6:212Þ

This relationship is obtained by considering that the work performed on an element of length dX as
the crack propagates, i.e. from the zone tip to the crack tip, due to a stress (σ c, craze stress) operating
over the same distance, is (Fig. 6.86):
0δ 1
ð
dU ¼ @ σ c dδA dX ð6:213Þ
0

Since dX ¼ da during propagation:

ðδ
dU
GI ¼ ¼ σ c dδ ¼ δ σ c ð6:214Þ
da
0

An alternative approach is the J-integral concept (Ramstainer et al. 2001). In standard form, it is
written as (Fig. 6.87):
ð
du
J¼ wdy  T ds ð6:215Þ
dx
Γ
348 6 Mechanical Properties

where T is the traction vector on the contour path with the increment ds and u is the corresponding
displacement vector. w is the work made by the load per unit volume, and Γ is the contour path of the
loop integral starting from the lower side of the crack (n is the normal vector). The integral defines the
energy entering the closed loop area minus the energy stored during crack growth.
If the material behaves elastically, the energy release rate:

∂U 1 ∂U
J¼ ¼ ð6:216Þ
∂A B ∂a

can be obtained by measuring the load–deflection curves for samples of different crack lengths
(McCrum et al. 1997). Crack growth is considered to occur when the energy release rate is equal to
or greater than a crack propagation resistance parameter JR. This criterion is similar to GI  GIC in
linear elastic fracture mechanics. In cases when the non-linearity is due to major yielding or
viscoelasticity rather than non-linear elasticity, the energy release rate due to crack growth is difficult
to measure. This is because energy is also dissipated remote from the crack, and this energy may
have no connection with crack growth. A complicating fact is also that since the plastic region does
not return to its original shape, some elastic energy still remains in the specimen. Therefore, a full
unloading curve can no longer be used to determine the energy release rate during crack growth.
A way of overcoming these problems is to measure load versus deflection for a series of specimens
with slightly different initial crack lengths to obtain a small region of stable crack growth (da). The
energy release rate, J, is then calculated at constant displacement. da can be obtained by freeze
fracturing the sample and visual examination of the crack extension. A crack initiation energy release
rate is defined as J0.2, i.e. the crack resistance at Δa ¼ 0.2 mm. This parameter is obtained by plotting
JR, the crack growth resistance, determined from SENB or CT tests as a function of Δa. J is
calculated from:
ηU
J¼ ð6:217Þ
BN ð W  a 0 Þ

where U is the energy absorbed by the specimen and BN is the net thickness of the specimen. η is a
constant which takes different values depending on specimen geometry and measuring method
(McCrum, Buckley and Bucknall 1997). JR increases with Δa, but dJR/da, which defines the increase
in resistance of the specimen to stable crack growth during the initial stages of tearing, decreases as
the crack extends. At the same time, the strain energy release rate J increases as the load increases.
Crack growth becomes unstable when:
dJ dJ R
> ð6:218Þ
da da

The essential work of fracture (EWF) procedure is a technique to isolate the energy for crack
propagation (essential work) from the plastic deformation energy and is especially suitable for thin
specimens under plane stress conditions with extensive plastic deformation involved in the fracture
process (Karger-Kocsis 1996; Ramsteiner, Schuster and Forster 2001). However, it has later been
applied to different geometries, which is compiled in a comprehensive review on EWF by Bálány,
Czigány and Karger-Kocsis (2010). Two deep notches are made on each edge of the specimen/sheet,
leaving a ligament of length l (Fig. 6.88). By determining the fracture energy, (Wf), from tensile tests
at constant strain rate for specimens with different ligament length (Fig. 6.88), the two processes can
be separated and their size determined. This is done by considering that the essential work (we) in the
inner fracture process zone (IFPZ) closest to the ligament is proportional to the ligament length,
whereas the plastic work (wp) in the outer plastic dissipation zone (OPDZ) outside the IFPZ is
proportional to the square of the ligament length:
6.19 Fracture of Polymers 349

Fig. 6.88 The procedure to obtain the essential work of fracture

W f ¼ we lB þ wp βl2 B ð6:219Þ

By rearranging the above equation and plotting the fracture energy versus the ligament length, the
essential work is obtained from where the curve (Eq. (6.220)) meets the y-axis. β is a shape factor of
the yield zone (OPDZ).
Wf
¼ we þ wp βl ð6:220Þ
lB

Ductile fracture occurs for very short or long cracks. For intermediate crack sizes, fracture is brittle
and catastrophic. A polymer may be brittle or ductile, depending on the size of the critical crack, the
yield strength of the material and the critical stress intensity factor. The critical crack size defines the
upper limit of ductile failure and the transition to brittle failure. The effect of increasing the specimen
thickness is to move the critical crack size to lower values. Thus, the tendency for yielding decreases,
which is because the system approaches plane strain conditions. As an example, the very ductile
polycarbonate appears brittle if the specimen thickness is sufficiently large. Normally, there is a
combination of plane strain and plane stress, since the surfaces of the specimen experience plane
stress and the interior experiences plane strain.
KIC increases with strain rate (crack speed). The increase in fracture stress with strain rate is,
however, smaller than the corresponding increase in yield stress. Hence, a polymer tends to be more
brittle at higher strain rates and impact loading conditions, therefore usually causing brittle fracture.
Since polymers in the brittle mode are only slightly viscoelastic, their time-dependent modulus can be
described by a simple exponential relationship:
EðtÞ ¼ E0 tn ð6:221Þ

where n < 0.1 for most polymers (often below 0.05). n is considered constant over a large range of
both temperature and time, and it is proportional to the loss tangent. Since KI depends only on the
loading condition and specimen geometry, it is possible to obtain the time dependence of the fracture
condition:
350 6 Mechanical Properties

K IC 2 ¼ EðtÞ GIC ¼ δc EðtÞ σ c ðtÞ ð6:222Þ

The type of fracture of a polymer that is activated at a specific temperature depends on the relative
sizes of the yield and fracture stresses and yield and fracture strains. Above 75  C, PMMA shows both
a yield and fracture strain and is relatively ductile. For polypropylene, the brittle/ductile transition
occurs at 40  C. Note that the yield strain increases with temperature for polypropylene, but
decreases with temperature for PMMA. The yield stress, on the other hand, decreases with tempera-
ture. The yield stress at temperatures below the brittle/ductile transition has been determined in
compression tests where brittle fracture is absent (Vincent 1961). The extrapolated value for the
fracture stress at 0 K is 280 MPa for both polymers, and it has been suggested that all polymers have
the same 0 K fracture stress. This ultimate stress can be obtained either by lowering the temperature or
by using an infinite strain rate. The ultimate fracture stress is determined by the strength of the bonds
on the individual chains since disentanglements or segment rotations are absent at zero absolute
temperature.
Kausch et al. (2001) evaluated the fracture behaviour of polypropylene using three approaches; the
J integral, essential work of fracture and LEFM with the addition of a plastic zone. It was concluded
that the modified LEFM was superior to the other methods and yielded useful effective toughness and
energy release rate parameters. The drawback with the J-integral is that it is elaborate and the size of
the crack during testing is difficult to determine in a non-transparent and tough material like
polypropylene. The drawback with the EWF approach is that it is difficult to assess the total work
used in the procedure from a stress strain curve of a tough material like polypropylene (Kausch et al.
2001).

6.19.3 Ductile Failure

Ductile failure/fracture is governed by shear stresses that are acting at a 45 angle to the draw
direction, whereas normal stresses perpendicular to the draw direction are operative in the brittle
failure mode. Below TBD, molecular motions connected with either the glass transition or the
ß-transition are frozen-in, and the failure is brittle. Above TBD, motions connected to either one or
both types of transition are activated, and the fracture is ductile. If the temperature is increased from
room temperature to 40  C, PMMA, which is brittle at room temperature, can yield before failure
(Arridge 1975). At 60  C, necking is observed, and the polymer fails in a ductile manner at a strain of
more than 100%, even though the polymer is still glassy. In fact, cold drawing of PMMA can be made
at a temperature between the β-transition temperature and Tg (Allison and Andrews 1967). Evidently,
the brittle/ductile transition temperature for PMMA is below Tg (105  C). The brittle/ductile
transition temperature for PMMA is closely related to the ß-transition temperature and the molecular
processes which are activated there. This is the reason why polycarbonate and PVC also become
ductile far below their glass transition. For polymers that lack a strong ß-transition (e.g. polystyrene)
or do not have a β-transition (e.g. natural rubber), the brittle/ductile transition is connected to the glass
transition.
In cases of strain hardening, i.e. when the polymer chains are oriented during deformation, the
nominal fracture stress is usually higher than the nominal yield stress. The true fracture stress is then
always higher than the yield stress. Nevertheless, in applications where plastic deformation is
unwanted, the yield stress is the more important ‘design’ parameter. In cases where yield is to a
certain degree acceptable, for example, for synthetic fibres in yarns or ropes, the size of the fracture
stress determines the suitability of the material. A strain hardening polymer can sustain higher-impact
loads and is capable of spreading the impact over larger volumes compared to a material which does
6.19 Fracture of Polymers 351

not strain harden. To obtain strain hardening, films and fibres are stretched after extrusion. For
example, PET fibre is relatively brittle at low degrees of chain orientation, but after large
deformations at low stress levels, strain hardening occurs and the material becomes strong and
tough. If the PET fibre is stretched several hundred per cent at higher temperature, where chain
disentanglement and recrystallisation can occur, a flexible, elastic material is obtained with far higher
fracture stress compared to the original material. The reason for the high fracture stress in this case is
due to the crystalline morphology. In an unoriented glassy semicrystalline polymer, for example,
PET, the presence of crystals decreases the fracture stress and strain. This is because the crystal
surface acts as a stress concentrator, and the molecular mobility in the amorphous phase is lower than
in a fully amorphous material (Grellman 2001). The glassy amorphous phase has limited possibilities
to relieve the stresses, and the molecular packing is constrained in the rigid amorphous region next to
the crystal. In oriented recrystallised PET, there are more tie chains, and therefore, crystal separation
demands a higher stress (Cagiao et al. 1993; Stribeck 1993). The oriented polymer becomes brittle in
the direction transverse to the draw direction due to the crystal fold surface plane being located
primarily in this direction.
For rubbery semicrystalline polymers, the ductility is sensitive to molar mass, molar mass
distribution and degree and type of chain branching and branch distribution. At low molar mass,
the number of tie molecules and trapped entanglements is low, and the crystals can, therefore, more
easily glide apart (Fig. 6.89) (Nilsson et al. 2012). Hence, the polymer becomes brittle.
By introducing chain branches, the crystals become thinner, and the number of tie molecules and
trapped entanglements increases compared to a linear polymer with the same molar mass (Fig. 6.90)

Fig. 6.89 Simplified

sketch of the lamellae
structure of polyethylene.
The low molar mass
polymer (left figure)
contains less/no tie
molecules (T) compared to
the higher molar mass
polymer (right figure).
Grey areas correspond to
crystals (chain tilt in the
crystal is not shown).
Trapped entanglements and
loose folds (loops) in the
right figure are not shown

Fig. 6.90 The same type

of sketch as in Fig. 6.89
(with the same
simplifications); it
illustrates in this case that
the number of tie molecules
increases for the branched
polymer (left) compared to
the linear analogue (right)
with the same molar mass
and that the majority of
branches are located
outside the crystals
352 6 Mechanical Properties

(Tr€ankner, Tr€ankner and Hedenqvist 1994). The associated lower crystallinity leads to a weaker
material (lower yield stress). The chain branches decrease the tendency for chain slippage through the
crystals when the material is stressed, which leads to the branched polymer becoming tougher than
the linear analogue.
Uniaxial orientation of a semicrystalline polymer increases the fracture stress due to the larger
number of strained tie molecules and entanglements/knots in this direction. At the same time, the
fracture strain decreases in the orientation direction. The fracture strain in the direction perpendicular
to the direction of initial orientation increases due to chains/crystals reorienting in the new stress
direction.
The fracture behaviour of an elastomer depends on several factors. The fracture stress is higher for
a rubber that crystallises than for those that do not. Further, it increases with molar mass in a
non-linear way, approaching a plateau value at high molar mass. On the other hand, the fracture
stress has a peak value at intermediate degrees of crosslinking, which is due to the number of polymer
chain segments that can orient in the draw direction and contribute to increased strength decreasing
with crosslink density.

6.19.4 Creep Failure

Creep failure refers to fracture in components which are exposed to a constant load. This mode of
failure is important to control and avoid in, for example, applications where plastic pipes are being
used. This includes the distribution of natural gas where the pipes are exposed to an internal
hydrostatic pressure. It is required that plastic pipes should last for 50 years up to even 100 years
(Gerets et al. 2017). In order for this to work, it is important that the failure mechanisms under static
loading are fully revealed. Creep rupture data of pipes exposed to an internal pressure are generally
presented in a log σ–log t plot, where σ is the circumferential/hoop stress in the pipe wall and t is the
lifetime (Fig. 6.91). Three different regions are often observed in, for example, polyethylene pipes
(Viebke et al. 1994).
Stage I refers to failures at high internal pressure, and the pipe lifetime is here a strong function of
the stress level. Stage I or ductile failure is characterised by large macroscopic deformation/yielding
and/or whitening prior to fracture (Fig. 6.92). A large ‘bubble’ is formed along the pipe, and the final
crack is usually growing in the circumferential direction. It should be mentioned that at very high
pressures, 50 MPa, pipes may fail in a brittle mode without any large-scale yielding, as has been
observed for PVC pipes (Kausch 1987).
At low stresses and long exposure times, the pipe fails due to chemically induced embrittlement
(Stage III). Stage III occurs at high temperature and/or in oxidative environments. The region in

Fig. 6.91 Typical creep

rupture curve for
polyethylene pipe exposed
to internal pressure
6.19 Fracture of Polymers 353

Fig. 6.92 Ductile (Stage I)

failure of a
polyethylene pipe

Fig. 6.93 Brittle (Stage II) failure of polyethylene (uniaxial notch test). Fracture surface contains numerous fibrils that
increase in size from right to left in the figure, i.e. in the direction of crack propagation (Tr€ankner et al. 1997). Reprinted
with permission, Copyright # 1996 Society of Plastics Engineers

between Stages I and III is referred to as Stage II. Stage II is characterised by a fracture which occurs
without, or with only little, macroscopic deformation and no chemical degradation. It is often initiated
at an existing defect. Stage II and Stage III are presented below together with a few measuring
techniques.
In Stage II, propagation may be slow (slow crack growth, SCG) or rapid (rapid crack growth,
RCG) (Leis 1989; Tr€ankner, Hedenqvist and Geede 1997). It occurs at low stresses (of the order of
5 MPa) and after long times. SCG is far more common than RCG. A typical SCG fracture surface is
shown in Fig. 6.93. The many fibrils present indicate that the failure is microscopically ductile. Slow
crack growth can be divided into an induction period (crack initiation) followed by stable and
unstable crack propagation periods. The induction period is considered to be the most time-
consuming step in the fracture process; therefore, in order to study SCG, specimens are notched to
limit the induction period. Although it is difficult to differentiate between crack initiation and growth
in, for example, hydrostatic pipe tests, there are ways to separate them (Redhead et al. 2013). The
crack starts at a defect that may be a flaw or a void or simply a scratch that originates from the
handling of the pipe.
The crack propagates in a stable manner through the formation of a craze. At a specific crack
opening, the fibrils break, mainly due to chain gliding and disentanglement, and the crack grows
further into the specimen. Unstable crack growth occurs when the last few intact fibrils are unable to
carry the load. This part is usually characterised by long stretched fibrils (left part of Fig. 6.93). Slow
crack growth has been studied extensively with several methods (Bradley, Cantwell and Kausch
(1998), Redhead et al. (2013)). Internal pressure testing where a pipe section is simply pressurised
from the inside is among the most common methods but is also the most complex method to evaluate
due to the multiaxial stress state. Alternative methods include those based on uniaxial tests. These can
354 6 Mechanical Properties

be performed on dumbbell or straight specimens, preferably notched by a razor blade on (usually) one
side (single edge notch test/Pennsylvania edge notch tensile test (PENT)) or on all four sides (full
notch creep test (FNCT)), either in static creep or under cyclic load. Cyclic loading on cracked round
bar (CRB) specimens has been performed to accelerate the testing, followed by extrapolation to static
load (Redhead, Frank and Pinter 2013). The advantage of having a circular geometry is that the test
condition is closer to LEFM, since plane strain conditions prevail during crack growth (Arbeiter et al.
2017). For CT specimens, there is a mixture of plane strain (in the centre) and plane stress (at the
edges) conditions, which increases the plastic zone and leads to larger deviations from LEFM. For
plane strain conditions, it is possible to use a Paris–Erdogan type of equation involving the stress
intensity factor, modified for static conditions, to estimate the sample slow crack growth lifetime
(Arbeiter et al. 2017). The constant tensile load test (CLT) is performed on a ring that is cut out from
the pipe. The ring is notched on both the inside and the outside perpendicular to the load direction.
The test is normally, for example, as in the FNCT test, used at elevated temperature with also the
addition of a surface energy reducing agent. An alternative to CTL is the constant tensile strain test
(CTS) which is similar to CTL with the exception that the specimen is exposed to constant strain
rather than constant load (Lustiger 1986). The latter test yields information about the ability of the
material to relax stress concentrations around defects. Examples of other methods to reveal creep
failure are three-point bending and slow strain rate tests. Due to today’s very tough polyethylene
grades, tests that can shorten the evaluation period drastically have been developed (including the
CRB test). Accelerated FNCT (aFNCT) is achieved by a slightly higher temperature than in the
FNCT test combined with a more efficient surface reducing agent (Gerets et al. 2017). The strain
hardening test (SHT) is, in relative terms, a very rapid test (within hours), which has been shown to
correlate well with stress cracking resistance (Gerets et al. 2017). The strain hardening modulus
obtained in the strain hardening region on tensile-tested specimens at elevated temperature and
constant strain rate (Fig. 6.94) is linearly related to the failure lifetime determined with the FNCT/
aFNCT methods. Further slow crack growth tests are given in Gerets et al. (2017).
Based on numerous measurements on polyethylenes tested in the notched uniaxial test, Brown and
co-workers showed that the failure time could be calculated from the following empirical equation:

Q
tF ¼ C σ n a0 m exp ð6:223Þ
RT

where σ and ao are the applied stress and initial notch depth, respectively, and Q is the activation
energy of SCG (Ward et al. 1991). N and m are constants, and C is a parameter characterising the

Fig. 6.94 Definition of the strain hardening modulus

6.19 Fracture of Polymers 355

fracture resistance of the polymer. C increases with increasing molar mass and increasing degree of
chain branching. The higher the molar mass or the higher the degree of chain branching, the slower is
the chain disentanglement rate and the slower is the crack growth.
Typical notched uniaxial test data are presented in Fig. 6.95 in the form of crack opening
displacement as a function of time. In this case, the crack opening is measured between two points
that are located on the specimen surface on opposite sides of the crack. A fracture/crack initiation
time (tF) is defined as the period from instant loading until the first fibrils break (close to the inflection
point at approximately 2000 h in Fig. 6.98).
When polyethylene pipes are exposed to high internal pressures at low temperatures (usually
below 0  C), they can fail in a rapid crack growth (RCG) mode (Leis 1989). The rate of crack
propagation can reach 400 m/s, and the crack oriented along the pipe usually has a sinusoidal pattern.
Crack initiation is induced at defects in the pipe, for example, flaws or surface scratches. RCG can
propagate several hundred metres and cause severe damage. A critical internal pressure usually exists
below which RCG does not occur. It has also been shown that below 15  C, the presence of a notch
is not necessary to produce RCG failure in polyethylene pipes (Leis 1989). The cause of RCG is not
fully understood, but it is believed that internal stresses in the material play a role in its occurrence.
Several techniques exist for measuring RCG. Standard impact tests, such as Charpy V (see below),
and the full-scale tests are examples. In the latter method, a pipe is notched, and the fracture is
produced by placing explosives inside the pipe (Tr€ankner 1990).
Creep failure times have been theoretically defined from different failure criteria, including
energy-based, time-dependent yield and viscoplastic models (Spathis and Kontou 2012). The latter
model, which is able to predict the stress dependence of the failure time for, for example, polycar-
bonate and polyamide 6,6, also considers viscoelastic contributions, especially at low strains.
Fracture that occurs in polymers under combined physico-chemical exposure and mechanical load
is referred to as environmental stress cracking (ESC) (Lustiger 1986; Birley et al. 1992; Robeson
2013). It is a serious problem since it causes premature failure in many products and can be difficult to
predict beforehand. Examples of ESC cases are polycarbonate exposed to cyanoacrylate adhesive,
poly(vinyl chloride) exposed to benzene and polyethylene exposed to soap (Robeson 2013). The
reason for ESC is that the active ESC medium/liquid swells the polymer and facilitates crazing and
chain disentanglement. In addition, surface active agents reduce the surface energy and facilitate the
formation of new surfaces and, consequently, crack propagation. Not only ‘typical’ solvents and
detergents have ESC potential, but ‘permanent’ or supercritical gases can also reduce the critical
strain and stress for craze initiation if their concentration in the polymer is high enough. Their
solubility increases as the temperature approaches the condensation point and also with increasing gas
pressure. It has been shown that ESC is absent in He and in vacuum. In order to limit the risk of ESC,
the material should have a high molar mass. This increases the number of chain entanglements and
the number of connecting points between different crystals, i.e. the number of tie molecules. Short-
chain branches are produced if hexene-1, for example, is copolymerised with ethylene. These
branches counteract chain disentanglement and produce a larger number of tie molecules
(Fig. 6.90), which increases the ESC resistance. Semicrystalline polymers have generally better
ESC resistance than fully glassy polymers. ESC can also be reduced by blending the polymer with
a rubber component (e.g. high-impact polystyrene), adding fillers such as glass fibres, crosslinking or
addition of a less ESC-sensitive polymer (ABS addition to polycarbonate) (Robeson 2013). Choosing
a polymer with a large difference in solubility parameter compared to the environment (e.g. a liquid)
will reduce ESC. It should be noted that ESC plays a lesser role at high stress levels, when the fracture
is fast, and also that ESC chemicals can both reduce the crack initiation time and increase the crack
propagation rate (Andean et al. 2013).
356 6 Mechanical Properties

Fig. 6.95 Crack opening

displacement as a function
of time at 100  C and
3.12 MPa for a medium
density polyethylene.
Drawn after Tr€ankner et al.
(1997)

Fig. 6.96 Test geometry for Charpy (a), Izod (b) and tensile impact (c) methods and the pendulum principle (d)

6.19.5 Impact Failure

Impact resistance (impact strength) is one of the most important properties for polymers and is
important in many applications, including cushionings, airbags and car fenders. It is also a property
that is strongly dependent on test method and specimen and test geometry, which makes it difficult to
directly compare data from different sources. In addition, it has been shown that the quality and
repeatability of notching can be a significant source of scatter in determining impact properties
(Agnelli and Horsfall 2013). The test methods available are divided into pendulum methods and
falling weight methods. There are three main types of pendulum method (Fig. 6.96). The Charpy test
is a high-speed three-point flexure test, the Izod test is based on a notched cantilever geometry and the
tensile impact test represents a high strain rate version of the conventional tensile test. Any friction in
the system will be detected by letting the pendulum swing freely without any specimen. The
maximum height of the swing will represent zero impact energy. In the presence of a specimen, the
pendulum, with mass (m), will lose potential energy, and the difference in ‘swing height’
Δh corresponds to the energy absorbed by the specimen (Ua), the kinetic energy of broken polymer
fragments (Uk) and machine/testing related energy absorption (Um):
m g Δh ¼ U a þ Uk þ Um ð6:224Þ

The notched impact strength, Sa, is impact energy per unit ligament area:
m g ðh1  h2 Þ  U k  U m
Sa ¼ ð6:225Þ
b ð w  aÞ
6.19 Fracture of Polymers 357

Fig. 6.97 Impact strength

as a function of notch
radius for PVC (a), PA (b),
ABS (c), POM (d) and
PMMA (e). Drawn after
Vincent (1971) and Nielsen
and Landel (1994)

where b, w and a are specimen width, thickness and notch length, respectively (Fig. 6.96) (Birley,
Haworth and Batchelor 1992). The total energy absorbed in the polymer consists of parts from crack
initiation and crack propagation. Like in creep measurements, crack initiation can be shortened by
using a specimen geometry that leads to a triaxial stress state (plain strain) or by the use of notched
specimens. The stress concentration due to the notch is expressed by the stress concentration factor
(SCF; compare with Eq. (6.187)):
 0:5
σ a
SCF ¼ c ¼ 1 þ ð6:226Þ
σ r

where σ c is the concentrated stress at the notch tip and σ is the nominal remote stress. r is the notch tip
radius. This equation predicts an infinite stress at an infinitely small radius, a case which is clearly
erroneous since local yielding in the notch tip limits the stress. Another limiting complication is that
Eq. (6.226) does not consider the change of mode of failure which may accompany changes in the
specimen geometry. Impact strength usually increases rapidly with increasing notch radius. When
comparing different materials in terms of impact resistance, it is of utmost importance that the notch
depth and geometry are identical for all materials. The reason for this is that the notch sensitivity
differs greatly between different polymers, as can be seen in Fig. 6.97. Rigid PVC, PC, POM and PA
are more notch-sensitive than, for example, ABS.
In the tensile impact method, the tensile movement is obtained by either a falling pendulum or gas
pressure. Visser, Caimmi and Pavan (2013) used a straight tensile impact specimen with a notch along
the thickness (edge) perpendicular to the strain direction on each side of the specimen. In contrast to
the Izod and Charpy tests, the deformation rate of the tensile specimen is essentially constant.
A drawback with the pendulum-type methods is that the test conditions do not resemble the impact
condition for the product in service. Therefore, more qualitative falling weight impact methods have
been developed (Birley, Haworth and Batchelor 1992). The polymer product (e.g. a pipe) is exposed
to a vertically falling load. The method is frequently used to establish the falling weight distance
which produces fracture in 50% of the specimens. This is obtained by either increasing the distance if
fracture is absent or decreasing the distance if fracture occurs until 50% of the specimens have failed.
Other parameters which may be varied in the test are impact energy and velocity, temperature and
specimen geometry. A drawback with the method is that the often complex specimen geometry
generates a complex stress state. The method is relevant for a number of different products, including
extruded pipes, sheet, PVC window profiles, biaxially oriented PE or PET films and blow-moulded
PE drums. Another falling weight multiaxial impact test method involves a clamped disc specimen
that is hit by a hemispherical striker in the centre. With this impact method, the DSGZ
(Duan–Saigal–Greif–Zimmerman) model has been successfully verified to describe ABS impact
behaviour in a broad range of strain rates, up to the maximum impact load (Duan et al. 2002). It
shows the necessity of including thermomechanical coupling (adiabatic effects) and hydrostatic
358 6 Mechanical Properties

Fig. 6.98 Instrumented falling weight impact test data for some commodity thermoplastic materials, obtained at 3 m/
s with a hemispherically ended missile of mass 25 kg. Force–deflection curves for an extrusion blow-moulded HDPE
container (a), an injection-moulded rubber-modified PS (b) and an injection-moulded unmodified PS (c). Fp and δp are
peak force and deflection. Drawn after Birley, Haworth and Batchelor (1992)

Table 6.3 IFWI data for car fenders (26  C) (Birley et al. 1992)
Thickness Fp Up UF
Material (mm) (N) (J) (J)
Polyurethane 3.7 583 5.8 9.5
Modified polybutylene terephthalate 3.4 820 2.9 5.2
PP homopolymer 2.6 1183 10.1 16.5
Glass-reinforced polyester 3.4 1324 4.1 14.5
PBTP/PC blend 3.2 2500 21.0 29.5
HDPE 4.1 1526 13.2 23.0

effects in describing impact behaviour in many cases. In instrumented falling weight impact tests,
IFWI, a transducer is fitted to the impactor head which measures the force during the impact event,
which lasts for a few milliseconds. The force–time data can then be analysed on a computer. If it is
assumed that the compactor head moves with a constant speed through the specimen, the force–time
data can be transformed into force–deflection data. This allows for the determination of stress, strain
and impact energy (Fig. 6.98).
Table 6.3 presents IFWI low-temperature data on ‘competitive’ materials in a car fender obtained
at a velocity of 3 m/s. The data have been normalised to 1 mm thickness to account for the effects of
thickness variations between different systems. Up and UF are peak and fracture energy.
If the material follows LEFM (a linear load–deflection curve, often the case when the notch is very
sharp), it is possible to obtain the material parameter fracture toughness (Gc) under impact conditions
(or KIC (Visser et al. 2013)). It is assumed here that the modulus of the polymer is independent of
strain rate or that the modulus–strain rate relationship is known. Finally, it is assumed that the fracture
energy, crack depth and modulus are explicit functions of each other. If all these conditions are
fulfilled, it is possible to estimate the total energy absorption at failure, which is the area under the
linear load–deflection curve with a force Ff at a failure deflection (δ) (Birley, Haworth and Batchelor
1992):

δF F 2
U F ¼ FF ¼D F ð6:227Þ
2 2

where the specimen compliance is:

6.19 Fracture of Polymers 359

δF
D¼ ð6:228Þ
FF

When the impact energy reaches a critical value (UF), the released strain energy (Gc, referred to as
critical strain energy release rate by Plati and Williams (1975)) is:
1 dU
Gc ¼ ð6:229Þ
b da

Rearranging Eq. (6.229) together with Eq. (6.227) yields:

FF 2 dD
Gc ¼ ð6:230Þ
2b da

Provided an accurate value of UF can be measured, it is meaningful to write:

U F ¼ Gc bw ϕ ð6:231Þ

which is a combination of Eqs. (6.227) and (6.230). ϕ is a dimensionless factor related to the
compliance and specimen geometry:
" #1
dD
ϕ ¼ D a ð6:232Þ
d w

ϕ can be determined for a range of different geometries. Plati and Williams determined solutions for
ϕ appropriate to Charpy and Izod tests, and they concluded that the fracture mechanics approach is a
helpful tool for all polymers except highly ductile materials. The most common procedure for
obtaining Gc is to vary specimen or notch dimensions and to plot the values of UF as a function of
bwϕ. The slope then yields Gc. Table 6.4 presents Gc values for a range of thermoplastics, and impact
strength generally increases downwards in the list.
At least up to moderate impact load rates (~1 m/s), the crack propagation can be described as going
through three phases. In phase I crack initiation and crack tip blunting occur with a rapid increase in
growth rate, and in phase II the crack propagates in a ‘non-stationary’ stable manner ending up in
phase III where the crack growth rate is high and constant (Lach and Grellman 2017). For brittle
polymers, unstable crack growth occurs without passing phase III, and for very ductile polymers,
constant crack growth also does not occur due to extensive crack tip blunting and the associated large
plastic zone.
It is well-known that the strength of low-temperature relaxations/transitions, associated with main
chain motions (β and γ transitions), is related to the level of the impact strength. The increased
mobility of the main chain, rather than side chain motions, provides high-impact strength (Ramsteiner
1983). Whenever a strong ß-relaxation exists, it coincides with the impact brittle–ductile transition.
PVC, PC and PET all have strong ß-transitions connected with main chain motions (as mentioned

Table 6.4 Gc for Material Gc (kJ/m2) Type of fracture

thermoplastics from
PMMA 1.3 Brittle
Charpy tests (Plati and
Williams 1975) PVC 1.4 Brittle
HIPS 16 Ductile
LDPE 35 Ductile
ABS 49 Ductile
360 6 Mechanical Properties

Fig. 6.99 The effect of

water on the impact
strength of PA6,6. The
material has been kept in
cold distilled water for
0 (●), 1 (○) and 4 (■)
weeks. Drawn after
Vincent (1971)

above). For example, the ß-transition in PVC is associated with the motion of the chlorine atom as a
result of main chain segment rotations. PIB, PS and PMMA lack strong low-temperature main chain
relaxations, and the brittle–ductile transition is, therefore, associated with the onset of the α-transition
(glass transition). High-impact strength is obtained if the polymer chains are able to rotate through α
or ß rotations and relax the imposed stress. At larger strains, stress is relieved through chain
reorientation by chain segment gliding, void formation and even chain scission. Large free volume
and high molar mass favour impact strength. The former provides high chain flexibility, and the latter
retards chain disentanglement. Weak secondary bonds favour chain flexibility and thus provide high-
impact strength. The number of strong polymer–polymer hydrogen bonds in PA6,6 is reduced by the
presence of plasticisers, for example, water (Fig. 6.99). High crosslink density decreases chain
flexibility and increases the risk of elastic energy concentration. A highly crosslinked material is,
therefore, brittle under impact.
The first important polymer blend was toughened or high-impact PS (HIPS) (McCrum, Buckley
and Bucknall 1992). By blending PS with rubber particles, an increase in ductility was obtained. For
optimal ductility it was observed that adhesion had to be good between the blend components. In the
1950s, Dow Chemical company introduced a high-impact PS (HIPS) where styrene–butadiene rubber
particles (10%) were blended with styrene during the PS polymerisation. In this way, styrene was
grafted onto the rubber particles (using the double bonds in the rubber), and this ensured good
adhesion between PS and the rubber. Even higher impact strength was obtained by replacing styrene
with styrene–acrylonitrile (SAN) polymer and blending it with different rubber copolymers. The
trade name for this family of polymers is ABS (acrylonitrile–styrene–butadiene). Both HIPS and ABS
contain rubber particles which are cellular and contain embedded PS or ANS nuclei (illustrated for
HIPS in Fig. 6.100). The cellular/salami morphology is due to incomplete phase inversion, and the
cellular particles resemble the morphology prior to phase inversion. The average rubber particle size
in ABS is smaller than in HIPS. On impact, a large number of small crazes are initiated at the largest
rubber particles (a few microns in size), and they absorb the major part of the impact energy. The
rubber particles also act as craze terminators, and the final morphology is, therefore, a network of
crazes which extends between rubber particles (Fig. 6.100). The large number of crazes causes stress
whitening since the craze structure scatters light. In ABS, besides crazing, shear bands are initiated
on impact in the directions of maximum shear stress, and they terminate crazes as efficiently as the
larger rubber particles in HIPS.

6.19.6 Dynamic Fatigue Failure

Polymeric products are often exposed to cyclic stress, which usually results in earlier failure (fatigue)
than under static conditions. Hence, it is important to understand fatigue and how to design against
6.19 Fracture of Polymers 361

Fig. 6.100 Schematic illustration of a HIPS system which shows bands of crazes which extend from rubber particle to
rubber particle

Fig. 6.101 Stress- (left) and strain- (right) controlled fatigue test. The maximum (max), minimum (min) and mean
values are displayed, as well as the amplitudes (subscript a). The x-axis is time. Drawn after Bier€
ogel and Grellman
(2014a, b, c)

premature fatigue failure. The response of polymers to cyclic loading is, however, generally more
complex than for metals because of the viscoelastic nature of the polymer. The evaluation of fatigue is
performed by exposing the material to a stress-controlled or strain-controlled test (Fig. 6.101). The
type of load (sine, triangular, square, random) and the frequency, stress/strain amplitude and mean
stress/strain are selected. The loading can be purely alternating with a mean load of zero, or skewed to
non-zero mean stresses. The test can be made under uniaxial load, flexure or torsion. Because of the
viscoelastic properties, the cyclic strain response in the stress-controlled system and the cyclic stress
response in the strain-controlled system can both experience an overall drift with time due to,
respectively, creep and stress relaxation (Bier€ogel and Grellman 2014a).
The most common way of representing the dynamic fatigue, or simply fatigue, properties is by
drawing W€ ohler diagrams (S-N-diagrams). The stress amplitude (assuming no drift in mean stress) is
plotted against the number of cycles to failure. Below a critical stress amplitude, the material does not
fail. This is referred to as the fatigue limit or fatigue/endurance strength. According to Kausch (1987),
the W€ohler diagram may be divided into three regions where the first region (I) is characterised by
immediate crazing, and in region II craze formation is preceded by a craze incubation time. The long
362 6 Mechanical Properties

Fig. 6.102 W€
ohler curve for annealed polycarbonate tested in uniaxial tension at 2 Hz (maximum stress is the
maximum in the sine wave with a minimum stress of 2.2 MPa). The solid and broken curves represent the thermal and
mechanical mode/region (the actual experimental data points are not shown). Drawn after Janssen et al. (2008)

life region (III) forms essentially the endurance limit of the material. Crack initiation has a very long
incubation period here. W€ohler curves for a substantial number of polymers and fatigue testing
conditions, showing more or less clear fatigue limits, were compiled by Bier€ ogel and Grellman
(2014b, c).
Fatigue failures can be further divided into two modes. At high stress amplitudes, thermal fatigue
due to the damping properties of the polymer occurs, which is characterised by thermally induced
softening or yielding before a crack is formed. At low stress amplitudes, crack initiation and
propagation occur without any signs of yielding. The two failure modes are denoted, respectively,
thermal and mechanical/brittle fatigue. These two modes/regions are illustrated in Fig. 6.102 for
annealed polycarbonate, showing a clear transition at 56 MPa. It is interesting to note that the same
constitutive model for describing the deformation of polycarbonate at constant load could be used to
describe its behaviour under cyclic load in the mechanical fatigue region (Janssen et al. 2008). It
consists of a spring and dashpot in series (representing intermolecular interactions) coupled with a
spring in parallel, representing strain hardening. This type of model has also been considered by
Winkler and Kloosterman (2015).
In brittle fatigue, the crack propagation can often be described by the Paris relationship, or
modifications of it, which was originally developed for metals (Nielsen and Landel (1994):
da
¼ Af ΔK m ð6:233Þ
dN

where ΔK is the stress intensity factor interval (ΔK ¼ Kmax  Kmin). Af and m are constants. For many
polymers and metals, m  4, but there are exceptions. In brittle fatigue, the craze propagation is
continuous, but the crack propagation is discontinuous due to the discontinuous rupture of the craze
fibrils. This results in a crack surface that contains concentric rings or parallel strips. A clamshell
pattern can be observed, but also significantly different patterns as in epsilon growth. Here, the main
6.20 Cellulose Fibre Systems and Related Materials 363

crack direction involving crazing is accompanied by shear bands deviating from the main crack
direction (Takemori 1984).
An increase in molar mass increases chain entanglement density and stabilises craze ligaments.
Hence, the fatigue lifetime increases with molar mass. Chemical crosslinks prevent craze formation
and lead to a shorter fatigue lifetime. The most fatigue-resistant glassy polymers are those that
contain stiff but mobile molecules. The high mobility ensures relaxation of stresses at the crack tip.
PC, PSU and PPO are examples of more fatigue-resistant glassy polymers, whereas PMMA and PS
are brittle fatigue-sensitive polymers. Semicrystalline polymers are generally more fatigue resistant
than glassy polymers due to additional fatigue energy being required to deform crystals. The fatigue
behaviour is, as with other mechanical features, also temperature dependent, as observed by, for
example, a reduction in the fatigue limit (Mura, Ricci and Canavese 2018). Interestingly, and not
really expected, notches can have a ‘supporting’ effect on the fatigue properties (Park et al. 2015).
Smooth, un-notched specimens can have a lower fatigue limit than notched specimens under uniaxial
load. This is explained by the resulting lower stress at the surface in the latter specimens.

6.20 Cellulose Fibre Systems and Related Materials

Cellulose fibre systems (including paper, paperboard and cellulose nanofibres) deserve a special
section here, since the normally porous material structure gives rise to mechanical properties different
from solid polymer materials and also may require different mechanical parameters. There is no strict
definition of paper. However, its principal component is always plant fibres from, for example, wood,
grass or cotton. Paper is a layered structure of flattened fibres together with smaller amounts of fibre
fragments. Some qualities of paper contain additives, for example, kaolin clay or talc. These fillers are
added to increase the whiteness of the paper and to decrease the price. Since the material is built up of
macromolecules (mainly cellulose, hemicellulose and lignin), the fibre systems show viscoelastic
properties. Figure 6.103 shows a tensile stress–strain curve of a cellulose nanofibre (CNF) material.
Observe that we use the term fibre here, although it is actually a fibril system (fibrillated fibres), if not
a native nanofibre system (e.g. bacterial cellulose). The absence of a clear peak indicating the yield
point is because the fibre system does not form a neck. The porous fibre network in typical paper gives
rise to only a weak or negligible thickness dependence of the mechanical properties. If the paper is

Fig. 6.103 The stress (FT/cross section) and the tensile index for a CNF paper. Data from Plackett et al. (2010)
364 6 Mechanical Properties

compressed to, for example, half the original thickness, the force necessary to generate fracture is
essentially unchanged. However, the fracture force depends on the mass of material per unit surface.
In addition, it is difficult to determine an ‘effective’ thickness of a very rough surface, as in the case of
normal pulp fibre paper systems. Hence, surface weight (grammage) is a more useful parameter than
the paper cross section when determining, for example, the fracture strength. The tensile index is
defined as (Fellers 2009):
FT
σw
T ¼ ð6:234Þ
bw

where FT is the tensile fracture (maximum) force and b is the width of the straight tensile specimen.
w is the surface weight. If compression tests are performed, then the force is expressed with the index
c. The unit for the tensile index is Nm/kg. The example given in Fig. 6.103 is for CNF ‘paper’, and not
a typical paper. It is considerably more dense than normal pulp fibre paper, which is evidenced by a
density of ca. 1400 kg/m3 and a more than an order of magnitude higher tensile index. In such a dense
system, it is more meaningful to use the ‘normal’ strength (force per cross section) than in a typical
paper. Also, the strength, the stiffness (tensile stiffness index) and the fracture energy (tensile energy
absorption index) can be expressed based on the surface weight, rather than the cross-sectional area
(Fellers 2009). The strength of paper is sometimes also expressed as ‘length’, which should be
interpreted as the specific length of a paper strip when it breaks due to its own weight. It is obtained by
dividing the tensile index with gravity.
The strength of a fibre system, like paper, depends on the fibre strength and the strength between
fibres. Based on this, the theory by Page was developed, which relates these properties to the paper
strength/tensile index (Page 1969; Van Der Akker et al. 1958) (Fig. 6.104):
1 9 1 12
¼ w þ τ lα ð6:235Þ
σw
T 8 σ ZS s

where the first and second terms to the right represent fibre strength and fibre-to-fibre strength,
respectively. σ w
ZS is the zero-span tensile index, which means the strength when the two clamps in the
tensile tester are initially in contact. This means that the fibres bridging the clamps are strained
without the possibility to be pulled out from the paper. Hence, σ w ZS corresponds to the intrinsic fibre
strength. The number in the term originates as a consequence of the fibres being oriented randomly in
the plane and not all in the tensile direction (Fig. 6.104). It is also due to corrections for there being no
possibility for the paper to shrink in the cross section (Poisson’s ratio for paper is theoretically 0.33)
during the zero-span test, which is different from a normal tensile test or a real situation (Page 1969).
In the second term, τs is the shear stress needed to break the bond between two fibres in contact. l is
the fibre length and α is the bonded area between fibres (area per mass of fibres). The latter can be
obtained from light scattering measurements. The equation clearly shows, in accordance with
experimental data, that the strength of paper increases linearly with the fibre strength, fibre length
and fibre-to-fibre bond strength (Page 1958). The strength of a paper is highest when there is a
combination of fibre fracture and fibre pull-out (Fig. 6.104). Full derivation of Eq. (6.230) can be
found in Page (1969) and Van Der Akker et al. (1958). Fibres are often lightly beaten to create a larger
fibre-fibre interfacial area and, consequently, a stiffer and stronger paper.
The special structure in paper leads to some unique characteristics (Fellers 2009). In contrast to
solid polymers, paper is stronger and more ductile (higher fracture strain) in tension than in
compression. Creasing is an indentation process that is used to, subsequently, be able to fold, for
example, a plane paperboard into a box-shaped packaging. Very conveniently, such an indent does
not have a critical effect on the strength of paper/board. Tearing is also uniquely tested for paper
products as a means of determining its fracture properties and crack resistance. It can be evaluated
with a pendulum apparatus in a similar way as impact resistance tests performed on solid polymers.
6.20 Cellulose Fibre Systems and Related Materials 365

Fig. 6.104 Randomly in-plane oriented fibres in a paper, before and after fracture. The grey fibres are those that break,
and the thick fibres are those that are pulled out from the opposite side of the crack

Cellulose fibre materials, as well as wood, take up a sizeable amount of water, and this leads to a
phenomenon referred to as mechano-sorptive creep (MSC) (Olsson and Salmén 2014; Dinwoodie
2000). In periodically varying relative humidity conditions, these materials creep faster than when
exposed to a constant relative humidity, even if the average relative humidity in the former case is
lower than the relative humidity at constant conditions. Different explanations for MSC have been
suggested, including effects of molecular mobility, physical ageing, inter-fibre mechanisms and
sorption-induced gradients of stress (Olsson and Salmén 2014). It has also been shown that single
wood fibres can experience MSC, although to a significantly lower degree than paper (Olsson and
Salmén 2014). Hence, both mechanisms within the fibre and in the fibre network contribute to MSC.
Whereas hemicellulose does not seem to have a decisive role in MSC, MSC increases with decreasing
microfibrillar angle of the S2 layer (S2 is the main fibril layer in the wood fibre). It should be noted
that MSC is not unique to wood-based materials. Apart from the mechano-sorptive effect, the
plasticising effect of water also leads to a tougher and more tear-resistant paper; however, the tensile
index decreases (Fellers 2009).
Many biopolymeric systems are polar and contain hydrogen bonds, and the hydrogen bonding
makes the materials stiff and strong but also brittle in dry conditions. The hydrogen bond network
prevents larger-scale chain motions. This also affects the rheological properties. The solution to this
has been to add a plasticiser, which facilitates the processing of the material and, importantly,
increases the material toughness. A trifunctional alcohol (glycerol) has been shown to be among
the most effective and also most used plasticisers for polysaccharides and proteins. It is miscible over
a large concentration range with these polymers and effectively breaks the rigid polymer–polymer
hydrogen bond network, in much the same way as water does (Özeren et al. 2020a,b). Water is an
even more ‘powerful’ plasticiser, but is not practical to use in these systems because of its low boiling
point.
366 6 Mechanical Properties

6.21 Summary

Polymers are viscoelastic materials, with often strong time-dependent mechanical properties. For
more rigid molecular systems, and at low temperature, the elastic component dominates, whereas at
higher temperature the viscous component dominates. Linear elastic fracture mechanics can be
applied to the former systems, whereas in the latter systems, elastic-plastic fracture mechanics
must be used to describe crack-dependent properties. Polymers with linear rigid segments surrounded
by very flexible bonds are tough at room temperature (refer to, e.g. polycarbonate), whereas polymers
with bulky side groups and, consequently, a less flexible chain are brittle (e.g. polystyrene). In the
regime of linear viscoelastic behaviour (low strain), it is possible to apply relatively simple models to
describe creep, stress relaxation and dynamic mechanical properties. A number of ‘internal’ (molar
mass, molar mass distribution, type of side group, branch content and branch distribution, crystallin-
ity, crosslink density) and external (time, temperature and environment) factors affect the overall
mechanical behaviour. Also important are the actual primary (C–C, C–O, etc.) and secondary
(dispersion, dipole, hydrogen bond) bond strengths and the entanglement density (dependent on,
e.g. molar mass). For a designer of a plastic product, it is important to measure the mechanical
properties with a method as close as possible to the real case. Creep rate measurements are adequate
to establish the critical properties for a plastic coat hanger, whereas the dynamic stiffness and
damping properties should be used for designing a car tire. Whereas rigid glassy polymers like
polystyrene show, especially at lower temperature, dominantly enthalpy-elastic properties, a molec-
ularly highly flexible crosslinked rubber shows entropy-driven mechanical properties. In the latter
case, a number of special mechanical features arise, for example, a material stiffness that increases
with increasing temperature.

6.22 Exercises

6.1. The stress state at a given point in a continuum is given as:

0 1
10 5 5
B C
σij ¼ @ 5 10 15 A Mpa ð6:236Þ
5 15 20

Calculate the size of the normal and shear stresses in a plane that goes through the point with
the normal vector:
0 1
1
1 B C
n ¼ pffiffiffiffiffi @ 3 A ð6:237Þ
35
5

6.2. A person is holding a leaning board. The angle is θ ¼ 45 , the board is 2 m long and 1 m wide and
the weight of the board is 30 kg. Assume that the centre of mass of the board is located at the
middle of the board. Determine the force needed to hold the board (arrow A) when the person is
holding the board at a distance of 1.5 m from where the board is in contact with the ground. What
is the force from the ground on the board (arrow B)? (Fig. 6.105)
6.22 Exercises 367

Fig. 6.105 A person

holding a leaning board

Fig. 6.106 Plane stress

condition

6.3. The stress state (in MPa) in a plane stress situation is given in (Fig. 6.106):
Calculate the two principal stresses and the angle from the above orientation of the square to
where the principal stress state occurs. Hint: The two principal stresses are on a circle (defined
by the points (12,5) and (7,-5)) along the x-axis (no shear stress).
6.4. Two rods are connected by means of an adhesive joint, the allowable shear stress (τcr) of which
is 10 MPa. Determine the permissible axial force (P) that the connection can carry. The adhesive
covers 40 cm axial length (L), and the outer diameter (D) of the inner rod is 10 cm.
6.5. Show that the expression λ1λ2λ3 ¼ 1 is valid at constant volume.
6.6. Derive an expression between volume strain (ΔV/V0) and principal strains expressed in
logarithmic strains. Show the exact solution for constant volume.
6.7. Calculate the density after a tensile strain of 4% of a material with a Poisson’s ratio of 0.45. The
density of the undeformed material is 1100 kg m3.
6.8. Three Maxwell elements are coupled in parallel. Calculate the resulting relaxation modulus as a
function of time at constant strain. The modulus for elements 1, 2 and 3 are E1 ¼ 1, E2 ¼ 2 and
E3 ¼ 1 GPa, and the respective stress relaxation times are τ1 ¼ 1, τ2 ¼ 10 and τ3 ¼ 100 s.
6.9. Derive a relationship for strain as a function of time after complete unloading of a Voigt–Kelvin
element.
6.10. Engine Inc. wants to have a polymer material as a replacement for steel in parts of a car engine.
For satisfactory use, it has been determined that the shear strain should not be higher than 0.5%
after 2 years of service at a shear stress of 0.2 MPa. The local temperature where the plastic will
368 6 Mechanical Properties

Fig. 6.107 Storage compliance as a function of shear rate

be situated is ca. 130  C. They found an interesting material with the following dynamic
mechanical data at 130  C (Fig. 6.107):
Determine if the material can be used in the application by calculating the shear strain after
2 years. Do the following:
– Calculate the retardation spectrum assuming linear viscoelasticity:

∂J 0 ðωÞ
L ðτ Þ ¼  ð6:238Þ
∂ ln ðωÞ

by using discrete differences, e.g. ΔJ0 (ω).

– Insert the retardation spectrum in the following equation:

ð
1
h  i
t
J ðtÞ ¼ J 0 þ jðτÞ 1  exp  dτ ð6:239Þ
τ
0

by knowing that L(τ) ¼ j(τ)τ and, therefore, j(τ) ¼ L(τ)/τ .

– Consider that the retardation strength is:
ð1
LðτÞ
∂τ ¼ J 1  J 0 ¼ J 0 0  J 1 0 ð6:240Þ
0 τ

and τ ¼ 2π/ω. Use a sum of terms as an approximation of the integral.

– Determine J(2 years) and multiply with the stress to obtain the strain.
6.22 Exercises 369

Fig. 6.108 An isochronous stress strain curve at 100 h and 20  C

6.11. A tensile bar of a polymer is under tensile load at 20  C. After 100 h, the strain is measured to be
1.3%. Estimate the modulus at 100 h and 1.3% strain. How large is the volume increase during
the actual loading? The isochronous stress–strain diagram of the polymer is given in
(Fig. 6.108). Assume a Poisson’s ratio of 0.34.
6.12. Link the terms (a) brittle, (b) crystalline, (c) entropy-elastic, (d) oriented and (e) highly
crosslinked to what they best describe: (1) rubber, (2) ceramic, (3) polyethylene fibre, (4) ther-
moset and (5) metal.
6.13. A polymer specimen is exposed to σ ¼ 2 MPa for the first 500 s, and it is exposed to an
additional σ ¼ 0.5 MPa during the following 1000 s. After 1000 s the stress is fully removed.
What is the tensile strain after 750 and 1200 s? Assume that it is possible to use Boltzmann
superposition principle. The creep compliance for the material is here described as:

DðtÞ ¼ 0:1 t0:5 ðGPaÞ1 ð6:241Þ

6.14. What is the creep compliance at 40  C for a polymer material that is described by the same
compliance as in exercise 6.13 at 30  C, and where can the Arrhenius equation be used with an
activation energy: ΔH ¼ 150 kJ mol1?
6.15. It takes 15 h for the stress of a strained polymer to decay by 100% at 15  C. What is the
relaxation time for the same stress decrease at 30  C? Tg for the polymer is 70  C. Use the
WLF equation to shift data.
6.16. Calculate a possible error in volume strain measurement caused by temperature variation.
Assume 1  C variation. Consider 1% strain, υ ¼ 0.4 and a linear thermal expansion coefficient:
αL ¼ (1/L0)(dL/dt) ¼ 10–4 ( C)1.
6.17. At what frequency (f50) does the maximum in damping (tan(δ)max) due to the glass transition of
a polymer material occur at 50  C if it occurs at 150 Hz at 20  C? The activation energy is
150 kJ mol-1. Is the material rubbery or glassy at 20  C from the beginning before the dynamic
load is applied? Is the material rubbery or glassy at 20  C and 500 Hz? At what frequency is the
loss in heat/energy highest at 20  C: 1, 150 or 500 Hz? What frequency (1, 150 or 500 Hz) is
closest to the glass transition at 20  C?
370 6 Mechanical Properties

Fig. 6.109 Creep fracture data

6.18. A polymer is creep tested at different tensile stresses, leading to a diagram below that shows the
relation between time to failure and stress (Fig. 6.109). The three data sets were obtained at
40, 60 and 80  C. Estimate the lifetime at 20  C when the stress applied is 5 MPa. Preferably,
make use of the equations given in the figure.
6.19. A strip with rectangular cross section of a rubber with isotropic properties is stressed in different
ways in the principal orthogonal directions (ignore any shear stresses and strains). The modulus
is 10 MPa and Poisson’s ratio can be taken as 0.5. Calculate the strain in the longitudinal
direction (33) due to a stress of 10 MPa in the same direction. Then, do the same when a stress of
the same size is also applied in one of the transverse principal directions. What is the strain in
the longitudinal direction when a third stress of the same size is applied in the third direction?
Make use of Eq. (6.89).
6.20. If the temperature and crosslink density are known, the short-term modulus/stiffness at a small
deformation of a rubber can be calculated/estimated (requires also the gas constant in the
calculation). (a) Determine the modulus of a rubber which has a density of 0.98 g cm-3 at
two different temperatures (20 and 40  C) for a molar mass between crosslinks that is
3000 g mol1. (b) Do the same for a lower crosslink density (molar mass between crosslinks
of 7000 g mol1).
6.21. The bulk modulus is the inverse of the compressibility (Eq. (6.77)). Calculate the bulk modulus
for a polymer with the volume–pressure data (Fig. 6.110) in the interval p ¼ 0–50 MPa.
6.22. Consider a polymer that has, depending on the plasticiser content, the nominal stress–strain
behaviour in (Fig. 6.111). The three curves correspond to (a) elastomeric, (b) ductile and
(c) brittle behaviour. Couple the three behaviours (a–c) with the curves 1–3. Estimate the
value of the initial modulus, yield stress and extensibility of curve II. How will these values
change if one uses an increased strain rate? Which of behaviours 2 and 3 correspond to the
toughest material?
6.23. A paper strip with a rectangular cross section is tensile tested, and the curve below is obtained at
50 mm/min strain (Fig. 6.112). The grammage of the paper is 60 g m2. The sample thickness is
300 μm and its width is 1.5 cm. Calculate the ‘normal’ fracture stress and the tensile index
(Eq. (6.234)).
6.24. Figure 6.113 shows three curves of the temperature dependence of the shear storage modulus for
a polymer with three different degrees of crosslinking. Which symbols represent the curves with
the lowest and highest degree of crosslinking. Essentially, a similar type of modulus behaviour
6.22 Exercises 371

Fig. 6.110 Isothermal volume–pressure data

Fig. 6.111 Stress–strain curves of a polymer having different amounts of plasticiser

can be obtained with a polymer of different crystallinity. Which curves should then represent the
polymer with the lowest and highest degrees of crystallinity?
6.25. Derive a relationship between area change and Poisson’s ratio in a two-dimensional case. What
is the value of Poisson’s ratio when there is no area change during deformation?
6.26. Estimate the yield stress for two polyethylenes with a crystal thickness of respectively 12.5 and
20 nm (use Fig. 6.77).
6.27. Consider a wide, thin plate of a brittle polymer with a central crack of length 2 cm. Assume that
plastic zone at the crack is negligible and linear elastic fracture mechanics hold. The critical
fracture energy (Gc) is 500 J/m2. Will the crack propagate at a stress of 6 MPa normal to the
crack (use the stress intensity factor)? The elastic modulus of the polymer is 4 GPa.
372 6 Mechanical Properties

Fig. 6.112 Force-strain curve for a paper strip

Fig. 6.113 Shear storage

modulus as a function of
temperature

6.28. Consider the case in exercise 6.27, but now the plate is thick (plane strain conditions; ignore
plane stress at the surface). The critical fracture energy (GIC) is 200 J/m2. Poisson’s ratio is 0.35.
Will the crack propagate?
6.29. Determine the crack growth rate per cycle of a crack which is initially 2 cm long and placed in
the middle of a wide, thin sheet. The stress varies between 1 and 3 MPa. The Paris law
(Eq. 6.233) has the following constants for this polymer: Af ¼ 0.125, m ¼ 4.

References

Agnelli, S., & Horsfall, I. (2013). Engineering Fracture Mechanics, 101, 59.
Aguikar-Vega, M. (2013). Chapter 21: Structure and mechanical properties of polymers. In E. Saldı́var-Guerra &
E. Vivaldo-Lima (Eds.), Handbook of polymer synthesis, characterization, and processing. New York: Wiley.
Aldhufairi, H. S., & Olatunbosun, O. A. (2018). Proceedings of the Institution of Mechanical Engineers, Part D:
Journal of Automobile Engineering, 232, 1865.
Allison, S. W., & Andrews, R. D. (1967). Journal of Applied Physics, 38, 4164.
Andean, L., Castellani, L., Castiglione, A., Mendogni, A., Rink, M., & Sacchetti, F. (2013). Engineering Fracture
Mechanics, 101, 33.
References 373

Arbeiter, F., Pinter, G., Lang, R. W., & Frank, A. (2017). Fracture mechanics methods to assess the lifetime of
thermoplastic pipes. In W. Grellmann & B. Langer (Eds.), Deformation and fracture behaviour of polymer
materials. Berlin: Springer.
Arridge, R. G. C. (1975). Mechanics of polymers. Fair Lawn: Oxford University Press.
Awaja, F., Zhang, S., Tripathi, M., Nikiforov, A., & Pugno, N. (2016). Progress in Materials Science, 83, 536.
Bálány, T., Czigány, T., & Karger-Kocsis, J. (2010). Progress in Polymer Science, 35, 1257.
Banks, H. T., Hu, S., & Kenz, Z. R. (2011). Advances in Applied Mathematics and Mechanics, 3, 1.
Bauwens-Crowlet, C., Bauwens, J.-C., & Homès, G. (1969). Journal of Polymer Science, Polymer Physics Edition, 7,
735.
Bier€ogel, C., & Grellmann. (2014a). Fatigue loading of plastics – Introduction. In K.-F. Arndt & M. D. Lechner (Eds.),
Polymer solids and polymer melts – Mechanical and thermomechanical properties of polymers (Vol. 6A3). Berlin:
Springer.
Bier€ogel, C., & Grellmann. (2014b). Tensile fatigue loading of thermoplastics – Applications. In K.-F. Arndt & M. D.
Lechner (Eds.), Polymer solids and polymer melts – Mechanical and thermomechanical properties of polymers
(Vol. 6A3). Berlin: Springer.
Bier€ogel, C., & Grellmann. (2014c). Torsional fatigue strength – Application. In K.-F. Arndt & M. D. Lechner (Eds.),
Polymer solids and polymer melts – Mechanical and thermomechanical properties of polymers (Vol. 6A3). Berlin:
Springer.
Birley, A. W., Haworth, B., & Batchelor, J. (1992). Physics of plastics, processing, properties and materials
engineering. Munich: Carl Hanser.
Bowman, K. (2004). Mechanical behaviour of materials. Hoboken: Wiley.
Boyd, R. H. (1984). Macromolecules, 17, 903.
Bradley, W., Cantwell, W. J., & Kausch, H.-H. (1998). Mechanics of Time-Dependent Materials, 1, 248.
Broitman, E. (2017). Tribology Letters, 65, 23.
Brooks, N. W., Ghazali, M., Duckett, R. A., Unwin, A. P., & Ward, I. M. (1999). Polymer, 40, 821.
Cagiao, M. E., Baltá-Calleja, F. J., Vanderdonckt, C., & Zachmann, H. G. (1993). Polymer, 34, 2024.
Case, J., & Chilver, A. H. (1959). Strength of materials, an introduction to the analysis of stress and strain. London:
Edward Arnold Ltd.
Chen, K., & Schweizer, K. S. (2011). Macromolecules, 44, 3988.
Chèriére, J. M., Bélec, L., & Gacougnolle, J. L. (1997). Polymer Engineering and Science, 37, 1664.
Corradini, P., Rosa, C. D., Guerra, G., & Petraccone, V. (1987). Macromolecules, 20, 3043.
Darras, O., & Séguéla, R. (1993). Journal of Polymer Science, Polymer Physics Edition, 31, 759.
Das, O., & Bhattacharyya, D. (2017). Development of polymeric biocomposites: Particulate incorporation in interphase
generation and evaluation by Nanoindentation. In A. N. Netravali & K. L. Mittal (Eds.), Interface/interphase in
polymer nanocomposites. Salem: Scrivener Publishing LLC.
Deblieck, R. A. C., van Beek, D. J. M., Remerie, K., & Ward, I. M. (2011). Polymer, 52, 2979.
Degennes, P.-G. (1979). Scaling concepts in polymer physics. Ithaca: Cornell University Press.
Devries, K. L., & Nuismer, R. J. (1985). ACS symposium series, vol. 285, Applied polymer science, Ch. 13, p. 277.
Dinwoodie, J. M. (2000). Timber, its nature and behaviour (2nd ed.). London: E. & F. N. Spon.
Duan, Y., Saigal, A., Greif, R., & Zimmermann, M. A. (2002). Polymer Engineering and Science, 42, 395.
Eremeeva, A. S. (1965). Mekhanika Polimerov, 1, 21.
Fan, J., Fan, X., & Chen, A. (2017). Chapter 8: Dynamic mechanical behaviour of polymer materials. In F. Yilmaz
(Ed.), Aspects of polyurethanes (p. 193). Zagreb: InntechOpen.
Fellers, C. (2009). Pulp and paper chemistry and technology. In M. Ek, G. Gellerstedt, & G. Henriksson (Eds.), Pulp
and paper physics and technology (Vol. 4). Berlin: De Gruyter.
Findley, W. N., & Khosla, G. (1955). Journal of Applied Physics, 26, 821.
Foreman, J. P., Porter, D., Behzadi, S., Curtis, P. T., & Jones, F. R. (2010). Composites: Part A, 41, 1072.
Francois, D., Pineau, A., & Zaoui, A. (2013). Fracture mechanics. In Mechanical behaviour of materials volume II:
Fracture mechanics and damage (p. 7). Berlin: Springer.
Gaucher-Miri, V., Elkoun, S., & Séguéla, R. (1997). Polymer Engineering and Science, 37, 1672.
Gedde, U. W., & Hedenqvist, M. S. (2019a). Rubber Elasticity. In Fundamental polymer science (chap. 3). Cham:
Springer Nature Switzerland.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Chain orientation. In Fundamental polymer science (chap. 9). Cham:
Springer Nature Switzerland.
Gerets, B., Wenzel, M., Engelsing, K., & Bastian, M. (2017). Slow crack growth of polyethylene – Accelerated tests
methods. In W. Grellmann & B. Langer (Eds.), Deformation and fracture behaviour of polymer materials. Berlin:
Springer.
Golovin, K., Kobaku, S. P. R., Lee, D. H., DiLoreto, E. T., Mabry, J. M., & Tuteja, A. (2016). Science Advances, 2,
e1501496.
374 6 Mechanical Properties

Grellman, W. (2001). New developments in toughness evaluation of polymers and compounds by fracture mechanics.
In W. Grellman & S. Seidler (Eds.), Deformation and fracture behaviour of polymers (p. 3). Berlin: Springer.
Grellman, W., & Lach, R. (2001). Toughness and relaxation behaviour PMMA, PS and PC. In W. Grellman &
S. Seidler (Eds.), Deformation and fracture behaviour of polymers (p. 193). Berlin: Springer.
Hedenqvist, M. S., Bharadwaj, R., & Boyd, R. H. (1998). Macromolecules, 31, 1556.
Hedenqvist, M. S., Yousefi, H., Malmstr€om, E., Johansson, M., Hult, A., Gedde, U. W., Trollsås, M., & Hedrick, J. L.
(2000). Polymer, 41, 1827.
Janssen, R. P. M., de Kanter, D., Govaert, L. E., & Meijer, H. E. H. (2008). Macromolecules, 41, 2520.
Karger-Kocsis, J. (1996). Polymer Engineering and Science, 36, 203.
Kausch, H.-H. (1987). Polymer fracture. Berlin: Springer.
Kausch, H.-H., & Dettenmaier, M. (1982). Colloid and Polymer Science, 260, 120.
Kausch, H-H, Grein, C, Béguelin P, & Gensler, R. (2001). International conference on Fracture ICF10, ICF 1001–007.
Keating, M. Y., Sauer, B. B., & Flexman, E. A. (1997). Journal of Macromolecular Science-Physics, 36, 717.
Klein, C. A., & Cardinale, G. F. (1993). Diamond and Related Materials, 2, 918.
Klüppel, M. (2014). Wear and abrasion of tires. In S. Kobayashi & K. Müllen (Eds.), Encyclopedia of polymeric
nanomaterials. Berlin: Springer.
Knauss, W. G. (1989). International series on the strength and fracture of materials and structures (Vol. 4, p. 2683).
New York: Pergamon Press.
Knauss, W. G. (2015). International Journal of Fracture, 196, 99.
Kohlrausch, R. (1847). Annalen der Physik und Chemie, 12, 393.
Krieg, M., Fl€aschner, G., Alsteens, D., Gaub, B. M., Roos, W. H., Wuite, G. J. L., Gaub, H. E., Gerber, C., Dufrêne,
Y. F., & Müller, D. J. (2019). Nature Reviews Physics, 1, 2019.
Lach, R., & Grellman, W. (2017). Modern aspects of fracture mechanics in industrial application of polymers. In
W. Grellmann & B. Langer (Eds.), Deformation and fracture behaviour of polymer materials. Berlin: Springer.
Lakes, R. (1987). Science, 235, 1038.
Lakes, R. S., & Witt, R. (2002). International Journal of Mechanical Engineering Education, 30, 50.
Leis, B. N. (1989). 11th A. G. A. Plastic Fuel Gas Pipe Symposium, San Francisco, U.S.A.
Liu, C.-Y., He, J., Keunings, R., & Bailly, C. (2006). Macromolecules, 39, 8867.
Lu, H., Zhang, X., & Knauss, W. G. (1997). Polymer Science and Engineering, 37, 1053.
Lustiger, A. (1986). In W. Brostow & R. D. Corneliussen (Eds.), Failure of plastics (p. 305). Munich: Carl Hanser.
McCrum, N. G., Buckley, C. P., & Bucknall, C. B. (1997). Principles of polymer engineering (2nd ed.). Oxford: Oxford
Science Publishers.
Mercier, J. P., & Groeninckx, G. (1969). Rheologica Acta, 8, 504.
Michler, G. H. (2001). Crazing in amorphous polymers – Formation of fibrillated crazes near the glass transition
temperature. In W. Grellman & S. Seidler (Eds.), Deformation and fracture behaviour of polymers (p. 193). Berlin:
Springer.
Mohammady, S. Z., Mansour, A. A., Knoll, K., & Stoll, B. (2002). Polymer, 43, 2467.
Molan, G. E., & Keskkula, H. (1968). Applied Polymer Symposia, 7, 35.
Moyassari, A., Gkourmpis, T., Hedenqvist, M. S., & Gedde, U. W. (2018). Polymer, 161, 139.
Moyassari, A., Gkourmpis, T., Hedenqvist, M. S., & Gedde, U. W. (2019). Macromolecules, 52, 807.
Muliana, A. H., & Haj-Ali, R. M. (2004). Mechanics of Materials, 36, 1087.
Mullikan, A. D., & Boyce, M. C. (2006). International Journal of Solids and Structures, 43, 1331.
Mura, A., Ricci, A., & Canavese, G. (2018). Materials, 11, 1818.
Naebe, M., Abolhasani, M. M., Khayyam, H., Amini, A., & Fox, B. (2016). Polymer Reviews, 56, 31.
Nielsen, L. E., & Landel, R. F. (1994). Mechanical properties of polymers and composites. New York: Marcel Dekker.
Nilsson, F., Lan, X., Gkourmpis, T., Hedenqvist, M. S., & Gedde, U. W. (2012). Polymer, 53, 3594.
Olabarrieta, I., G€allstedt, M., Sarasua, J.-R., Johansson, E., & Hedenqvist, M. S. (2006). Biomacromolecules, 7, 1657.
Olsson, A.-M., & Salmén, L. (2014). Wood Science and Technology, 48, 569.
Özeren, H. D., Nilsson, F., Olsson, R. T., & Hedenqvist, M. S. (2020a). Materials Design, 187, 108387.
Özeren, H. D., Guivier, M., Olsson, R. T., Nilsson, F., & Hedenqvist, M. S. (2020b). ACS applied polymer materials., in
press.
Page, D. H. (1969). Tappi Journal, 52, 674.
Park, S.-H., Park, C.-M., Kim, J.-H., & Kim, T. (2015). Advances in civil, environmental, and materials research
(ACEM 15), Incheon, Korea, August 25–29.
Pennings, J. P., Pras, H. E., & Pennings, A. J. (1994). Colloid and Polymer Science, 272, 664.
Petermann, J., & Ebener, H. (1999). Journal of Macromolecular Science-Physics Edition, B38, 837.
Pinto, M. B. (2007). European Journal of Physics, 28, 171.
Plackett, D., Anturi, H., Hedenqvist, M., Ankerfors, M., Lindstr€ om, T., & Siró, I. (2010). Journal of Applied Polymer
Science, 117, 3601.
References 375

Plati, E., & Williams, J. G. (1975). Polymer Engineering and Science, 15, 470.
Plazek, D. J. (1960). Journal of Colloid Science, 15, 50.
Porter, D., & Gould, P. J. (2009). International Journal of Solids and Structures, 46, 1981.
Qausar, M. (1989). Pure and Applied Geophysics, 131, 703.
Qiu, W., & Kang, Y.-L. (2014). Chinese Science Bulletin, 59, 2811.
Ramsteiner, F. (1983). Kunststoffe, 73, 148.
Ramsteiner, F., Schuster, W., & Forster, S. (2001). Concepts of fracture mechanics for polymers. In W. Grellman &
S. Seidler (Eds.), Deformation and fracture behaviour of polymers (p. 3). Berlin: Springer.
Redhead, A., Frank, A., & Pinter, G. (2013). Engineering Fracture Mechanics, 101, 2.
Roa, J. J., Oncins, G., Diaz, J., Sanz, F., & Segarra, M. (2011). Recent Patents on Nanotechnology, 5, 27.
Robeson, L. M. (2013). Polymer Engineering and Science, 53, 453.
Rodgers, B., & Waddell, W. (2013). Chapter 9: The science of rubber compounding. In J. E. Mark, B. Erman, & C. M.
Roland (Eds.), The science and technology of Rubber (4th ed.). Amsterdam: Elsevier.
R€osler, J., B€aker, M., & Harders, H. (2007). Mechanical behaviour of polymers. In Mechanical behaviour of
engineering materials. Berlin: Springer.
Saada, A. S. (1993). Elasticity theory and applications (3rd ed.). Malabar: Krieger Publishing Company.
Sadler, D. M., & Barhan, P. J. (1990). Polymer, 31, 46.
Salazar, A., Rodrı́guez, J., & Martı́nez, A. B. (2013). Engineering Fracture Mechanics, 101, 10.
Schapery, R. A. (1969). Polymer Engineering and Science, 9, 295.
Schapery, R. A. (1997). Mechanics of Time-Dependent Materials, 1, 209.
Schmieder, K., & Wolf, K. (1952). Kolloid Zeitschrift, 127, 65.
Schrauwen, B. A. G., Janssen, R. P. M., Govaert, L. E., & Meijer, H. E. H. (2004). Macromolecules, 37, 6069.
Schwarzl, F. R. (1969). Rheologica Acta, 8, 6.
Schwarzl, F. R. (1970). Rheologica Acta, 9, 382.
Schwarzl, F. R. (1971). Rheologica Acta, 10, 165.
Schwarzl, F. R. (1975). Rheologica Acta, 14, 581.
Schwarzl, F. R., & Staverman, A. J. (1952). Journal of Applied Physics, 23, 838.
Schwarzl, F. R., & Struik, L. C. E. (1968). Advances in Molecular Relaxation Processes, 1, 201.
Sharma, P., Chauhan, U., Kumar, S., & Sharma, K. (2018). International Journal of Applied Engineering Research, 13,
363.
Shur, Y. J., & Rånby, B. (1975). Journal of Applied Polymer Science, 19, 1337.
Siviour, C. R., & Jordan, J. L. (2016). Journal of Dynamic Behaviour of Materials, 2, 15.
Spathis, G., & Kontou, E. (2012). Composites Science and Technology, 72, 959–964.
Stachurski, Z. H. (1997). Progress in Polymer Science, 22, 407.
Stribeck, N. (1993). Colloid and Polymer Science, 271, 1007.
Takahashi, M., Chen, M. C., Taylor, R. B., & Tobolsky, A. V. (1964). Journal of Applied Polymer Science, 8, 1549.
Takayanagi, M. (1983). Pure and Applied Chemistry, 55, 819.
Takayanagi, M., & Goto, K. (1985). Polymer Bulletin, 13, 35.
Takemori, M. T. (1984). Annual Reviews on Materials Science, 14, 171.
Talamini, B., Mao, Y., & Anand, L. (2017). Journal of the Mechanics and Physics of Solids, 111, 434.
Tr€ankner, T. (1990). Studsvik report, Studsvik, Sweden.
Tr€ankner, T., Hedenqvist, M. S., & Gedde, U. W. (1994). Polymer Engineering and Science, 34, 1581.
Tr€ankner, T., Hedenqvist, M., & Gedde, U. W. (1997). Polymer Engineering and Science, 37, 346.
Van der Akker, J. A., Lathrop, A. L., Voelker, M. H., & Dearth, L. R. (1958). Tappi Journal, 41, 416.
Van Holde, K. (1957). Journal of Polymer Science, 24, 417.
Van Krevelen, D. W., & Te Nijeuhuis (2009). Properties of polymers (4th ed.). Amsterdam: Elsevier.
Viebke, J., Elble, E., Ifwarson, M., & Gedde, U. W. (1994). Polymer Engineering and Science, 34, 1354.
Vincent, P. I. (1961). Plastics, 26, 141.
Vincent, P. I. (1971). Impact test and service performance of thermoplastics. London: Plastics Institute.
Visser, H. A., Caimmi, F., & Pavan, A. (2013). Engineering Fracture Mechanics, 101, 67.
Vogt, B. D. (2018). Journal of Polymer Science, Part B: Polymer Physics, 56, 9.
Wagner, M. H., & Narimissa, E. (2018). Journal of Rheology, 62, 221.
Wang, J., Xu, Y., Zhang, W., & Ren, X. (2019). Polymers, 11, 654.
Ward, I. M. (1985). Mechanical properties of solid polymers (2nd ed.). New York: Wiley.
Ward, A. L., Lu, X., Huang, Y.-L., & Brown, N. (1991). Polymer, 32, 2172.
Weng, P., Tang, Z., & Guo, B. (2020). Polymer, 190, 122244.
Wetton, R. E., Foster, G., & Corish, P. J. (1991). Polymer Testing, 10, 175.
Williams, J. G. (1978). Applications of linear fracture mechanics in failure. in Polymers, Advances in polymer science,
27, ed. Cantow, H.-J., p. 67, Berlin: Springer.
376 6 Mechanical Properties

Williams, G., & Watts, D. C. (1970). Transactions of the Faraday Society, 66, 80.
Winkler, A., & Kloosterman, G. (2015). Frattura ed Integritá Strutturale, 33, 262.
Wolf, K. (1951). Kunststoffe, 41, 89.
Wu, S. (1992). Polymer International, 247, 229.
Xu, H., Tang, Y., Liu, Z., Cai, Y., & Wang, Y. (2017a). IOP Conference Series: Materials Science and Engineering,
231, 012123.
Xu, T., Jia, Z., Wu, L., Chen, Y., Luo, Y., Jia, D., & Peng, Z. (2017b). Applied Surface Science, 423, 43.
Young, R. (1974). The Philosophical Magazine: A Journal of Theoretical Experimental and Applied Physics, 30, 85.
Young, R. (1988). Journal of Materials Forum, 11, 210.
Chapter 7
Transport Properties of Polymers

7.1 Introduction

Mass transport occurs just about everywhere. The driving force is nature striving for equalisation of
the chemical potential. This transport plays important roles in a broad range of processes, including
transport of species through cell membranes, migration from packaging, diffusion-limited oxidation,
transport of water in plants and membrane separation. Polymers are generally poor electric and heat
conductors but usually good ‘solute molecule conductors’. In contrast to metals, polymers normally
consist of a relatively ‘open’ structure with a lower degree of atomic packing and high free volume,
especially in the amorphous phase. The ‘loose’ connectivity of the polymer molecules leads to good
electric and heat-insulating properties. At the same time, the large amount of free volume favours
diffusion and solubility of gas, vapour or liquid molecules within the polymer matrix. In this chapter,
the focus is on mass transport in mainly homogeneous/single component (amorphous/semicrystal-
line) solid polymers but also includes a small section on thermal transport in these systems. Mass
transport in heterogeneous systems such as polymer blends, including interpenetrating networks and
composites, is extensively described in the book Transport Properties of Polymeric Membranes by
Thomas et al. (2018). It is also a good compilation of membrane processes. Mass transfer in porous
systems, involving, for example, Knudsen and capillary flow, is not dealt with here, but is covered in,
for example, the review by Tartakovsky and Dentz (2019) and the book by K€argel, Ruhrven and
Theodorou (2012). For foams, refer to, for example, the papers by Kimball and Frisch (1991) and
Pilon, Fedorov and Viskanta (2000). For ion and charge transport, the reader is referred to Aziz et al.
(2018), Weber et al. (2014), Geise, Paul and Freeman (2014) and Jaiswal and Menon (2006), and for
diffusion coupled with reaction, some of the basic treatments are presented in Crank (1986).
This chapter starts with a description of the typical characteristic trajectory, the random walk, of
solute molecules and the effects of temperature and solute size on diffusion. In the following section,
solute solubility is discussed with a focus on how to determine solubility from equation of state
models. Fick’s laws of diffusion are then derived for different standard geometries, and the ways to
solve them analytically and numerically are illustrated. In a large section of the chapter, the
characteristics of mass transfer in the different types of polymer are described, including amorphous
and semicrystalline, rubbery and glassy polymers. Techniques for measuring mass transport are
described, and two common cases where mass transport is central are briefly presented (barrier
and membrane applications). Complex behaviour, such as anomalous behaviour, concentration-
dependent transport properties, swelling and build-up of mechanical stresses, and clustering are
also dealt with. The chapter ends with as, mentioned above, a short section on thermal transport
properties.

# Springer Nature Switzerland AG 2021 377

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_7
378 7 Transport Properties of Polymers

7.2 Diffusion

7.2.1 Basics of the Random Walk

Interstitial and vacancy/substitutional diffusion are the two most common diffusion mechanisms in
solid crystalline/close-packed materials (Welty et al. 2001). In interstitial diffusion, the solute
molecules jump from one interstitial hole to the next, and the process may involve a dilation or
distortion of the lattice (Fig. 7.1a). In vacancy diffusion, the solute molecule moves whenever a
vacancy occurs next to the atom (Fig. 7.1b).
Solute diffusion in amorphous polymers or in the amorphous part of semicrystalline polymers is
normally neither a pure interstitial process, nor a pure vacancy process, but rather a combination of
the two modes. Vacancies, i.e. free volume holes, are not as clearly defined in polymers as in metals
because of the low degree of atomic close-packing in the amorphous phase of the former. Diffusion of
solutes in polymers is a thermally activated process (Vieth 1991). Atomistic simulations provide
detailed understanding of the diffusion processes of solute molecules in, for example, polymer
matrices. By introducing solute molecules into a polymer matrix, it is possible to analyse the
dynamics of the solute–polymer system (Fig. 7.2). Note the large free volume in the system when
the atoms are ‘frozen’. A large part of the observed free volume in Fig. 7.2 is, however, occupied by
the vibrational space of the atoms.
By recording the distance travelled by the solute molecule (r) as a function of time (t), it is possible
to obtain the diffusivity (D) with the Einstein relation (Eq. (7.1)) (Crank 1986):

2
r ¼ 6Dt ð7:1Þ

which is the mean square displacement (MSD) and is derived below (cf. Chap. 5). It is based on the
assumption that the system consists of atoms and molecules, simply referred to here as particles,
which move according to Brownian motion. Brownian motion implies that the particles are subjected
to the force (Bian, Kim and Karniadakis 2016):

FðtÞ ¼ hFðtÞi þ FðtÞ f ð7:2Þ

where the first term on the right side is due to friction that acts so as to reduce the particle velocity
hv(t)i:

a) b)
Fig. 7.1 Interstitial (a) and vacancy/substitutional (b) diffusion in close-packed systems
7.2 Diffusion 379

Fig. 7.2 Atomistic system

of amorphous starch
containing 10% glycerol
molecules (dark molecules)
close to the true
(experimental) density at
300 K. Courtesy of H. D.
Özeren, Fibre and Polymer
Technology, KTH Royal
Institute of Technology.
Note that the sizes of the
atoms correspond to their
van der Waals volumes

hFðtÞi ¼ ξ hvðtÞi ð7:3Þ

where ξ is the friction constant and F(t)f is a fluctuating force which arises from the constant collisions
between particles and is the force that actually keeps the particles in continuous motion. Eqs. (7.2)
and (7.3) can be combined with Newton’s first law to yield the Langevin equation (Bian, Kim and
Karniadakis 2016):
dvðtÞ
m ¼ ξ hvðtÞi þ FðtÞ f ð7:4Þ
dt

The next step is to obtain an expression for the mean square displacement during time t, and, for
simplicity, we consider it along the x-direction: hx(t)2i. Eq. (7.4) is multiplied with x on both sides and
rewritten for one dimension:
"    2 #
d2 x d dx dx dx
mx 2 ¼ m x  ¼ ξ x þ x Fx ðtÞ f ð7:5Þ
dt dt dt dt dt

where Fx(t)f is the x-component of the ‘collision’ force. Averaging the equation above leads to:
"   * 2 +#   D E
d dx dx dx
m x  ¼ ξ x þ xFx ðtÞ f ð7:6Þ
dt dt dt dt

The collision force is independent of x and, therefore, the last term in Eq. (7.6) vanishes. The
relationship may be further simplified to:
    
d dx dx
m x ¼ kT  ξ x ð7:7Þ
dt dt dt

and
380 7 Transport Properties of Polymers

    
dx kT m d dx
x ¼  x ð7:8Þ
dt ξ ξ dt dt

by considering that the particle kinetic energy is:

*  +
2
1 dx 1
m ¼ kT ð7:9Þ
2 dt 2

The solution to the differential Eqs. (7.7) and (7.8) is of the form:
 
dx kT
x ¼ þ C exp ðt=τB Þ ð7:10Þ
dt ξ

where C is a constant and τB ¼ m/ξ is the relaxation time of the Brownian motion. With the initial
condition x(0) ¼ 0, C ¼ kT/ξ, which yields:
 

dx 1 d x2 kT
x ¼ ¼ ð1  exp ðt=τB ÞÞ ð7:11Þ
dt 2 dt ξ

Further rearrangement and time integration with respect to x yields:

2  2kT
x ¼ ðt þ τB exp ðt=τB ÞÞ þ C1 ð7:12Þ
ξ

and with the initial condition hx2(0)i ¼ 0, C1 ¼ 2kTτB/ξ and

2  2kT
x ¼ ½t  τB ð1  exp ðt=τB ÞÞ ð7:13Þ
ξ

At short times, in the ballistic regime t < <τB, and using the Taylor series
2
exp ðyÞ=1  y þ y2 þ Oðy3 Þ, the average square distance travelled by the solute along x is:

2  kT 2
x ¼ t ð7:14Þ
m

At long times, in the random walk regime (t >> τB), the following expression is obtained:

2  2kT
x ¼ t ¼ 2D t ð7:15Þ
ξ

where D ¼ kT/ξ, i.e. the diffusivity D is inversely proportional to the friction coefficient. This is the
diffusivity determined in absence of a gradient in chemical potential/concentration and without any
interfering bulk flow. In three dimensions, Eq. (7.15) turns into Eq. (7.1).
To obtain sufficient statistics, i.e. a straight line with slope 6D, new ‘origins’ can be taken along the
solute molecule trajectory or, alternatively, several molecules are inserted in the polymer matrix and
monitored during a shorter time. These are techniques to achieve the t ! 1 condition without having
to sample to ‘infinity’. Figure 7.3 shows the mean square displacement of glycerol molecules in a
glycerol environment, to illustrate the random walk in a ‘mobile’ system (small relaxation time of
Brownian motion, t> > τB occurs early) where the slope is described by Eq. (7.1). Note the ballistic
regime (an almost immediate increase) at the very beginning of the curve.
7.2 Diffusion 381

Fig. 7.3 The mean square displacement (MSD) in three dimensions of glycerol molecules in a liquid glycerol system at
300 K (400 molecules). (Courtesy of H. D. Özeren, Fibre and Polymer Technology, KTH Royal Institute of
Technology)

1
2

Fig. 7.4 In order to obtain sufficient statistics in using Eq. 7.16, new velocity vector origins are taken along the velocity
trajectory. This means that the vector multiplications of v’s in Eq. (7.16) are taken pairwise along the velocity trajectory
at time 1 (1 in the figure) and between every second velocity vector at time 2, and so on

7.2.2 The Velocity Autocorrelation Function

An alternative approach to calculate the diffusivity from a random walk is through the velocity
autocorrelation function:
ZðtÞ ¼ hvðtÞ vð0Þi ð7:16Þ

which is the projection of the particle velocity at time t (v(t)) onto the initial velocity (v(0)) averaged
over all initial conditions (Chandrasekhar 1943). The technique is illustrated in Fig. 7.4.
Figure 7.5 shows the velocity autocorrelation function as a function of time for glycerol molecules
in an all-glycerol environment, the same system as in Fig. 7.3. A typical feature is the loss in
correlation as a function of time, and for times much longer than any molecular relaxation process
in the matrix, the velocity correlation is completely decorrelated and Z approaches zero. Negative
382 7 Transport Properties of Polymers

Z values imply that the moving particle changes its diffusion direction, resulting in velocity vectors
which are anti-parallel.
The relationship between the velocity autocorrelation function and the diffusivity is now derived.
The solution of Eq. (7.4) can be written as (Bian, Kim and Karniadakis 2016):
ð
1 t
vðtÞ ¼ vð0Þ exp ðt=τB Þ þ dτ exp ððt  τÞ=τB Þ FðτÞ f ð7:17Þ
m 0

and the mean square velocity is:

D E D E
vðtÞ2 ¼ vð0Þ2 exp ð2t=τB Þþ
ðt D E
2
exp ðt=τB Þ dτ exp ððt  τÞ=τB Þ vð0Þ FðτÞ f þ ð7:18Þ
m 0
ð ðt
1 t
dτ exp ð ð t  τ Þ=τ B Þ dτ0 exp ððt  τ0 Þ=τB Þ Γδðτ  τ0 Þ
m2 0 0

The middle term disappears since hv(0) F(τ)fi ¼ 0. The fluctuating force follows an independent
Gaussian white noise signal (with strength Γ). The third term is:
Γ
ð1  exp ð2t=τB ÞÞ ð7:19Þ
2ξm

which is obtained knowing that:

D E
Γδðτ  τ0 Þ ¼ FðτÞ f Fðτ0 Þ
f
ð7:20Þ

Hence, Eq. (7.18) can now be expressed as:

D E D E
Γ
vðtÞ2 ¼ vð0Þ2 exp ð2t=τB Þ þ ð1  exp ð2t=τB ÞÞ ð7:21Þ
2ξm

At long times, we have:

D E
kT
vðtÞ2 ¼ ð7:22Þ
m

and by combining Eqs. (7.21) and (7.22), the noise strength becomes:
Γ ¼ 2ξkT ð7:23Þ

Hence:
D E D E
kT
vðtÞ2 ¼ vð0Þ2 exp ð2t=τB Þ þ ð1  exp ð2t=τB ÞÞ ð7:24Þ
m

By multiplying both sides of Eq. (7.17) with v(0) and taking the average, the following equation is
obtained:
D E
kT
hvðtÞvð0Þi ¼ vð0Þ2 exp ðt=τB Þ ¼ exp ðt=τB Þ ¼ ZðtÞ ð7:25Þ
m

By integrating Eq. (7.25) over time, the diffusivity D is obtained:

7.2 Diffusion 383

Fig. 7.5 Velocity autocorrelation function for the same glycerol system as in Fig. 7.3. (Courtesy of H. D. Özeren, Fibre
and Polymer Technology, KTH Royal Institute of Technology)

ð1 ð1
kT kT
ZðtÞdt ¼ exp ðt=τB Þdt ¼ ¼D ð7:26Þ
0 m 0 ξ

Hence, by time integration of the whole velocity autocorrelation function, the diffusivity is
obtained, which is another approach compared to using the mean square displacement (Eq. 7.1). It
should be noted that Fig. 7.5 shows a much smaller time period than in Fig. 7.3.

7.2.3 Diffusion; Temperature and Solute Size

When studying methane diffusion in polystyrene by molecular dynamics simulations, Han and Boyd
(1996) were able to distinguish between two different solute jump mechanisms in the polymer matrix
(Fig. 7.6). At high temperature, well above the glass transition temperature, the diffusion process is
liquid-like and characterised by a more or less continuous solute flow, and discrete jumps are less or
absent. The solute molecule moves along with the highly mobile matrix chain segments. At a lower
temperature, but still above Tg, the diffusion process changes. The solute is trapped between polymer
chain segments and oscillates in the same place. The solute makes occasional jumps into adjacent
positions. This diffusion process is referred to as confined or cage-like diffusion and is typical for
small solute molecules in polymers at ambient or lower temperatures. The fact that the change from
confined to liquid-like diffusion, when increasing the temperature, does not occur at the onset of
larger chain segment displacements (associated with the glass transition) leads to the conclusion that
small solute diffusion is a local event and separated from larger-scale dynamics. Confined diffusion is
observed early in the mean square displacement–time curve as a non-linear part with decreasing slope
(Fig. 7.7). At longer times, the random walk dominates and the curve becomes linear.
The jumps in the liquid-like region are less distinguishable, and there is a broad range of jump
lengths, in contrast to the confined diffusion region where jumps are more clearly distinguishable
(Fig. 7.8). Molecular dynamics reveal that small molecule jump lengths vary between different
polymers. A methane molecule jumps on average more than twice as far in atactic polypropylene
384 7 Transport Properties of Polymers

Fig. 7.6 High-temperature

liquid-like small molecule
diffusion (upper left) and
low-temperature cage-like
small molecule diffusion
(lower right)

Fig. 7.7 Schematic curves

showing mean square
displacement with (thin
line) and without (thick
line) confined diffusion.
The two curves are
arbitrarily drawn and
should not be compared

Fig. 7.8 Simulated

distribution of jump lengths
for a methane molecule in
atactic polystyrene at
various temperatures
(in  C). (Drawn after data
of Han and Boyd (1996))

(2.8 Å) as in atactic polystyrene (1.2 Å). A methane molecule jump is on average 1.4 Å in
polyisobutylene and 2.4 Å in amorphous polyethylene. All these values were obtained from molecu-
lar dynamics simulations at 127–140  C (Han and Boyd 1996).
As a thermally activated process, the temperature dependence of diffusion follows the Arrhenius
relationship over limited temperature regions, and, as discussed later, the diffusion activation energy
increases with the size of the solute, since it requires more space for displacement. Over larger
temperature intervals, one has to consider that the activation energy decreases with increasing
7.2 Diffusion 385

Fig. 7.9 Ethyl acetate (○) and propyl acetate (●) diffusivity in poly(methyl acrylate). (Drawn after data of Fujita,
Kishimoto and Matsumoto (1960))

Fig. 7.10 Diffusion coefficients of gases and vapours in polystyrene at 25–30  C as a function of the mean diameter,
calculated from their Lennard-Jones (gas viscosity) volume or liquid molar volume. The line represents the trend among
near-spherical molecules. (Drawn after data of Berens and Hopfenberg (1982))

temperature, which means that the slope of the log (D) versus T1 curve decreases with increasing
temperature (Fig. 7.9).
For a given polymer, the diffusivity decreases with increasing size of the solute molecule. The
correlation between the size and the diffusivity of the solute is evident, as shown in Fig. 7.10. The
same trend is also observed if the diffusivity is plotted as a function of the square of the solute
diameter. Correlations have also been made using solute molar volume at the critical point. The
spherical/near-spherical molecules follow essentially a single line, but for the non-spherical
molecules, the diffusivity is higher than predicted considering the mean diameter obtained from,
for example, the liquid molar volume or gas viscosity, indicating that the solute molecules may orient
386 7 Transport Properties of Polymers

during diffusion. One way to consider the possible orientation of non-spherical molecules during
diffusion is to use the empirical equation suggested by Michaels and Bixler (1961b):

d2
de ¼ ð7:27Þ
dm

de is the effective diameter; d is, for example, the Lennard-Jones diameter; and dm is the maximum
dimension of the solute molecule.

7.3 Solubility

The trend in solute solubility is more complex than the trend in diffusivity. Solubility is a strong
function of the solute–polymer intermolecular forces and may be rationalised in terms of the
solubility parameter concept ‘equal dissolves equal’ (Hansen 2007; Venkatram et al. 2019). Refer
also to Gedde and Hedenqvist (2019a). In order to avoid a large sorption/uptake of the solute in
the polymer, it is important to choose a polymer–solute combination with a sizeable difference in
solubility parameters. Thus, for a polar environment, e.g. water (solubility parameter: δ ¼ 48 J/cm3),
a non-polar polymer, e.g. poly(tetrafluoroethylene) (δ ¼ 13 J/(cm)3), polyethylene (δ ¼ 16 J/cm3) or
polyisoprene (δ ¼ 17 J/(cm)3), should be selected. Consequently, if the environment is non-polar
(n-hexane (δ ¼ 15 J/(cm)3) or carbon tetrachloride (δ ¼ 17.5 J/(cm)3)), a polar polymer,
e.g. polyamide 6,6 (δ ¼ 28 J/(cm)3) or polyacrylonitrile (δ ¼ 31.5 J/(cm)3), should be selected. All
solubility parameter values here were obtained from Grulke (1989). For elastomers and polyethylene,
it has been shown that the logarithm of the solubility of a large number of small molecules increases
linearly with the solute intermolecular energy, measured via the Lennard-Jones force constant ε/k,
where ε is the depth of the potential well and k is the Boltzmann constant. Michaels and Bixler
(1961a) found the following relationship for the solute solubility in the amorphous component of
polyethylene and the Lennard-Jones force constant:

ε
ln ðSa Þ ¼ 0:022  5:07 ð7:28Þ
k

This relationship originates from the fact that the solubility increases with condensability of the
gas. High condensability, in turn, is indicated by a high critical temperature, normal boiling point or
Lennard-Jones force constant of the gas. Less condensable gases prefer to stay in the surrounding
atmosphere and avoid being trapped in the polymer matrix, which involves a condensation step.
The solubility coefficient (S or Sa) of solutes in rubbery amorphous and semicrystalline polymers
follows Henry’s law at low solute concentrations (Fig. 7.11, stage I):
C¼Sp ð7:29Þ

where C is the concentration of solute in the polymer and p is the partial pressure of the solute in the
vapour phase. At higher pressures, i.e. at higher solute concentrations in the polymer, the polymer
becomes plasticised and a deviation from Henry’s law occurs (Fig. 7.11, stage II). The upturn at
higher pressures can also be due to solute clustering/self-association. Stages III and IV occur in glassy
polymers. Dual sorption representing stage III will be discussed in a later section. It should be noted
that other curve shapes also exist for glassy polymers; examples include s-shaped curves and those
that are similar to stage I or stage I/II shapes (Minelli and Sarti 2020). At high pressure, even for
permanent gases like N2, the difference between the real gas and the ideal gas becomes noticeable,
due to attractive forces between the molecules. In this case, the term fugacity, rather than the applied
‘partial’ pressure, is meaningful to use and correlate with, for example, the activity or concentration.
7.3 Solubility 387

Fig. 7.11 Solute

concentration–partial
vapour pressure isotherms:
in stage I, Henry’s law is
valid; in stage II,
plasticisation/clustering
occurs; stage III represents
dual sorption; and in stage
IV plasticisation/clustering
occurs. (Drawn after
Hedenqvist and Gedde
(1996))

It is essentially the partial pressure of the equivalent ideal gas at the same temperature and molar
Gibbs free energy as the real gas. By this definition, the fugacity will have a lower value than the
applied pressure, but becomes equal to the applied partial pressure at low pressure. The fugacity can
be calculated from, for example, the van der Waals gas model or determined from experiments.
The solubility–solute concentration dependence was described by Rogers, Stannet and Szwarc
(1960) (refer also to Rogers (1988)) according to:
S ¼ Sc0 exp ðσCÞ ð7:30Þ

where Sco is the zero solute concentration solubility coefficient and σ describes the size of the
concentration dependence. Note that this equation cannot describe the stage III behaviour in
Fig. 7.11. In a limited temperature region, the Arrhenius equation can be used to describe the
temperature dependence of the solubility (Rogers 1988):
S ¼ S0 exp ðΔH s =RT Þ ð7:31Þ

where ΔHs is the enthalpy of solution, which, in turn, is the sum of the molar enthalpy of condensation
(ΔHc) and the partial molar enthalpy of mixing (ΔHm). ΔHc is negative, whereas ΔHm for permanent
gases is positive and larger than ΔHc. This means that the solubility of permanent gases increases
with temperature. For condensable vapours (e.g. water) on the other hand, the enthalpy of condensa-
tion is usually larger than the enthalpy of mixing, leading to a heat of mixing which is negative and a
solubility which decreases with increasing temperature. The temperature dependence of the concen-
tration of solute in the polymer also depends on the temperature dependence of the vapour pressure.
The solubility of liquids usually increases with increasing temperature due to plasticisation effects
(Rogers 1988).
Examples of solubility models based on statistical thermodynamics will be given here (Nilsson
and Hedenqvist 2011). There are also other models, based on, for example, the UNIFAC model,
which is not further discussed here. For further reading, please refer to Wei and Sadus (2000), Rolker
et al. (2007), Radfarnia et al. (2005), Wibawa and Widyastuti (2009), Wang (2007) and Serna et al.
(2008). In the statistical thermodynamics models, the pressure–volume–temperature properties of the
pure components (gas and polymer) are first determined, and a rule of mixing is then applied to obtain
the solubility data.
The sorption of a solute/gas in a polymer can be estimated by equating the chemical potential of
the solute in the pure gas state (index g) with the chemical potential of the solute in the solute/polymer
mixture (index m):
388 7 Transport Properties of Polymers

g
μ1 ðT, pÞ ¼ μm
1 ðT, ϕ1 Þ ð7:32Þ

where index 1 refers to solute (index 2 is polymer). ϕ1 is the solute volume fraction in the close-
packed solute/polymer mixture. By using the Sanchez–Lacombe equation of state model (SL-EOS)
(Sanchez and Lacombe 1976, 1977, 1978; Lacombe and Sanchez 1976; Challa and Visco, Jr. 2005;
De Angelis et al. 1999; Sadat-Shojaei, Movaghar and Dehghani 2015), the following relationship is
obtained, where the left side represents the pure solute/gas state and the right side the solute/polymer
mixture (De Angelis et al. 1999; Nilsson et al. 2013):
" #
e
ρ1 Pe1 ð1  e
ρ1 Þ ln ð1  e
ρ1 Þ ln ðe ρ1 Þ
r1  þ
0
þ þ ¼
Te1 Te
eρ1 e
ρ1 r 01
M1 ΔP∗ ð7:33Þ
ln ðϕ1 Þ þ ð1  ϕ1 Þ þ e
ρ ð 1  ϕ1 Þ 2 þ
ρ∗
1 RT
e
ρ Pe ð1  e
ρÞ ln ð1  e
ρÞ ln ðe
ρÞ
r1  þ 1þ þ
eρ
Te1 Te e
ρ r1

where Pe = P=P∗ is the reduced pressure, e ρ = ρ=ρ∗ the reduced density and Te = T=T ∗ the reduced
temperature of the mixture. Those with index 1 refer to the same parameters for the pure solute. The
unitless parameter r 01 is the ‘size’ of the solute molecule in the gas phase is given by:

r 01 ¼ M1 = v∗ ∗
1 ρ1 ð7:34Þ

where:

v∗ ∗ ∗
1 ¼ RT 1 =P1 ðsolute molar volumeÞ ð7:35Þ

The corresponding solute size in the mixture is:

r 1 ¼ v∗ 0 ∗
1 r 1 =v ð7:36Þ

where the close-packed volume of the mixture (on a mer basis) can be obtained as (v∗
2 is the polymer
‘mer’ molar volume):
v∗ ∗
1 v2 1
v∗ ¼ ¼ or v∗ ¼ RT ∗ =P∗ ð7:37Þ
ϕ1 v∗
2 þ ð1  ϕ1 Þv∗
1 ϕ 1 =v ∗ þ ð1  ϕ Þ=v∗
1 1 2

The reduced density of the mixture is related to the other reduced parameters through:
 
ρ2 Pe
e ϕ1
e
ρ ¼ 1  exp  þ  1  eρ ð7:38Þ
Te Te r1

and the corresponding density of the pure solute is:

 
ρ21 Pe1
e 1
e
ρ1 ¼ 1  exp  þ  1 0 eρ ð7:39Þ
Te1 Te1 r1 1

The solute solubility in the polymer is obtained by simultaneously solving Eqs. (7.33), (7.38) and
(7.39) for e
ρ, e
ρ1 and ϕ1.
7.3 Solubility 389

P∗ ¼ ε∗/v∗ is the characteristic pressure representing the hypothetical cohesive energy density of
the close-packed mixture (at 0 K), where ε∗ is the ‘mer–mer’ interaction energy. The latter can be
obtained from a suitable mixing rule (Sadat-Shojaei, Movaghar and Dehghani 2015). ρ* is the
corresponding characteristic mass density, and the characteristic temperature is T∗ ¼ ε∗/k or
T∗ ¼ ε∗/R. The corresponding parameters for the pure components are given with indices 1 and
2. For the mixture, the pressure term can be obtained as:

P∗ ¼ ϕ 1 P∗ ∗
1 þ ð1  ϕ2 ÞP2  ϕ1 ð1  ϕ1 ÞΔP
∗
ð7:40Þ

where ΔP∗ is the net change in cohesive energy during mixing at 0 K and can be obtained as:

ΔP∗ ¼ P∗ ∗ ∗ ∗ 0:5
1 þ P2  2ζ P1 P2 ð7:41Þ

where the interaction parameter ζ describes the deviation of ΔP∗ from the geometric mean. It is close
to unity for gas–polymer mixtures without strong interactions (non-polar components) and can be
estimated from solubility parameters (Nilsson et al. 2013).
ρ* for the mixture can be estimated as:
ρ∗ ∗
1 ρ2 1
ρ∗ ¼ ¼ ð7:42Þ
w 1 ρ∗
2 þ ð1  w1 Þρ∗
1 w1 =ρ∗ ∗
1 þ ð1  w2 Þ=ρ2

where w1 and w2 are the weight fractions of the solute and polymer which can be obtained from the
volume fraction:
ϕ1
w1 ¼ ∗ ¼ 1  w2 ð7:43Þ
ϕ1 þ ð1  ϕ1 Þρ∗
2 =ρ1

T∗ for the mixture is obtained as:

P∗
T∗ ¼ or T ∗ ¼ T ∗ v∗ =R ð7:44Þ
ϕ 1 P∗ ∗
1 =T 1 þ ð1  ϕ1 ÞP∗ ∗
2 =T 2

The parameters T∗, ρ∗ and P∗ for the polymer are usually determined by least-squares fitting of
the model (Eq. 7.39, with index 2 and using r2) to the polymer experimental density as a function of
pressure and temperature and for the gas using vapour pressure data. Figure 7.12 shows the fit to
experimental N2 data in PDMS using the model with two different interaction parameters.
Glassy polymers are not in equilibrium and cannot be described by conventional equation of state
models. The non-equilibrium lattice fluid (NELF) model was developed by the Sarti group in Bologna
for obtaining solubility data of solutes in glassy polymers (De Angelis et al. 2007; De Angelis and
Sarti 2011; Minelli, De Angelis and Sarti 2017). The concept relied on using the polymer density as
an internal state variable describing the departure from equilibrium. Basically, only the PVT data of
the pure components and the density of the solid mixture were needed in advance. By considering a
pseudo-equilibrium between the chemical potentials of the solute/gas (μG) and of the solid (μS; we
now use index s is used for the solute–polymer mixture), it is possible to calculate the solute volume
fraction (ϕ1) as a function of gas pressure and temperature:

μG ðT, pÞ ¼ μS T, e
ρ S , ϕ1 ð7:45Þ

The right-hand side of Eq. (7.45) is obtained from:

  0 ∗ ∗ ∗ ∗
r 1  r 01 S r 1 v1 P1 þ P  ϕ2 ΔP
2
μS
¼ ln e ρS ϕ1  r 01 þ ln 1  e
ρ S
 r 1  e
ρ ð7:46Þ
RT e
ρS RT
390 7 Transport Properties of Polymers

Fig. 7.12 N2 concentration as a function of vapour pressure in polydimethylsiloxane (○). The solid line is the equation
of state model using ζ ¼ 0.801, and the broken line corresponds to ζ ¼ 1. (Drawn after data of De Angelis et al. (1999))
pffiffiffiffiffiffi pffiffiffiffiffiffi
where most parameters are described above. ΔP∗ = P∗ ∗ 2
1 - P2 Þ is a first-order approximation
where ζ ¼ 1. e
ρ , the reduced solid density, can be estimated by using:
S

 
ρ ρ w1 1  w1
e
ρ ¼
S
¼ þ ð7:47Þ
ð1  w1 Þρ∗ ð1  w1 Þ ρ∗1 ρ∗
2

where ρ is here the density of the solute-filled polymer. In the case of negligible swelling (gases at low
pressure), this density can be approximated by the density of the pure polymer ρ2. The left-hand side
of Eq. (7.45) can be obtained from the Sanchez–Lacombe equation of state, which results in:

μG ρ r 0 v ∗ P∗
e
ρÞ  r 01 ln ð1  e
¼ ln ðe ρÞ  r 01  1 1 1 ð7:48Þ
RT RT

The equilibrium density of the gas is obtained by minimising the chemical potential, which means
solving Eq. (7.39).
If dilation data is missing, and swelling cannot be neglected, the polymer pseudo-equilibrium
density can be estimated by considering that it decreases in a linear fashion with increasing solute
pressure (Giacinti Baschetti, Doghieri and Sarti 2001):
ρð pÞ ¼ ρ2 ð 1  k pÞ ð7:49Þ

where k is the swelling coefficient and ρ2 is the pure polymer density. When the pressure goes to zero,
the solubility coefficient at the limit of zero pressure can be obtained from (De Angelis, Sarti and
Doghieri 2007):
2 0   ∗    13
T∗ ∗
1 P2 ρ2 ρ02 T∗ ∗
1 P2
6 M P∗ B 1 þ T ∗ P∗  1 ρ0 ln 1  ρ∗ þ T ∗ P∗  1 C7
T STP 6 1 B 2 C7
So ¼ exp 6 ∗ 1∗ B 2 1 2 2 1
C7 ð7:50Þ
TpSTP 4ρ1 RT 1 @ ρ02 T ∗  ∗ ∗ ∗
 A5
þ ∗ ∗ P1 þ P2  ΔP12
1
ρ 2 P1 T
7.4 Fick’s Laws of Diffusion 391

where STP refers to standard temperature and pressure. The solubility of gases and vapours in glassy
polymers has also been modelled by a non-equilibrium perturbed hard sphere chain (NE-PHSC)
model (Doghieri et al. 2006). The perturbed hard sphere chain theory (PHSC) is here combined with
a ‘non-equilibrium thermodynamics for glassy polymers’ approach (NET-GP). The residual
Helmholtz free energy in the PHSC is obtained as the sum of two terms: Ares ¼ Aref + Aper, where
Aref represents the chain connectivity and hard sphere interactions and Aper represents mean field
forces. The perturbation is described by either a van der Waals approach or a square well potential
having a variable width. The solubility of CO2 in glassy polycarbonate (Doghieri et al. 2006) was
successfully described by the model.
The statistical associated fluid (SAFT) models are an important class of statistical thermodynamic
models. Different approaches with the SAFT concept have been developed and used, including the
‘variable range statistical associating fluid theory’ (VR-SAFT) (Gil-Villegas et al. 1997; Galindo
et al. 1998; McCabe et al. 2001), the perturbed chain SAFT model (PC-SAFT) (Gross and Sadowski
2001), simplified PC-SAFT model (Von Solms, Michelsen and Kontogeorgis 2005) and the group-
contribution simplified PC-SAFT (GC-PC-SAFT) model (Tihic et al. 2008). In the PC-SAFT model,
a perturbation with a hard chain reference fluid is used, and the compressibility is considered to have a
contribution from a hard chain (hc) and a dispersion (disp) part:

Z ¼ 1 þ Zhc þ Z disp ð7:51Þ

For full details of the model, see Gross and Sadowski (2001). The PC-SAFT model predicts the
PVT behaviour better than the original SAFT models for non-spherical molecules like toluene since
the ‘chain’ feature is implemented in the former (Gross and Sadowski 2001). For further reading on
the equation of state theory, refer to Gedde and Hedenqvist (2019a).

7.4 Fick’s Laws of Diffusion

Fick’s laws of diffusion are some of the best known equations in physics today, and a large range of
mass transfer cases can be dealt with by the original or modified versions of these laws. Ironically,
these laws were first put on paper by a medical doctor, interested in mathematics, who did most of his
work on the mechanics of different organs in the human body (Cussler 1975). They date back to a
paper by Adolf Eugen Fick in 1855 who put a theory behind the experimental work of Thomas
Graham by considering the analogy between mass transfer and Fourier’s law of heat conduction and
Ohm’s law of electrical conduction (Fick 1855a, b, 1995). Fick considered that the flow rate of a
solute molecule through a unit cross section of a matrix material is proportional to the gradient in
concentration of the solute in the direction of interest (e.g. x-direction):
∂C
F ¼ Fx ¼ D ð7:52Þ
∂x

where C is solute concentration and D is a proportionality constant which determines the rate at which
the solute moves in the matrix material (the diffusivity). This is referred to as Fick’s first law of
diffusion. The concentration gradient is the driving force for mass transfer. This is the case for an ideal
mixture of solute and matrix (or an approximation to this case). In non-ideal cases, it is actually the
gradient in chemical potential that is the driving force. The amount of solute transferred through unit
cross section per time unit is referred to as the flux (F). According to Fick’s first law, the flux depends
only on the diffusivity rate (D) and the concentration gradient (∂c/∂x) (Crank 1986). Consider the
392 7 Transport Properties of Polymers

Fig. 7.13 A plate exposed

to a flow of solute. The
straight slope in
concentration corresponds
to a case where D is
constant and the non-linear
slope to a case where
D increases with
concentration

plate in Fig. 7.13 which is subjected to an unlimited amount of a gas on the left side. At steady state, it
is possible to calculate the flux as:
 
∂c c  c1
F ¼ D ¼ D 2 ð7:53Þ
∂x x2  x1

if the concentration gradient is linear (solid line in the figure) between c1 and c2 that are the solute
concentrations in the plate at the two boundaries x1 and x2. Henry’s law (C ¼ S  p) gives a
relationship between the solute vapour pressure ( p) just outside the plate and the solute concentration
(C) in the plate through the solubility coefficient (S). The law is, at low pressures, valid for most gas/
non-glassy polymer combinations. Assuming that Henry’s law holds, Eq. (7.53) can be rewritten as:
 
p1  p2
F¼DS ð7:54Þ
x2  x1

This equation can be further simplified by introducing the permeability coefficient (P), defined as:
P¼DS ð7:55Þ

In essence, this means that the solute permeability depends on two factors, the diffusivity (D) and
the solubility (S). By controlling these factors, it is possible to steer towards high barrier properties or
specific membrane characteristics.
By using the flux, it is possible to derive the increase in concentration, or simply mass uptake,
within a material element as a function of time (Fick’s second law of diffusion). It is possible to do this
for any basic element, and here it will be derived for the most common elements: the slab/plate, the
cylinder and the sphere. Only unidimensional transport is considered, but the derivations may easily
be extended to two- and three-dimensional transport. Consider the slab in Fig. 7.14.
The mass uptake rate is equal to the mass that enters surface A minus the mass that leaves the slab
through surface A+ per unit time (Crank 1986). The mass uptake rate may also be written as the
increase in concentration as a function of time multiplied by the volume of the slab:
∂C
Ax 2dx ¼ F A  Fþ Aþ ð7:56Þ
∂t
7.4 Fick’s Laws of Diffusion 393

Fig. 7.14 Unidimensional

diffusion in a slab

The flow rates and cross-sectional areas can be expressed with reference to the flow rate at x:
   
∂C ∂ ð Fx Ax Þ ∂ðFx Ax Þ
Ax 2dx ¼ Fx Ax  dx  Fx Ax þ dx ð7:57Þ
∂t ∂x ∂x

In the case of a slab, the cross sections are independent of x and Ax ¼ A ¼ A+. Hence, Eq. (7.57)
reduces to:
∂C ∂F
¼ x ð7:58Þ
∂t ∂x

which, combined with Eq. (7.53), yields Fick’s second law of diffusion:
 
∂C ∂ ∂C
¼ D ð7:59Þ
∂t ∂x ∂x

and if D is independent of concentration and space, then one arrives at:

2
∂C ∂ C
¼D 2 ð7:60Þ
∂t ∂x

It should be pointed out that if the uptake of solute is high, D is increasing with solute concentra-
tion, then the concentration profile is non-linear (Fig. 7.13) and the concept of a permeation constant
(P), as defined above, is invalid. In addition, the use of Henry’s law is not correct either. However,
P in Eq. 7.55 may be approximated using ‘average’ D and S values. The diffusivity in the derivations
here is based on a constant volume, despite the net mass transfer across the slab (Fig. 7.14) or along
the cylinder or sphere elements below. This is an acceptable condition when the solute concentration
is low (e.g. with permanent gases at low pressures). As the concentration increases, it may be
necessary, depending on the approximations made and the accuracy required, to consider changes
in volume (swelling). If the concentration dependence on the diffusivity is strong, the effects of
volume changes may, however, be of lower importance.
Fick’s second law of diffusion is now derived for a cylinder. Consider the cylinder element in
Fig. 7.15.
The mass uptake rate in the cylinder element is equal to the mass flow in through surface A minus
the mass flow out from the surface A+:
   
∂C ∂ðFr Ar Þ ∂ ð Fr Ar Þ
Ar 2dr ¼ Fr Ar  dr  Fr Ar þ dr ð7:61Þ
∂t ∂r ∂r

where Ar ¼ rdθ dz. Eq. (7.60) is simplified to:

394 7 Transport Properties of Polymers

Fig. 7.15 Diffusion along

radius in a cylinder

∂C ∂ðFr r Þ
r¼ ð7:62Þ
∂t ∂r

With the use of Eq. (7.52) for radial mass flow, Eq. (7.62) can finally be written as Fick’s second
law of diffusion in cylindrical coordinates:
 
∂C 1 ∂ ∂C
¼ rD ð7:63Þ
∂t r ∂r ∂r

If D is independent of concentration and space, then the above equation becomes:

 2 
∂C 1 ∂C ∂ C
¼D þ ð7:64Þ
∂t r ∂r ∂r 2

Consider now radial diffusion in a sphere (Fig. 7.16).

The mass uptake rate in the sphere element, bounded by the sides: 2dr, rdθ and rsin(θ)dϕ, is:

∂C 2 ∂ðFr r 2 Þ
r ¼ ð7:65Þ
∂t ∂r

which, with Eq. (7.52), is simplified to the final form:

 
∂C 1 ∂ ∂C
¼ 2 2
r D ð7:66Þ
∂t r ∂r ∂r

If D is concentration (and space) independent, the above equation is simplified to:

 2 
∂C 2 ∂C ∂ C
¼D þ 2 ð7:67Þ
∂t r ∂r ∂r

Besides the equations describing the diffusion of the solute in the material, the boundary
conditions also affect the flow through the specimen. The simplest boundary condition is where the
surface concentration is constant, such as in the case when the vapour pressure of a gas is constant
immediately outside the specimen, or when there is a steady fast removal of diffusing solute from the
low concentration side. If the rate of removal is slow enough to affect the overall flux, this has to be
considered, which is usually accomplished through an evaporation condition at the surface. The rate
of mass transfer at the surface depends on the relationship between the flux at the surface and the rate
7.5 Methods of Solution of the Diffusion Equation 395

Fig. 7.16 Radial diffusion in a sphere

of evaporation (the latter described by an evaporation constant FE). For unidirectional flow in a plate,
the following relationship governs the surface condition (Crank 1986):
∂C
D ¼ FE ðC  C0 Þ ð7:68Þ
∂x

The rate of mass transfer to the surface (left-hand side) equals the evaporated mass (right-hand
side). The gradient in concentration can be estimated from the gradient in concentration just within
the slab surface, and C can be taken as the concentration just within the surface. C0 is the surface
concentration in equilibrium with the concentration/vapour pressure of the solute outside the low
concentration side. If solute is effectively removed by, for example, a flow of air, then C0 will be zero.
If the side with the surface A+ in Figs. (7.14) to (7.16) is impermeable, the solute gradient at A+ is
maintained at zero (a mirroring condition/isolated point). If the inflow of solute into the slab is
symmetrical with the same flux through surfaces A and A+, only half of the slab is enough to
consider, where now the solute gradient at Ax is maintained at zero (Fig. 7.14). Further information
about boundary conditions are presented in Crank (1986).

7.5 Methods of Solution of the Diffusion Equation

There are several ways to solve Fick’s second law of diffusion if the diffusion coefficient is constant.
The method of separation of variables yields solutions in the form of orthogonal Fourier series for
plate geometries and Bessel series for cylinder geometries. The method of reflection and superposi-
tion yields solutions containing series of error functions. The error functions are preferably used at
short times due to their rapid convergence there and to problems in infinite or semi-infinite
geometries. The orthogonal series, however, converge rapidly at longer times and are, therefore,
also more suited for solutions at longer times. The Laplace transform methods can also be used to
396 7 Transport Properties of Polymers

solve the diffusion equation, and they can be applied to both short- or long-term solutions. These
methods mentioned above are described in detail in Crank (1986) and Carslaw and Jaeger (1959).
One example where the method of separation of variables is used will be given here. Consider a
plate which at time 0 is exposed to an environment that yields the surface concentration C0. An
expression for the concentration as a function of time and space (x) within the plate of thickness 2 l
will be derived. Assume that D is constant. The concentration of the solute can be described by two
independent terms (Crank 1986):
C ¼ XðxÞ T ðtÞ ð7:69Þ

where the first is a function of x only and the latter is a function of time only. Differentiation and
insertion of Eq. (7.69) in Eq. (7.60), provided D is constant, yield:
2
∂T ∂ X
X ¼ DT ð7:70Þ
∂t ∂x2

By rearranging, Eq. (7.70) becomes:

2
1 ∂T D ∂ X
¼ ð7:71Þ
T ∂t X ∂x2

The right- and left-hand sides must now be equal to a constant which is chosen, for convenience, to
be equal to λ2 D. Hence, Eq. (7.71) may be written as:
1 ∂T
¼ λ2 D ð7:72Þ
T ∂t

and:
2
1∂ X
¼ λ2 ð7:73Þ
X ∂x2

The general solutions to Eqs. (7.72) and (7.73) are:

T ¼ exp λ2 D t ð7:74Þ

and:
X ¼ A sin ðλxÞ þ B cos ðλxÞ ð7:75Þ

The general solution is obtained by adding a series of terms:

X
1
C ¼ C0 þ ðAm sin ðλm xÞ þ Bm cos ðλm xÞÞ exp λ2m D t ð7:76Þ
m¼1

The initial condition (I.C.) is:

C ¼ 0,  l < x < l, t ¼ 0 ð7:77Þ

The boundary conditions (B.C.) are:

C ¼ C0 , x ¼ l, x ¼ l, t > 0 ð7:78Þ
7.5 Methods of Solution of the Diffusion Equation 397

The B.C. yield Am ¼ 0 and λm ¼ mπ/(2 l ). Hence, Eq. (7.76) can be written as:
X1    
mπ m2 π 2
C  C0 ¼ Bm cos x exp  2 D t ð7:79Þ
m¼1
2l 4l

Bm is obtained by first multiplying both sides of Eq. (7.79) by cos(pπx/l ) and integrating between
0 and l (this is taken twice since the thickness is 2 l ):
0 l 1 0 1
X1 ð  X1 ðl    
@Bm 2 cos pπ x cos mπ x A exp  m π D t
2 2
ðC  C 0 Þ @2 cos pπ x A ¼
m¼1
2l m¼1
2l 2l 4l2
0 0
ð7:80Þ

The integral on the right-hand side is:

ðl  
pπ mπ
2 cos x cos x dx ¼ l ð7:81Þ
0 2l 2l

for m ¼ p and 0 for m 6¼ p. The integral on the left-hand side is:

ðl 
pπ 4l
2 cos x dx ¼ ð1Þn ð7:82Þ
0 2l pπ

Since only odd numbers of p and m yield a non-zero solution, we can use n ¼ 0,1,2,3. . . and
p ¼ m ¼ 2n + 1. Eq. (7.79) then becomes:
X1   X1  
4 ð2n þ 1Þ2 π 2
ðC  C0 Þ ð1Þn ¼ ðBn Þ exp  D t ð7:83Þ
n¼0
π ð2n þ 1Þ n¼0 4l2

The initial condition now provides an expression for Bn:

4C0
Bn ¼  ð1Þn ð7:84Þ
π ð2n þ 1Þ

that leads to the final expression:

" #
1    
4 X ð1Þ
n
ð2n þ 1Þπ ð2n þ 1Þ2 π 2
Cðx, tÞ ¼ C0 1  cos x exp  D t ð7:85Þ
π n¼0 ð2n þ 1Þ 2l 4l2

The concentration profiles for this special case are shown in Fig. 7.17. The curves are generated
using 50 terms in Eq. (7.85).
The ‘normalised’ mass uptake in the plate as a function of time is obtained by integrating
Eq. (7.85):
ðl
Mt 1
¼ Cðx, tÞ dx ð7:86Þ
M1 2l C0
l

yielding:
398 7 Transport Properties of Polymers

Fig. 7.17 Concentration

profiles in the region
0 < x < l generated with
Eq. (7.85)

X1  
Mt 8 1 ð2n þ 1Þ2 π 2
¼1 exp  Dt ð7:87Þ
M1 n¼0
π 2 ð2n þ 1Þ2 4l2

by knowing that:
ðl  
ð2n þ 1Þπ 4l
cos x dx ¼ ð1Þn ð7:88Þ
l 2l ð2n þ 1Þπ

When the diffusion coefficient is not constant, Fick’s second law of diffusion has to be solved
numerically (Kahaner, Molar and Nash 1989, Greenberg 1988). The simplest numerical method is the
explicit Euler forward method. Explicit means that the concentration at time ‘j + 1’ (Cj + 1) is
calculated from the known concentration at the previous time (Cj):

C jþ1 ¼ C j þ h jf C j ð7:89Þ

where hj is the size of the time step. The Euler forward finite difference representation can be
expressed as:
 
∂C ∂ ∂C
¼ DðCÞ !
∂t ∂x ∂x
ð7:90Þ
Ci,jþ1  Ci,j Diþ0:5,j Ciþ1,j  Ci,j  Di0:5,j Ci,j  Ci1,j
¼
hj hi 2

where i is the coordinate in the x-direction, hi is the size of the spatial step in the x-direction and
Di  0.5 ¼ Di  1 + Di)/2. Although explicit methods are straightforward and simple to implement,
they have seen very limited use in this field due to their numerical instability. For most practical
problems, the time step must be kept very small in order for the method to converge, if it will ever
converge. Implicit numerical methods, on the other hand, are more stable and less sensitive to the size
of the time step. The term implicit refers to the ‘unknown’ concentration in the next time step now
existing on both sides of Eq. (7.89). The ‘backwards’ implicit scheme is simply:

C jþ1 ¼ Cj þ hjf C jþ1 ð7:91Þ

7.5 Methods of Solution of the Diffusion Equation 399

Fig. 7.18 A volume fixed

section of a slab defined as
where there is no net bulk
flow out of or into the
section (considering only
molecules A)

Cj + 1 on the right-hand side can be estimated from the Jacobian: df(C)/dC. A well-known implicit
method is the trapezoidal or Crank–Nicolson method. When even more stable methods are needed,
for example, when the diffusivity changes rapidly with changes in concentration, multistep back-
wards differentiation implicit methods have been shown to be very suitable, which involves a
treatment of spatial gradients in concentration involving also the Laplace operator on both sides
(Edsberg and Wedin 1995, Hedenqvist, Krook and Gedde 2002):

6 1 3 2
∇2 C jþ1  ∇2 C j ¼ h f t jþ1 , C  ∇ C j þ ∇C j ð7:92Þ
11 j jþ1
hj 2

The diffusivity obtained can be different depending on how the reference frame is chosen. When,
for example, a slab swells during sorption of the solute, it matters how the reference frame is chosen.
In a highly swelling system, it can be convenient to use the unswollen cross-sectional area as the
reference frame (Hedenqvist 1998). In highly swelling systems, depending on how the diffusivity is
defined, the increase in thickness following the uptake may also have to be considered in Fick’s
second law.
A relationship will be derived here between the mutual diffusion coefficient D, which is including
both true molecular intrinsic diffusion and bulk flow, and the intrinsic diffusion coefficient of A
molecules (DA) in a polymer slab (B), using a fixed total volume in the system (Crank 1986)
(Fig. 7.18). DA is, in other words, the diffusion coefficient of A in the absence of a bulk flow of A
molecules inside the polymer slab (B).
Consider that the rate of accumulation of volume on the left side of the section can be defined as:
∂CA ∂CB
ψ ¼ V A DA þ V B DB ð7:93Þ
∂x ∂x

where Vi and Ci are, respectively, the specific volume (e.g. m3 kg1 of A) and concentration
(e.g. kg of A per total m3) in the volume fixed frame of species i. Assume also that the Vi’s are
constant. The bulk flow that counteracts the accumulation as described by Eq. 7.93 for A is ψCA. The
net flux of A corrected for bulk flow can now be written as:
∂CA ∂CA
D ¼ DA  ψ CA ð7:94Þ
∂x ∂x
400 7 Transport Properties of Polymers

where the first term on the right side is ‘true’ diffusion and the second is bulk flow. Further, it is
possible to write:
∂CA ∂CB
VA þ VB ¼0 ð7:95Þ
∂x ∂x

since the total volume (fraction) consists of only A and B: VACA + VBCB ¼ 1. Combining Eqs. (7.94)
and (7.95) yields:
∂CA ∂CA ∂CA ∂CB
D ¼ DA  CA V A DA  CA V B DB ð7:96Þ
∂x ∂x ∂x ∂x

which becomes:
D ¼ DA þ V A CA ðDB  DA Þ ð7:97Þ

Since the diffusivity of the polymer is negligible compared to that of the solute DB ! 0, which
leads to the final expression for the relation between the mutual and intrinsic diffusivity:
D D
DA ¼ ¼ ð7:98Þ
1  V A CA 1  vA

where vA is the volume fraction of A. Again, if the concentration of the solute (A) is fairly small in the
polymer, DA  D, i.e. the intrinsic and mutual diffusivities are equal. In a non-ideal thermodynamic
system (e.g. non-ideal mixing), the gradient in chemical potential should be considered rather than
just the concentration gradient. This is well-described in Koros, Burgess and Chen (2015).

7.6 Transport Properties of Elastomers and Melts

7.6.1 Introduction

The earliest studies of transport properties of polymers were conducted on rubbers. It was observed
early on that many rubbers are highly permeable to solutes. Butyl rubber is, however, one exception.
The two methyl groups on the isobutylene backbone are bulky, and the stereoregularity assures
efficient chain packing. The dense packing (low free volume), in turn, makes it difficult for solute
molecules to penetrate the polymer. Additionally, solute diffusivity is reduced by the low chain
flexibility due to the high contact area and friction between the densely packed chains (high
damping). In contrast, highly elastic rubbers, e.g. natural rubber and poly(butadiene), possess high
diffusivity towards most small molecules because of their high free volume and high chain flexibility.
The ranking of the permeability and diffusivity to gases among elastomers relates fairly well to their
specific glass transition temperatures (Matteucci et al. 2006). The solute solubility in the elastomer is
high for solutes with solubility parameters similar to the polymer solubility parameter, and the
solubility decreases with crosslink density and, usually, also with fillers.
A good starting point when describing transport properties of different polymers is the Hopfenberg
map (Fig. 7.19), which was based on diffusivity data on hydrocarbons. It is, however, ‘general’ and
can be used for other types of solutes. The map is modified here to also consider mechanical stress
effects which occur at high solute concentrations (dashed region).
7.6 Transport Properties of Elastomers and Melts 401

Fig. 7.19 Diffusion mechanisms of hydrocarbons in amorphous polymers as a function of penetrant activity and
temperature: (a) pure Fickian behaviour with constant D, (b) concentration-dependent D, (c) anomalous diffusion,
(d) Case II diffusion and (e) Case II diffusion coupled with crazing. The dashed region corresponds to concentration-
dependent D coupled with mechanical stress effects. The thick line is the solute concentration dependence of the glass
transition. (Drawn partly after Hopfenberg and Frisch (1969))

Solute diffusivity in elastomers and rubbery semicrystalline polymers, i.e. polymers above Tg, may
be divided into three modes, as shown in Fig. 7.19. At low solute concentrations, the solute diffusivity
can be described as pure Fickian diffusion which means that D is constant. Typical cases are diffusion
of permanent gases, i.e. O2, or non-solvents. At higher solute concentrations (activities), the solute
plasticises the polymer, and the diffusivity usually increases with solute concentration. In the third
mode, the solute concentration is high enough to deform the sample, and mechanical stresses have to
be implemented in the description of diffusion. The first two modes will be discussed in this chapter,
whereas stress effects will be discussed later.

7.6.2 Models of Diffusion in Elastomers

Most diffusion models for elastomers consider the diffusion and solution processes as being a result
of the formation and disappearance of holes or microcavities in the polymer matrix. A solute
molecule is ‘sorbed’ in a hole of suitable size, and with sufficient thermal energy, it may subsequently
jump to an adjacent hole. The distribution of the sizes of holes may vary in space, but the average hole
size and hole size distribution are constant at a given temperature. Diffusion, or the solute jump
process, is considered as a thermally activated process, and the temperature dependence usually
follows the Arrhenius equation in a limited region. The models of diffusion may be classified as either
of molecular or free volume types. The former models are based on the analysis of specific motions of
polymer segments and solutes and include intermolecular interactions, whereas the latter models
consider that the diffusivity is mainly determined by the available free volume in the system. Some
models are described briefly below.
402 7 Transport Properties of Polymers

Fig. 7.20 Jump

mechanism according to
Meares (1954)

7.6.2.1 Molecular Models

In the Meares model, the activation energy of diffusion Ed is considered as the energy required to form
a cylinder in the polymer matrix of sufficient diameter to allow for the solute to move (Fig. 7.20)
(Meares 1954; Koros, Burgess and Chen 2015). Hence, it is not identical to the energy required to
form a hole of sufficient size for the solute to jump into.
Meares derived the following equation for the activation energy:
1
Ed ¼ π σ 2 N A λ ðCEDÞ ð7:99Þ
4

where λ is the jump length, σ is the solute collision diameter, NA is Avogadro’s constant and CED is
the cohesive energy density of the polymer. By the use of this equation, Meares calculated jump
lengths for He and Ar in elastomers to be, respectively, 26.5 Å and 28.9 Å. The equation also predicts
a linear increase in activation energy as a function of σ 2. Van Amerongen (1964) proposed an
alternative relationship based on a comparison between the activation energies of diffusion and
viscous flow:
Md ðCEDÞ
Ed ¼ V ð7:100Þ
Mv n

where Md and Mv are the typical chain lengths involved in diffusion and viscous flow, respectively.
V is the molar volume of the kinetic chain segment in flow or diffusion, and n is a constant that, for
liquids, takes values between 3.5 and 4. The relationship between the cohesive energy density and the
activation energy has been verified for some elastomers (Meares 1954). Since CED decreases
strongly with increasing temperature, so does Ed (Eq. 7.100).
Barrer considered that a solute jump takes place when a total energy which is equal or larger than a
critical value E* is absorbed in a small region surrounding the solute molecule (Barrer 1957). The
energy in this activated zone is divided between the polymer chain segments and the solute molecule
and is stored in f degrees of freedom. The increase in degrees of freedom yields a loosening of the
polymer network and a rearrangement of polymer segments (Fig. 7.21).
Synchronised moves of polymer chain segments then force the solute molecule into a new
position. Barrer derived the following equation for the diffusion coefficient:
"  ∗  f 1  #
1 2 Xmax
f¼ f
E 1 E∗
D¼ λ ρfν exp  ð7:101Þ
2 f ¼1
RT ð f  1Þ! RT
7.6 Transport Properties of Elastomers and Melts 403

Fig. 7.21 Jump

mechanism according to
Barrer (1957). The grey
area is the activated zone
E*

where λ is the jump length, ρf is the probability that the f degrees of freedom (e.g. segmental and
intermolecular displacements) will cooperate in a diffusion jump and ν is the solute molecule
vibration frequency (1012 s1). E* is related to the apparent activation energy through the Arrhenius
relationship:
∂ð ln ðDÞÞ
Eapp ¼ R ¼ E∗  ð f  1Þ RT ð7:102Þ
∂ T1

where f ranges, according to the model, from 15 to 38, which indicates that rather long chain segments
cooperate in the diffusion step. Further, f, and hence the size of the activation zone, is temperature
dependent.
In order to create a path for a diffusing molecule through the polymer, Brandt considered that it
was also necessary to ‘bend’ neighbouring polymer chains to overcome their intermolecular energy.
This is referred to as the Brandt model (Brandt 1959; Brandt and Anysas 1963). The activation energy
thus consisted of a bend term (Eb) and an intermolecular term (Ei), and the total energy (E*) is:

E∗ ¼ Eb þ Ei þ Eth ð7:103Þ

where Eth is the thermal energy. Eth does not contribute to the activation energy. Brandt calculated
diffusivities based on the activated zone theory and found that the Barrer model yielded 25–70%
lower values compared to experiment (ethane in polyethylene) (Crank and Park 1968). It was claimed
that the deviation is due to that not all diffusion mechanisms is encountered in the activated zone
theory. In the Brandt model, the activated process, i.e. the displacement of neighbouring chains, is
only necessary if the free space adjacent to the solute molecule is too small to allow for a jump.
Hence, the model suggests that the smallest solutes may diffuse without the need of an activation
process (Fig. 7.22).
To verify the model, Brandt and Anysas (1963) measured the diffusion of CO2, CH3Cl, CHF3 and
SF6 in different fluoropolymers. In addition, when plotting the apparent activation energy for these
solute–polymer pairs, it was found that the apparent activation energy increased in a non-linear way
with σ 2 (Fig. 7.23). The curve exhibited a downward concavity with Eapp disappearing at σ 2 ¼ 5–8 Å2.
This indicates that solutes smaller than σ ¼ 2.2–2.8 Å are able to diffuse without the need of an
activation energy. This value was suggested to refer to the average distance between molecular
surfaces of neighbouring chains.
The molecular DiBenedetto and Paul model is more detailed than, e.g. the Meares model, because
it contains more molecular factors (DiBenedetto 1963a, b; DiBenedetto and Paul 1964, 1965). The
polymer is considered to consist of NA number of ‘n-centre’ polymer segments, where a centre is the
repeating unit (-CH2- for polyethylene) and NA is Avogadro’s number. The solute molecule is
404 7 Transport Properties of Polymers

Fig. 7.22 Jump

mechanism according to
Brandt (Brandt 1959;
Brandt and Anysas 1963).
The upper part shows the
features of a jump requiring
activation energy, and the
lower part illustrates a
jump without the need of
activation energy

Fig. 7.23 Correlation of

the apparent activation
energy for diffusion and the
square of the solute
molecular diameter for a
range of gases in poly(vinyl
acetate). (Drawn after data
of Brandt and Anysas
(1963))

Fig. 7.24 A solute

molecule about to enter
between four parallel
polymer chains (a).
Figure b shows the
activated state, i.e. the
instant when the solute
makes a jump along the
parallel chains. (Drawn
after DiBenedetto (1963b))

considered as a three-dimensional harmonic oscillator placed in an equilibrium position between four

parallel chains (Fig. 7.24).
In order for the solute molecule to jump to a new position, a cylindrical hole of volume equal to
2λn (π/4) dg2, where 2λ is the length of the centre and dg is the diameter of the cylinder, has to be
created. The interaction energy between centres on different chains is described by the Lennard-Jones
6-12 potential, and interactions between the solute molecule and the polymer chains are neglected.
Compressive stress on the neighbouring chains due to the expansion of the system with the four
7.6 Transport Properties of Elastomers and Melts 405

Fig. 7.25 Model for chain

interaction. Dashed lines
are the chain axes. (Drawn
after DiBenedetto (1963a))

parallel polymer chains is also neglected in the analysis. The total interaction energy between a single
centre and 2 J + 1 centres on a neighbouring chain is:
"   ∗ 6 #
X J
r∗
12
r
∗
Uik ¼ ε 2 ð7:104Þ
k¼J
r ik r ik

where ε* and r* are energy and distance parameters and rik is the distance between a specific centre
(k) and centre i (Fig. 7.25).
The treatment is limited to ten centres on each side of a specific centre, and the interactions are
considered to disappear beyond 5–10 Å from the centre. The activation energy (Ed) is proportional to
ΔU, which is the energy needed to create the activated state.
Ed can be written (Crank and Park 1968):
( " 11  11 # " 5  5 #)
4nN A ε∗ r ∗ v∗ 2 v∗ 2 v∗ 2 v∗ 2
Ed ¼ 0:77   2:32  ð7:105Þ
2λ va v va v

where va and v are the average volumes per centre in the activated and non-activated states. Assuming
that Ed is the same as the apparent activation energy, the model was tested on different gases by
plotting activation energy against normalised solute collision diameter, defined as:

2λn π4 dg 2
d∗
2
g ¼ nv
ð7:106Þ

In Fig. 7.26, the experimental diffusion activation energies of gases in a rubbery ethylene–-
propylene copolymer are shown together with a best fit of the model using suitable data of ε*, r*
and v. n was varied until the agreement between the model and the experiment was optimal. n ¼ 8.6
yielded the best fit, which suggests that a number of 8–9 repeating units per chain are involved in the
diffusion jump/step. The figure also indicates that a limiting activation energy is approached when dg
increases. This may be explained by the decrease in the interchain interaction energy at larger
distances. Contrary to the earlier models, the activation energy disappears as the collision diameter
goes to zero. This is because the activation energy in this theory is connected only to the work of
creating the extra volume necessary for the solute displacement.
The Pace and Datyner model incorporates features from both the DiBenedetto and Paul model and
the Brandt model (Pace and Datyner 1979a, b, 1980a, b). The amorphous polymer is considered to
have the same basic structure as shown in Fig. 7.24. The solute molecule is, in this model, able to
move along the parallel polymer chains and also in a direction perpendicular to them, the latter in
accordance with the Brandt model (Fig. 7.24). Diffusion along the polymer chains does not involve
406 7 Transport Properties of Polymers

Fig. 7.26 Normalised

experimental diffusion
activation energy in
rubbery
ethylene–propylene
copolymer (Ed (in cal/mol)/
RT, where T ¼ 296 K)
versus normalised solute
collision diameter. The
curve is the best fit of the
model (Eqs. (7.105) and
(7.106)). (Drawn after data
of DiBenedetto and Paul
(1965))

Fig. 7.27 The solute jump

mechanism perpendicular
to the polymer chain axis in
the relaxed (a) and the
activated stage (b). (Drawn
after Pace and Datyner
(1979a))

any activation energy, whereas diffusion perpendicular to the chains requires an activation energy
large enough to separate the chains to a distance which allows for the solute to pass (Fig. 7.27).
These two processes are suggested to occur in series since, if they occurred in parallel, only the
activation energy-free process would be observed. In series, the first process is halted whenever the
solute meets an obstacle, e.g. an entanglement, and the solute has to move to another ‘tube’ through
the activated process. The activated process is thus the rate-limiting jump event, and, therefore, it
determines the experimentally observed activation energy. Pace and Datyner (1979a) derived the
following relationship for Ed:
8   93
>
>
11 10 >4
 ∗ ∗ 3 > 0
ðr e  10σ Þ  r∗
r∗
< 0:077 re
re
r∗þσ 0 >
>
=
ε r 1
4
Ed ¼ 5:23 B  5  ð7:107Þ
λ2 >
σ4 >
1
4 >
>
>
: 0:58 r∗
re
ðr e  4σ 0 Þ  r∗ r∗þσ 0
>
;
re

where re, B and λ are the equilibrium chain separation distance, chain bending modulus and the
average chain length between polymer centres (backbone units). B is related to the polymer chain
stiffness. r* and ε* are, respectively, the average distance and the energy parameters per polymer
centre in the Lennard-Jones 6-12 equation. σ is the collision (hard sphere) diameter of the solute and
σ‘¼σ + r*-re. B is 12·103 (polyethylene), 20·103 (isotactic polypropylene) and 5.5·103 (poly
(ethylene terephthalate), average of overall and flexible bond modulus) J nm mol1 (Pace and
Datyner 1979a). The activation energy of diffusion, according to this theory, increases with the
7.6 Transport Properties of Elastomers and Melts 407

polymer chain stiffness and the cohesive energy density of the polymer (through ε*), decreasing
equilibrium chain separation distance and increasing solute size. The apparent activation energy is
related to Ed through: Eapp ¼ d(Ed/T )/d(1/T ). Eapp is thus constant if Ed is a linear function of T. Ed is
temperature dependent through re which increases with temperature, and Ed is also approximately a
linear function of σ. Pace and Datyner found satisfactory agreement between calculated Ed as a
function of σ for a number of small solute molecules in amorphous and semicrystalline polymers
(Pace and Datyner 1979a, b).

7.6.2.2 Free Volume Models

An early free volume model of viscosity and mobility (due to Bueche) was based on the Slater
fluctuation theory (Bueche 1953, 1956, Crank and Park 1968). Bueche adopted the theory by first
considering the fluctuation in volume V around the average value V0 of a given number of particles
(n), also involving the isothermal compressibility (ß):
 2
V  V0 k Tß
¼ ð7:108Þ
V0 V0

The volume distribution function f(V) around V0 is:

 
1 ðV  V 0 Þ2
f ðV Þ ¼ exp  ð7:109Þ
ð2πßV 0 RT Þ0:5 2βV 0 RT

According to Bueche, when, due to volume fluctuations, a polymer segment (occupying Vo)
experiences an increase in volume to Q, it will be able to move. The number Ð 1 of segments which
are able to move relates to the number of groups that have a volume Q: Q f ðV ÞdV . Hence, the
frequency/probability of motion can be formulated as:
ð1 (  0:5 )
1 ðV  V 0 Þ2
P ¼ P0 f ðV ÞdV ¼ P0 1  erf  ð7:110Þ
Q 2 2βV 0 RT

where erf is the error function and Po is a constant. The average polymer volume Vo is considered to
be linearly dependent on temperature above the glass transition:
V 0 ¼ V 00 ½1 þ ðα1 þ α2 Þ T  ð7:111Þ

where (α1 + α2) is the total volume expansion coefficient of the melt, α1 is the volume expansion
coefficient of the glass and α1 is the melt volume expansion coefficient minus that of the glass. Voo is
the volume of the molten polymer at zero K, extrapolated from the melt. The minimum volume
necessary for motion is, in the same way, given by:
Q ¼ Q00 þ α1 V 00 T ð7:112Þ

and Q00 is Q at zero K. If Eqs. (7.111) and (7.112) are inserted in Eq. (7.110), the final relationship for
P is:
  
1 A
P ¼ P0 1  erf  B T 0:5
ð7:113Þ
2 B T 0:5
408 7 Transport Properties of Polymers

Diffusion coefficient (cm /s)

Fig. 7.28 Toluene and

2
trichloroethane diffusivity
-6
in nitrile rubber versus 10
viscosity of nitrile rubber.
(Drawn after data of
Vahdat (1991))

-7
10
0.004 0.007 0.01
2
Viscosity (cm /s)

where:
 
Q00  V 00
A ¼ α2 ð7:114Þ
2βR

and:
 0:5
V 00
B ¼ α2 ð7:115Þ
2βR

Bueche used Eq. (7.114) to match viscosity data, and Kumins and Roteman (1961) applied the
theory to mass transport data by considering the analogy between viscosity and diffusivity. The
correlation between diffusivity and viscosity in the liquid-like/rubbery state is illustrated in Fig. 7.28
for several solutes in nitrile rubber.
Kumins and Roteman (1961) considered that the distribution of free volume in the polymer is
constant in time. However, the free volume distribution in space varies with time. They measured
diffusion coefficients of several non-solvating and non-reacting gases of different sizes at different
temperatures in poly(vinyl chloride-co-vinyl acetate), and from the best fit of the data, it was possible
to determine A and B and, subsequently, also Qoo and Voo. The relationship between the solute
occupied volume (Qoo  Voo) and the van der Waals diameters of the gases was good. The smallest
molecule (H2) requires a free volume hole of 56 Å3, whereas the largest molecule (CO2) requires
690 Å3. Ne needs a free volume hole of 224 Å3. If it is assumed that the free volume necessary for a
diffusive jump is of spherical shape, it means that the minimum hole diameters for He, Ne and CO2
are 2.37, 3.76 and 5.48 Å, respectively. This should be compared with the molecular diameters of,
respectively, 2.34, 2.8 and 3.23 Å. The large difference for CO2 is due to the non-spherical shape of
the molecule. The corresponding values for the segment occupied volume (Voo) are 76, 698 and
3000 Å3. The higher values of Qoo and Voo for larger solutes are due to the larger polymer segments
involved in the solute jumping process. The ratio between Qoo and Voo is approximately 1.3 for most
solutes, except for the smallest molecules, where the ratio is larger (it is 1.74 for H2) (Kumins and
Roteman 1961). This is suggested to be due to short chain segments being involved here, and they
have to be perturbed to a larger extent in order to ‘open up’ the required free volume, thereby
increasing Qoo.
In the Cohen, Turnbull and Fujita free volume model (CTF theory), the probability of finding a
free volume in a homogeneous liquid which is greater than v* is the starting point:

Pðv∗ Þ ¼ exp ðbv∗ = < v >Þ ð7:116Þ

b is a constant which is close to 1 and <v> is the average free volume in the system of identical
molecules (Cohen and Turnbull 1959). The free volume is here considered as the volume of a box
7.6 Transport Properties of Elastomers and Melts 409

containing a molecule minus the molecule volume. Eq. (7.116) is valid for a pure homogeneous liquid
but may be applied to a concentrated polymer solution (binary system). If <v> is set equal to f, i.e. the
fractional free volume or the free volume per unit volume instead of per molecule, and if the product
bv* is denoted B, it is possible to write:
PðBÞ ¼ exp ð B=f Þ ð7:117Þ

It is assumed that the mobility of a solute molecule in a polymer matrix is dependent on the
probability of finding a hole in the vicinity of the solute which is large enough to accommodate the
solute, or part of it. The mobility, md, is therefore proportional to P(B):
md ¼ Ad exp ð Bd =f Þ ð7:118Þ

where Bd is the critical value of B for a diffusive jump to occur. The proportionality factors Ad and Bd
are dependent only on the size and shape of the solute molecule. Consequently, they should be
independent of temperature and concentration of solute. Fleischer (1984) analysed the diffusion of n-
alkanes in polyethylene (n 16, n is the number of carbons) and found the following n dependence
for Bd:
Bd ¼ 0:4 þ ðn  2Þ 0:1 ð7:119Þ

The methyl group contributes with a value of 0.2 to the Bd-factor, whereas the methylene group
contributes with only 0.1. Von Meerwall et al. (1982) found that the Bd value remained practically
constant for n-alkanes with n ¼ 8–36. This suggests that when the solute is sufficiently long, only
parts of the molecule are jumping between the accessible holes. Kulkarni and Stern (1983) considered
that the Bd-factor is the ratio of the size of the solute molecule and the size of a polymer segment
involved in the viscous flow of the pure polymer due to solute movement. According to this view, Bd
can only take values between 0 and 1. Fleischer (1984), however, published Bd values greater than
unity. Another assumption that is central in this free volume theory is the volume additivity on
mixing. This assumption is reasonable for many systems, particularly those involving weak intermo-
lecular interactions. Thus, the fractional free volume ( f ) can be expressed as:
f ¼ f 1 ϕ1 þ f 2 ϕ2 ð7:120Þ

where vi are volume fractions, fi are the fractional free volumes and 1 and 2 represent the solute and
polymer, respectively. Fleisher arrived at the following expression relating diffusivity to free volume:
D ¼ A exp ðB=f Þ ð7:121Þ

It has been shown that the CTF free volume theory is applicable to systems which sorb less than
30% solute (Blackadder and Keniry 1974), although Kobayashi et al. (1994) managed to apply the
model to systems with higher solute concentrations.
It should be mentioned that the Fujita type CTF free volume theory has been less successful in
describing diffusion in highly polar systems, e.g. of water in poly(methyl acrylate) (PMA) (Aifantis
1980).
A well-known free volume model of diffusion is the Vrentas–Duda free volume model. This model
is more detailed than the Fujita model (Duda, Ni and Vrentas 1979; Vrentas, Duda and Lau 1982;
Vrentas, Duda and Ling 1989; Vrentas and Chu 1989; Sharma, Tewari and Arya 2017; Fang and
Vitrac 2017). It is considered here that the specific volume of a liquid or an amorphous polymer
consists of three terms:
410 7 Transport Properties of Polymers

Vb ¼ Vbð0Þ þ VbFI þ VbFH ð7:122Þ

where Vbð0Þ is the equilibrium liquid volume at zero absolute temperature and VbFI is the interstitial
free volume, which is assumed not to be rearranged in space without any system energy change. This
term, therefore, does not contribute to solute diffusion. VbFH is the free volume holes in the system. In a
solute–polymer system where the cohesive energy density is low, the solute diffusion coefficient can
be defined as:
∗
!
γ Vb1
D1 ¼ D01 exp ð7:123Þ
VbFH

where D01 is a temperature-independent constant (assuming that the critical energy that a solute has to
overcome to avoid the attractive forces that holds it to its neighbours is negligible). γ is an overlap
factor which accounts for the free volume hole being accessible to more than one jumping unit and is
∗ ∗
between 0.5 and 1. Vb1 and Vb2 are the specific critical hole free volumes needed for the solute and the
polymer to make a diffusive jump. It is assumed that VbFH can be redistributed without an energy
change in the system. The final expression for the self-diffusion (Eq. 7.124) contains an important
parameter, ξ, which is the ratio in the size of the solute molecule and the size of a ‘jumping’ polymer
segment:
0  ∗ ∗ 1
γ w1 Vb1 þ ξw2 Vb2
D1 ¼ D01 exp @ A ð7:124Þ
VbFH

where wi is mass fraction of component i. The challenge with this model, and partly also the CTF
model, is that it requires several input parameters that are not always available. Those include the zero
concentration diffusivity and its temperature dependence, rheology data, values on the interaction
parameter and the density of the solute and polymer as a function of temperature.
Thornton et al. (2009) compiled a large set of permeation and diffusion data for gases in both
rubbery and glassy polymers with a wide range in fractional free volume (5–35%) and permeability
ranging from 0.1 to 9700 Barrer. They found that the following equation (Eq. 7.125), with two
empirical constants (here denoted αP and βP), could describe the permeability (P) data as a function of
the fractional free volume ( f ). Whereas αP increased in a strongly non-linear way with decreasing
solute size, β increased linearly with increasing solute size (Fig. 7.29a):
P ¼ αP exp ðβP f Þ ð7:125Þ

It was also shown that the diffusivity could be described with a similar equation, although the data
of the smallest molecules deviated from the trend, which was suggested to be due to the amount of
data available for these being less than for the larger molecules (Fig. 7.29b):
D ¼ αD exp ðβD f Þ ð7:126Þ

The available/accessible (fractional) free volume in the polymer for a given solute can be
estimated in different ways. It can, for example, be estimated from the specific volume Bondi
group contribution method (Bondi 1964) where, however, the fact that the available free volume
can be different for different solutes is not considered. To consider this, a new approach was
developed (Park and Paul 1997; Greenfield and Theodorou 1993); the available fractional free
7.6 Transport Properties of Elastomers and Melts 411

Fig. 7.29 (a) αP and βP as a function of the square of the solute kinetic diameter. (b) The same but for αD and βD.
(Drawn after data of Thornton et al. (2009))

volume for a specific solute (n) can be obtained from the polymer total free volume (v) and the
occupied volume (v0), where the latter is different for different solutes:
 
f n ¼ v  ðv0 Þn =v ð7:127Þ

The occupied volume can be calculated based on the different groups the polymer repeating unit
consists of (K is the number of the different groups), where each group has a specific van der Waals
volume (vw) and an empirically determined gas–polymer group interaction parameter (γ nk):

X
K
ðv0 Þn ¼ γ nk ðνw Þn ð7:128Þ
k¼1

Along the same lines, Nilsson et al. (2013) showed, using polynomial functions, that it was
possible to obtain αP and βP for a large range of gas molecules as a function of solute kinetic
diameter. With this, it was possible to predict N2 diffusion in glassy poly(ether-ether-ketone) and
polycarbonate with Eq. (7.126) without using any adjustable parameter.
Rather than using the overall free volume, the concept shell-like free volume (vf) has been
introduced (Ohashi, Ito and Yamaguchi 2010). It is considered here that the free volume is a
consequence of the volume that arises when the molecules are constantly hitting each other. This
volume can be calculated based on the surface area of the solute (s) and the free volume thickness (δ)
(Fig. 7.30). The solute diffusivity is obtained as:

D ¼ D0 exp vp =v f ð7:129Þ

where Do is a pre-exponential factor and vp is the solute core volume. In a multicomponent system, the
total shell-like free volume can be obtained as the sum of the contributions from the different
‘interfaces’ made up of the solute and each of the other molecules (i):
!
X
vf ¼ s σ i δi ð7:130Þ
i

where σ i is the surface area fraction of the solute in contact with molecule i (Fig. 7.29). More details
and the later development of the free volume theories of diffusion can be found in the review by
Sharma, Tewari and Arya (2017).
412 7 Transport Properties of Polymers

Fig. 7.30 Solute molecule

in contact with two
different molecules 1 and
2. (Drawn after Ohashi
et al. (2010))

In the context of free volume, it is here of importance to question what happens when a polymer
above Tg is exposed to a gas/vapour under a high hydrostatic pressure. There are many cases when
polymers are exposed to high pressure gases/vapours. There are two ‘competing’ effects: a high
hydrostatic pressure densifies the polymer (reduces the free volume) and the sorption of the gas
plasticises the polymer (increases the free volume) (Koros, Burgess and Chen 2015). In the former
case, the diffusion and permeation decrease, and in the latter case they increase. The resulting effects
on the transport properties are determined by which of the two processes dominates. It has been
shown that this depends on the specific gas/vapour and its solubility in the polymer, as well as the
actual pressure. For example, with nitrogen in silicon rubber, the diffusivity first goes down with
increasing pressure, due to the hydrostatic effect, and then increases at higher pressure due to
plasticisation (Koros et al. 2015). For a readily condensable vapour (ethylene), the diffusivity
increases even at low pressure due to its high solubility and, consequently, high plasticisation
power in the rubber.

7.7 Transport Properties of Semicrystalline Polymers

7.7.1 Geometrical Effects of Crystals

Dense polymer crystals are obstacles for most solutes, which means that they have to circumvent the
crystals in the diffusion process (Vieth 1991). Different expressions have been used to consider the
tortuosity/labyrinth effect of these (Hedenqvist and Gedde 1996). These relate the transport properties
of the amorphous phase, where the solute is able to diffuse, to the macroscopically observed transport
properties of the semicrystalline material. The moisture permeability (P) has been observed to depend
on the permeability of the amorphous phase (Pa) and the volume fraction of this phase (va) for three
semicrystalline polymers: polyethylene (above a density of 940 kg/m3), polyamide 6,10 and poly
(ethylene terephthalate) (PET) (Losaski and Cobbs 1959):

P ¼ P a ϕa 2 ð7:131Þ

A clearly non-linear relationship between the polyethylene amorphous content and the permeabil-
ity of moisture has also been shown by Shah et al. (1998). If it is assumed that the diffusivity (D) and
solubility (S) are solute concentration-independent and depend linearly on the content of amorphous
material, the above equation can also be obtained from: D ¼ va Da, S ¼ va Sa and P ¼ DS, where Da
7.7 Transport Properties of Semicrystalline Polymers 413

and Sa are the diffusivity and solubility of the amorphous phase. It should be noted that different from
the results on moisture permeation (Losaski and Cobbs 1959), for semicrystalline materials with a
glassy amorphous phase (like PET and polylactic acid (PLA)), the gas solubility and/or diffusivity
depends on the content and properties of the rigid amorphous fraction (RAF), which varies with the
crystallinity and crystallisation history (Lin et al. 2002; Liu et al. 2004; Zekriardehani et al. 2017;
Guinault et al. 2012; Fernandes Nassar et al. 2017). This leads to relationships between the gas
permeability and the amorphous content which are different from Eq. (7.131). For a comprehensive
compilation of mass transport data in semicrystalline polymers, refer to Kanehashi et al. (2010).
The decrease in the diffusivity due to crystals can be considered to originate from two separate
effects, the geometrical impedance of the impenetrable crystals and their constraining effects on the
amorphous phase, the latter due to the amorphous chain segments being less mobile when they are
attached to crystals. These two effects had already been discussed in 1958 by Myers et al. (1958). The
former ‘tortuous’ effect can be expressed with a tortuosity factor (τ), and the constraining effect can
be expressed by the β-factor (Michaels and Bixler 1961b, Michaels, Bixler and Fein 1964). This leads
to the following relationship:
Da
D¼ ð7:132Þ
τβ

By assuming that the constraining effect is negligible when a small molecule like helium diffuses
in the amorphous phase, Michaels and Bixler (1961b) observed that the tortuosity effect decreased
with increasing amorphous content according to:
τ ¼ ϕa n ð7:133Þ

where the parameter n ranges between 1 and 1.88 for different semicrystalline polymers (Michaels
Bixler 1961b, Michaels, Vieth and Barrie 1963, Vieth and Wuerth 1969). This is a pure empirical
relationship. However, the tortuosity should depend on the aspect ratio of the crystals (w/l, where w is
width and l is thickness), and the Fricke theory (Fricke 1924) has been used to assess the geometrical
effects of the crystals on the solute diffusivity (Michaels and Bixler 1961b, Michaels, Vieth and
Barrie 1963, Michaels and Parker 1959, Neway et al. 2003, Mattozzi et al. 2007). The theory was
originally developed for describing the electrical conductivity of a system with a discrete phase in a
continuous matrix (Fricke 1924). It should be noted that when adopting the theory, the crystals are
considered as oblate spheroids, randomly arranged in an amorphous matrix. This is quite far from the
true structure in most semicrystalline polymers, except, for example, polyethylene with very low
crystallinity (Neway, Hedenqvist and Gedde 2003). The relationship between tortuosity, volume
crystallinity (vc), crystal width and thickness is:
 2
1  0:616  0:785  w=l
1
1
¼ 1 þ ϕc  ð7:134Þ
τ 2
3 0:616  0:785  w=l
1

From the experimentally determined τ for polyethylene, Michaels and Bixler (1961b) back-
calculated the aspect ratios to range between 17 and 60, which are reasonable values (Tr€ankner,
Hedenqvist and Gedde 1994).
Peterlin (1975, 1984) used a somewhat different approach to the effects of crystals:
ψ Da
D¼ ð7:135Þ
B
414 7 Transport Properties of Polymers

Fig. 7.31 The orientation

of crystal and amorphous
layers. (Drawn after
Hedenqvist and Gedde
(1996))

The diffusivity in the amorphous phase (Da) was considered to depend on the crystallinity (thereby
including constraining effects), and the tortuosity was here expressed as a detour factor (ψ). B is a
blocking factor that takes into account that amorphous interlayers can be too narrow for a solute
molecule to enter.
If the semicrystalline material is again approximated as a system of randomly oriented stacks of
crystal and amorphous layers (Fig. 7.31), upper and lower bound estimates of the solute diffusivity
can be derived using the approach by Boyd (1983), which was originally developed for the dielectric
constant. The ‘local’ diffusivity parallel (DH, horizontal) and perpendicular (DV, vertical) to the
crystal surface are, respectively:
DH ¼ ϕa Da þ ϕc Dc ð7:136Þ

and:
1 ϕ ϕ
¼ aþ c ð7:137Þ
DV Da Dc

where Dv vanishes for impermeable crystals and the parallel diffusivity becomes simply: DH ¼ va Da.
The average crystal orientation function is:
1
< cos 2 θ >¼ ð7:138Þ
3

where θ is defined in Fig. 7.31. By transformation of the local crystal/amorphous coordinates to a

specific overall coordinate system, the upper (U) and lower (L) bound diffusivities become (Pant and
Boyd 1992):
1
DU ¼ ðDV þ 2DH Þ ð7:139Þ
3

 
1 1 1 2
¼ þ ð7:140Þ
DL 3 DV DH

Because of the impermeable crystals, the difference between the lower (DL ¼ 0) and upper bound
is large, which limits the applicability of this approach:
2 2
DU ¼ DH ¼ va Da ð7:141Þ
3 3
7.7 Transport Properties of Semicrystalline Polymers 415

Fig. 7.32 Radially

growing crystals in the
spherulite. (Drawn after
Hedenqvist and Gedde
(1996))

Further, the upper bound predicts an unphysical solute diffusivity at 0% crystallinity. Boyd (1983)
went further and also derived upper and lower bound expressions for a spherulitic semicrystalline
morphology where crystals grow along the spherulite radius (Fig. 7.32). For this system, the upper
and lower bound diffusivities are:
"  0:5 #
1 4ðDV þ DH Þ
DU ¼ DH 1 þ 1 þ ð7:142Þ
2 DH

"   0:5 #
1 16DV
DL ¼ DH 1 þ 1 þ ð7:143Þ
2 DH þ DV

Again, the difference between the lower bound (Dv ¼ 0) and the upper bound is large, and the
predicted diffusivity at 0% crystallinity is unphysical:
1:24
DU ¼ ϕa Da ð7:144Þ
2

In Fig. 7.33, the two upper bounds (Eqs. (7.141) and (7.144)) are plotted as a function of the
crystallinity.
With Monte Carlo-generated random walks in computer-generated spherulite structures, it has
been possible to model the solute diffusion in a very complex geometry (Fig. 7.34) (Nilsson, Gedde
and Hedenqvist 2009). The tortuosity as a function of the volume crystallinity/amorphous phase
calculated in such an environment is given in Fig. 7.33. Note the relatively good fit to experimental
n-hexane diffusivity data (taken from n-hexane swollen polyethylenes where the constraint effect is
considered small). The simulation data could be described by Eq. (7.141), with the constants k1 and k2
equal to 1 (Fig. 7.33). Further findings about the Monte Carlo simulations and diffusivity data of
polyethylene are provided by Gedde and Hedenqvist (2019b) and in Chap. 5.
Da 1 þ k1 ð1  ϕa Þ
τ¼ ¼ ð7:145Þ
D 1  k2 ð1  ϕa Þ
416 7 Transport Properties of Polymers

Fig. 7.33 Ratio of the overall diffusivity (D) and the amorphous diffusivity (Da) as a function of the volume
crystallinity using Eqs. (7.132) and (7.134) with a crystal aspect ratio of (a) 12 (ß ¼ 1) and (b) 100 (ß ¼ 1). Curves
(c) and (d) represent upper bounds using Eq. (7.141) and (d) Eq. (7.144), respectively. The broken curve (e) is generated
by Monte Carlo simulations, and the solid curve (f) is obtained by Eq. (7.145) with the constants k1 and k2 equal to
1 (Nilsson et al. 2009). The experimental data is represented by n-hexane diffusivity (data taken for the polymer swollen
to 15 vol.% n-hexane) for polyethylene with different crystallinities (○). (Nilsson, Gedde and Hedenqvist 2009)

Fig. 7.34 The computer-generated spherulite and crystal geometry. (Drawn from illustrations in Nilsson, Gedde and
Hedenqvist (2009))

Crystallisation or changes in the semicrystalline structure during drawing/orientation of both

rubbery and glassy polymers affect the mass transport properties. It will be illustrated here for
uniaxially cold-drawn flexible semicrystalline polymers. The drawing reduces the solute diffusivity,
as illustrated by methylene chloride diffusion in high-density (linear) polyethylene and polypropylene
7.7 Transport Properties of Semicrystalline Polymers 417

Fig. 7.35 Methylene

chloride zero concentration
diffusivity as a function of
cold draw ratio for linear
PE (LPE) and isotactic
polypropylene (iPP).
(Drawn after Hedenqvist
and Gedde (1996))

(Fig. 7.35) (Peterlin 1975; Williams and Peterlin 1971; Hedenqvist and Gedde 1996). There is a rapid
decrease in diffusivity at a draw ratio of ca. 8, which is claimed to be due to the spherulite morphology
turning into a fibrous structure (Williams and Peterlin 1971). The change occurs over a larger draw
ratio interval for polypropylene due to a more gradual change in morphology in this polymer (Vittoria
et al. 1986). The change is also more gradual for crosslinked and non-crosslinked low-density
(branched) polyethylene than for high-density polyethylene, possibly also due to a more gradual
change in morphology with cold-drawing. It was observed by Webb et al. (1993) that the tortuosity
factor in polyethylene started to increase beyond a draw ratio of 6. The development of an oriented
semicrystalline fibrous structure makes the diffusivity, along with the mechanical properties,
non-isotropic (Takagi and Huttori 1965, Takagi 1965).

7.7.2 Molecular Constraints

For a semicrystalline polymer, the mobility of the amorphous chain segments is, as mentioned
previously, more constrained than in a fully amorphous analogue due to the segments being anchored
to crystals. This will have an effect on the permeability, diffusivity and solubility properties (as well
as the mechanical properties) since several repeating units of a polymer chain are involved in the
solute jumping process.
As mentioned above, the size of the effects of molecular constraints on the diffusivity can be
described by the size of the immobilisation factor (ß) (Eq. 7.132). Michaels and Bixler (1961b)
calculated the ß-factor for solute diffusion in polyethylene. It was observed that ß increased with the
size of the solute, which is understandable considering that the size of polymer segments that are
involved in the jumping process of a larger solute is larger. The constraining effects affect the size of
the fractional free volume obtained from diffusion data (e.g. with Eq. 7.117). The fractional free
volume of a rubbery semicrystalline polymer like polyethylene can be divided into two to three
components, that of the pure amorphous phase (equivalent to that of a 100% amorphous material) and
that of the amorphous crystal interphase (in turn sometimes divided into liquid-like and crystal core-
like interphases) (Hedenqvist and Gedde 1996). The fractional free volume is highest for the pure
amorphous phase, and the total free volume decreases with increasing crystal content (where the free
volume is zero), due to the interphase component with a lower free volume increasing relative to the
pure amorphous component, as observed for n-hexane diffusivity in polyethylene (Fig. 7.36). The
fractional free volume dependence on the crystallinity is different for semicrystalline glassy polymers
(PET and PLA). The rigid amorphous fraction (RAF) is more constrained than the mobile amorphous
phase (MAF), for the same reasons as the amorphous–crystal interphase component is more
418 7 Transport Properties of Polymers

Fig. 7.36 The free volume fraction of the amorphous component of two different polyethylenes as a function of
volume fraction of crystal core. (Drawn after data of Mattozzi et al. (2005))

constrained than the pure amorphous component in rubbery semicrystalline polymers. However, the
RAF has a higher free volume than MAF, due to poorer packing in the former case. Hence, the free
volume change with crystallinity may have more complex behaviour than that observed in Fig. 7.36
(Zekriardehani et al. 2017). This is due to the relative content of the RAF, and its fractional free
volume, varying with the crystallinity, in addition to the specific production/crystallisation process
used (Lin, Shenugin and Nazarenko 2002; Liu et al. 2004; Zekriardehani et al. 2017; Guinault et al.
2012; Fernandes Nassar et al. 2017).

7.8 Concentration-Dependent Diffusivity and Swelling in Flexible Polymers

The diffusivity is normally a function of solute concentration. In most cases, it increases with
increasing solute concentration due to plasticisation effects. At even higher solute concentrations,
concentration-dependent diffusion is accompanied by swelling-induced mechanical stresses. A way
to include solute concentration dependence is to consider the diffusivity to be exponentially depen-
dent on the solute concentration:
D ¼ Dc0 exp ðγCÞ ð7:146Þ

where Dc0 is the diffusivity at zero penetrant concentration and γ is a constant (Masclaux et al.
2010). A case for an exponential relationship can be argued based on the free volume theory (see
above). Alternatively, the exponent may be written (αa1) where (a1) is the activity of the penetrant
and α is a constant (Rogers, Stannett and Szwarc 1960). The increase in D with increasing concen-
tration can be several orders of magnitude. Kokes and Long (1953) showed, on the basis of diffusion
of different organic solutes in poly(vinyl acetate), that γ is proportional to the Flory–Huggins
interaction parameter. The general trend, although not always obeyed, is that the γ parameter
increases with increasing crystallinity and decreasing temperature (Kokes and Long 1953). These
two trends show that γ is related to the increase in segmental mobility of the polymer caused by the
presence of penetrant molecules.
7.8 Concentration-Dependent Diffusivity and Swelling in Flexible Polymers 419

Fig. 7.37 Normalised 1

Normalized mass increase

n-hexane mass increase in a
low-density polyethylene 0.8
sheet at ambient
temperature. The solid
0.6
curve represents a fit to the
experimental data. (Drawn
after data of Hedenqvist 0.4
and Gedde (1999))
0.2

0
0 1 2 3 4 5 6 7
t0.5 (h)

Blackadder and Keniry (1972, 1973, 1974) proposed, on the basis of differences between diffusion
coefficients obtained by time lag, steady-state permeability and sorption–desorption measurements
for systems with solvating solutes, that these anomalous effects are due to the build-up of internal
stresses caused by the heterogeneous swelling in the initial stage of sorption. It was noted that
diffusion coefficients were higher determined from the steady state than from transient sorption–de-
sorption measurements, which in turn yield higher values than in the time lag approach. The swelling-
induced thermal stresses are decreasing as a function of time (stress relaxation) due to the viscoelastic
nature of the polymers. The Blackadder model is, however, only capable of explaining differences
between diffusivities obtained from sorption and desorption data. The result of the swelling-induced
stresses is that the solute surface concentration increases as a function of time until an equilibrium
value is reached (Hedenqvist, Krook and Gedde 2002). Several boundary conditions, e.g. Eq. (7.147),
have been proposed (Hedenqvist et al. 1993, 1996a, b). A relaxation time (τ) is used to characterise
the rate at which the equilibrium is reached. Hedenqvist et al. (2000) found a correlation between this
relaxation time and the relaxation time determined from mechanical stress relaxation measurements.
For a polyethylene sample saturated with p-xylene at 30  C, the relaxation time was reported to be
64 h (Nisizawa 1969). Eq. (7.147) has been useful in modelling, for example, n-hexane sorption in
natural rubber and polyethylene (Hedenqvist and Gedde 1999) and methyl iodide sorption in cellulose
acetate (Long and Richman 1960). Note that the last example is for a glassy polymer and the
mechanisms in that system may be different from the first two rubbery systems, although the
boundary conditions used are similar. The sorption curve of a swelling system, when plotted with a
square root of time scale, may exhibit a characteristic S-shape (Fig. 7.37). This can be due to a time-
dependent solute surface concentration that can be described with an exponential expression:
C ¼ Ci þ ðCf  Ci Þ ð1  exp ðt=τÞÞ ð7:147Þ

In this equation, it is considered that the surface material sorbs a limited amount of solute
instantaneously (Ci) and that the final equilibrium value (Cf) is reached exponentially with time.
Hedenqvist and Gedde (1999) modelled the S-shape by considering the sorption to consist of two
stages. The first stage at small times is characterised by one-dimensional swelling due to the dry
central region of the sheet being unable to swell (Fig. 7.38). Once solute molecules have reached the
central region, the sheet is free to swell in three dimensions, and this results in a steeper slope of the
sorption curve (Fig. 7.37). For a material with a small bulk modulus, for example, a rubber, the shift
from the initial stage I to the later stage II occurs gradually, and the shift in the slope of the sorption
curve is less distinct (Ritums et al. 2005).
In several cases, a ‘non-intuitive’ change in solute diffusivity is observed at higher solute activity.
This often happens in moisture transport in polar polymers and is due to clustering. At high moisture
content/relative humidity, the effects of water molecules clustering become noticeable. The clusters
420 7 Transport Properties of Polymers

Fig. 7.38 The two stages

of swelling in a Stage II
polyethylene sheet exposed
to n-hexane. Stage I and
stage II are characterised,
respectively, by
one-dimensional and three-
dimensional swelling.
(Drawn from Hedenqvist Stage I
and Gedde (1999))

Fig. 7.39 Water diffusivity versus water activity in a wheat gluten protein material at 80  C. (Drawn after data of
Angellier-Coussy et al. (2011))

move slower than individual water molecules, and the determined diffusivity decreases. Figure 7.39
shows how the water diffusivity first increases exponentially with water activity/concentration due to
plasticisation effects in the hydrophilic protein material and then decreases at higher activity due to
clustering effects starting to dominate.
The cluster size can be obtained by analysing the sorption isotherm using statistical mechanics in
combination with, for example, the Guggenheim-Anderson-de Boer (GAB) model (Zeppa et al. 2009;
Masclaux, Gouanvé and Espuche 2010). In the above system, the mean cluster consists of more than
two water molecules at a water activity of 0.9 (Alipour et al. 2019). Water clustering in, for example,
dimers can be studied more directly experimentally with infrared spectroscopy (Davis et al. 2012).

7.9 Transport Properties of Glassy Polymers

7.9.1 Introduction

It is possible to divide solute diffusion in polymers into three categories with respect to the relative
rates of diffusion and polymer molecular relaxation (reorientation) (Crank 1986). Fickian diffusion,
which is a common case in elastomers, is characterised by a higher rate of polymer relaxation
7.9 Transport Properties of Glassy Polymers 421

compared to solute diffusion. When solute diffusion is faster than the rate of polymer relaxation, the
sorption process is normally characterised by a steep front of solute molecules moving into the
interior of the polymer specimen. This is the so-called Case II diffusion, which occurs in glassy
polymers exposed to various solvents, e.g., PMMA exposed to methanol (Thomas and Windle 1980).
It is sometimes accompanied by a phase transformation. The dual solubility and dual mobility
concepts are ‘models’ used to interpret the case of comparable diffusion and relaxation rates and
the typical III/IV sorption isotherm (Fig. 7.11). This is observed for glassy polymers exposed to low
solute concentrations. Examples are CH4 in polystyrene and CO2 in PET (Vieth 1991). At higher
solute concentrations, the dual mode generally turns into what is referred to as ‘anomalous diffusion’.
It should be noted that, as also mentioned above, there are other types of sorption isotherms in glassy
systems that the dual solubility concept cannot describe. An example is ethanol sorption in a super-
glassy polymer (S-shape) (Minelli and Sarti 2020).

7.9.2 Dual Solubility Model

For molecularly dissolved solutes, their concentration in the polymer is, as mentioned previously, a
linear function of the solute partial pressure ( p) in the surrounding medium (Henry’s law):
CD ¼ kD p ð7:148Þ

where index D refers to dissolved species (Vieth 1991). This is valid, for example, for low
concentrations in flexible polymers (stage I in Fig. 7.11). At higher concentrations, plasticisation of
the polymer leads to Henry’s law constant kD increasing with the partial pressure (stage II in
Fig. 7.11). Note that, for clarity, a different notation of Henry’s law constant is used here than in
Eq. (7.29).
It was observed early on that the sorption isotherms for several solute–polymer pairs did not obey
the linear relationship at small solute pressures. Instead, a small increase in solute pressure resulted in
a large increase in solute concentration (stage III). At higher pressures, the curve levelled out. This
behaviour was explained by sorption in two modes in the polymer. The first mode is the normal
molecular dissolution in the polymer according to Henry’s law, and the second mode is due to the
glassy polymer containing voids/free volume holes, i.e. regions of poor packing. The solute molecules
are trapped in these holes and adsorbed on the hole surfaces. The second mode is described by a
Langmuir isotherm. Hence, the equilibrium concentration of solutes in the case of dual solubility
consists of dissolved species (CD) and adsorbed species (CH):
C0H b p
C ¼ CD þ CH ¼ kD p þ ð7:149Þ
1 þ bp

b is the hole/void affinity constant and C0 H is the hole saturation constant. At low pressures, bp << 1,
and the sorption isotherm is linear:

C ¼ kD p þ C0H p ð7:150Þ

and at high pressures, bp >> 1, the saturation concentration, C0 H, in the holes is reached, which yields
a linear sorption isotherm again (the region between stages III and IV in Fig. 7.11):
C ¼ kD p þ C0H ð7:151Þ
422 7 Transport Properties of Polymers

At even higher pressures, plasticisation occurs and Henry’s law constant increases with pressure
which results in an upturn of the sorption isotherm (stage IV, Fig. 7.11).
Henry’s law constant follows the Arrhenius relationship:
kD ¼ kDo exp ðΔH D =RT Þ ð7:152Þ

where ΔHD is the enthalpy of dissolution. The temperature dependence of the hole affinity constant is
described in a similar way. The large solubility enthalpy, sometimes observed for glassy
polymer–solute pairs, may be explained by the hole affinity enthalpy being large. For methane in
polystyrene, it was found that ΔHD ¼ 6.3 kJ/mol and that the hole affinity enthalpy (ΔHH) was
15.1 kJ/mol (Vieth 1991). For the same polymer–solute pair, the hole saturation constant decreases
with increasing temperature.
The dual solubility concept has also been applied to predict sorption of gas mixtures. It is assumed
here that the gases compete in occupying the voids. The dissolution in the polymer matrix for each gas
is, however, independent of the presence of the other gases. The concentration of the i:th gas is:
C0Hi bi pi
C i ¼ k i pi þ ð7:153Þ
Pnt
1þ b jp j
j¼1

where nt is the total number of gases in the mixture.

7.9.3 The Dual Mobility Model

The diffusivity of solutes has to be characterised by a dual mode concept similar to that for solute
solubility (Vieth 1991). When describing this model here, it should be stressed that several models
exist that are applied to solute diffusion in the glassy state; refer to, for example, the Vrentas–Duda
free volume theory or modifications of it (Vrentas and Vrentas 1992; Sturm et al. 2019). In the dual
mode model, it is assumed that the solutes are either dissolved in the polymer matrix or adsorbed in
microvoids and that local equilibrium exists between the two modes. In the earliest models, it was
assumed that the adsorbed molecules were completely immobilised and that diffusion occurred
exclusively in the polymer matrix. It was also assumed that the ‘true’ diffusivity is constant, both
as a function of solute concentration and position in the polymer. The sorption kinetics can then be
written as:
2
∂ ∂ CD
ðCD þ CH Þ ¼ DD ð7:154Þ
∂t ∂x2

where DD is the diffusion coefficient of the dissolved solutes. CH can be replaced by considering that
the partial pressures for the two modes are the same. This leads to:
2 3
C0H b 2
∂CD 6 kD 7 ∂ CD
41 þ  25
¼ DD ð7:155Þ
∂t ∂x2
1 þ kbD CD

In later theories it was considered that the solute population in the holes had a limited mobility. By
using the fact that the driving force for diffusion is determined by the chemical potential gradient, it is
possible to also account for the mobility of the Langmuir population:
7.9 Transport Properties of Glassy Polymers 423

∂C ∂ 1 ∂μ
¼ ðDTD CD þ DTH CH Þ ð7:156Þ
∂t ∂x R T ∂x

where DTD and DTH are the thermodynamic diffusion coefficients of the dissolved and adsorbed
solutes. The partial immobilisation model relies on the concentration gradients of the two populations
being known. The flux according to Fick’s first law is written as:
∂CD ∂CH
J ¼ DD  DH ð7:157Þ
∂x ∂x

An alternative expression is:

∂C
J ¼ Deff ð7:158Þ
∂x

where:
2 32 31
C0H b DH C0H b
6 kD DD 76 kD 7
Deff ¼ DD 41 þ  25 4
1þ 25
ð7:159Þ
1 þ bkCDD 1 þ bkCDD

The drawback with the use of an effective diffusivity is that it is difficult to make a physical
interpretation of it. Koros and Paul (1978) were able to describe the CO2–PET system below Tg with
Eqs. (7.158) and (7.159). Above Tg it was sufficient to use only DD (Fig. 7.40). They attributed the
change at Tg as being due to the mobilisation of frozen-in voids when passing Tg. The ratio of the
diffusivities of the adsorbed and dissolved species below Tg was roughly 0.07.
Koros and Paul (1978) showed that the hole saturation constant was related to the unrelaxed or
excess free volume of the glass:
 
Vg  Vl ρ
C0H ¼ ð7:160Þ
Vg

where Vg is the volume of the glass at the specific temperature and Vl is the volume of the ‘liquid’
extrapolated from above Tg. ρ* is the ‘liquid-like’ molar volume of CO2. The hole saturation constant
decreases with annealing time due to volume compaction. Figure 7.41 shows that there is a clear
correlation between the hole saturation constant and the enthalpy relaxation (Chan and Paul 1980).

Fig. 7.40 Carbon dioxide

sorption isotherms in PET
at 25 (●), 65 (○) and 95  C
(■). (Drawn from Vieth
(1991))
424 7 Transport Properties of Polymers

Fig. 7.41 Size of sorption

(hole saturation constant)
versus size of enthalpy
relaxation for
polycarbonate–CO2 at
125 (●) and 135  C (○).
(Drawn from Vieth (1991))

Fig. 7.42 Schematic

curves of mass increase as a
function of time for

Mass increase
methanol in a sheet of
I
PMMA at 62 (I), 15 (II) and II
0  C (III). (Drawn after
III
data of Thomas and Windle
(1980))

Time

The dual mobility model has been extended to include acid dye diffusion in PA6 films. Sada et al.
(1987) measured a decreasing diffusivity with increasing acid dye concentration. Koros (1980)
extended the theory to include steady-state co-current permeation of a gas mixture.
The dual mode model parameters are correlated with the rigid amorphous fraction in semicrystal-
line PET (Zekriardehani et al. 2017). The occurrence of the RAF, with poorer packing/packing
‘defects’, increases the hole saturation constant C’H and decreases Henry’s law constant kd. Hence, it
appears as if the packing defects in RAF may be, at least partly, the ‘holes’ in the dual model.

7.9.4 Anomalous and Case II Diffusion

The mass increase (Mt), as well as the position of the steep solute concentration gradient, as a function
of time can be formulated as:
M t ¼ k tn ð7:161Þ

where k is a constant. n is 0.5 for Fickian diffusion and 1 for Case II diffusion (Samus and Rossi
1996). Hence, Case II diffusion is characterised by a constant increase in mass as a function of time
for the major part of a sorption process (Fig. 7.42, curve II). Anomalous diffusion is characterised by a
combination of Fickian and Case II diffusion with 0.5 < n < 1. For Case II diffusion, a sharp front of
solute moving at constant rate into the specimen is observed. Ahead of the solute front, the polymer is
in the glassy state, and behind the front the polymer, it is either in a rubbery state or still in the glassy
state. Hence, a phase transformation may accompany the sorption of the solute. Examples of
solute–polymer pairs that exhibit anomalous or Case II diffusion are alcohols in PMMA, acetone
7.9 Transport Properties of Glassy Polymers 425

and n-alkanes in polystyrene and benzene in crosslinked epoxy (Thomas and Windle (1980)).
Figure 7.42 shows the typical features of Case II and anomalous diffusion sorption curves. For
methanol-PMMA, anomalous diffusion is observed at the highest temperature, and at intermediate
temperature the system follows Case II behaviour. The sorption behaviour at the lowest temperature
is referred to as Super-Case II diffusion (curve III in Fig. 7.41). The increase in the rate of mass uptake
at longer times is attributed to the interactions of the ‘Fickian tails’ that precede the two Case II fronts
approaching the midplane of the sheet. The glass transition for the methanol-saturated PMMA system
is in the vicinity of 24  C. Hence, sorption at 62  C is accompanied by a phase transformation,
whereas sorption at 0  C occurs entirely in the glassy state.
It has been suggested that the sorption curves in Fig. 7.42 are a result of the relative propagation
rate of the front of phase transformation or plasticisation (v) and the ‘Fickian-driven’ solute diffusion
(D) (Rossi 1996). Note here that the shape of the uptake curves in Fig. 7.42 is a consequence of the
large jump in activity/concentration associated with the integral sorption experiment. If the sorption
experiment is performed in discrete small steps in solute activity, different behaviour is expected. The
propagation of plasticisation occurs at constant rate, whereas solute diffusion has the square root of
time–rate dependence. The slowest of the two processes controls the position of the sharp solute front
(Fig. 7.43). For the Fickian case, solute diffusion is the slowest process (τ1 ¼ D/v2 approaches zero),
and for Case II diffusion, the rate-limiting step is the propagation of plasticisation (τ1 approaches
infinity) (Fig. 7.43).
Schematic Case II solute concentration profiles in a sheet are drawn in Fig. 7.44. When solute
diffusion is rapid behind the front, i.e. when the front propagation is accompanied by a phase
transformation, the concentration gradient behind the front is small. In this case, the diffusion kinetics
may be described by a step function where, at the front, the diffusivity changes from D0 to D1

Fig. 7.43 Competing

mechanisms involved in
the propagation of the front
of solute molecules. x(t) is
the distance covered by the
front. τ1 determines the
shift from plasticisation-
controlled to diffusion-
controlled solute front
movement. Lines 1–2 refer
to plasticisation kinetics,
and lines 3–4 refer to
solvent transport. (Drawn
after Rossi (1996))

Fig. 7.44 Typical

concentration profiles in a
polymer sheet during
Concentration

sorption in the case of Case

II or anomalous diffusion

Distance
426 7 Transport Properties of Polymers

(D1 >> D0) (Samus and Rossi 1996). Hence, solute molecules diffuse rapidly to the front where they
are slowed down. Subsequently, they move with the front at a rate determined by the rate of
propagation of the phase transformation.
This implies that phase transformation occurs at a specific solute concentration C01, i.e. where Tg
occurs. The diffusivity can, therefore, be written as:
D1  D0
DðCÞ ¼ D0 þ ð7:162Þ
1 þ exp ðLðC  C01 ÞÞ

where L determines the spatial size of the glass transition region. An alternative approach to the
‘constant plasticisation-rate’ approach is the use of the Deborah number:
τrel τ
De ¼ ¼ xrel ð7:163Þ
τdif 2
D

where x is the sheet thickness and τrel and τdif are the characteristic polymer segment relaxation time
and diffusion time, respectively (Camera-Roda and Sarti 1990). A Deborah number close to unity
corresponds to Case II or anomalous diffusion. Criticism has been put forwards to the use of the
Deborah number due to the unrealistic implicit thickness dependence (Samus and Rossi 1996).
Finally, it should be mentioned that swelling-induced mechanical stresses also occur for glassy
polymers, and the solute may cause extensive swelling of the polymer or even crazing and premature
failure due to the pressure exerted by the solute on the surrounding polymer matrix (Samus and Rossi
1996). In case of extensive swelling, the sorption curve becomes S-shaped, and the stage I/II
behaviour illustrated previously for natural rubber and polyethylene also occurs here (Fig. 7.45).
The regions close to the surface of the sheet can only swell in one direction in contact with the solute
due to the restraint caused by the glassy interior of the sheet. It is not until solute reaches the middle
plane of the sheet that the polymer can swell freely in three dimensions. This occurs at the position
where the slope increases in Fig. 7.45 (indicated by an arrow).
The glass transition of this EVOH resin is approximately 70  C, and the saturation concentration of
methanol at 29  C is 10.5 wt.% which, according to Fig. 7.46, indicates that the sorption of methanol
is accompanied by a phase transformation (Samus and Rossi 1996).
A comprehensive overview of the different models of Case II diffusion is presented by Bargmann
et al. (2011), and the effects of a sharp diffusion front in glassy polymers are presented by Vesely
(2008). A summary of other non-ideal features of diffusion in glassy polymers, such as antiplasti-
cisation and solute-induced conditioning effects, is described by Koros, Burgess and Chen (2015).
Extensive chemical crosslinking, as in the case of thermosets, usually improves the chemical
resistance of glassy polymers (R€omhild, Bergman and Hedenqvist 2009). Crosslinking can also

1
Normalised mass increase

0.8

0.6

Fig. 7.45 Uptake kinetics 0.4

for methanol in an EVOH
sheet at 29  C. The arrow 0.2
indicates the time when the
first solute molecules reach 0
the midplane of the sheet. 0 200 400 600 800 1000 1200
(Drawn after data of Samus
and Rossi (1996)) t 0.5 l -1 (s mm-1)
7.10 Barriers and Membranes 427

Fig. 7.46 Decrease in the 0

glass transition temperature
as a function of the mass -20
fraction of methanol in
EVOH (27–44 mol% -40
ethylene). (Drawn after

Tg
data of Samus and Rossi -60
(1996))
-80

-100
0 0.04 0.08 0.12 0.16
mass fraction methanol

improve the barrier properties by reducing molecular mobility. Examples include epoxy-based
coatings on PET bottles to improve oxygen barrier properties (Hedenqvist 2018). However, for
glassy polymers, the molecular mobility is often already low, and the effects of crosslinking are
usually less than on initially flexible polymers. For thermosets, other factors than the degree of
crosslinking are also of importance, including monomer molecular stiffness, bulkiness and polarity,
which leads to a quite large span in barrier properties for these materials (Lange et al. 2002).

7.10 Barriers and Membranes

7.10.1 Barriers

Two important applications where mass transport plays a central role are barriers and membranes
(selective barriers). These will be briefly presented below.
There are different ways to obtain high barrier solutions for polymer-based applications in, for
example, food packaging (Hedenqvist 2018). These can be divided as outlined in Fig. 7.47. It should
be noted that the different approaches are often interdependent, for example, orientation leads to
closer molecular packing and even crystallisation. The use of rigid polymer molecules leads to a
lower molecular mobility and also, in some cases, to a liquid crystalline state/order. Polymer crystals
are impermeable to most solutes, and hence a high crystallinity increases the barrier properties.
Liquid crystalline polymers are worth a separate mention here. Liquid crystals are also tight, although
their molecular order is lower than in pure crystals. Nevertheless, it is possible to obtain a higher
(liquid) crystallinity in these systems than it is with polymers with ‘true’ crystals, which makes the
former polymers among those with the best barrier properties. Nematic liquid crystalline polymers are
both very good moisture and gas barriers.
Uniaxial and biaxial molecular orientation leads to tighter packing of atoms, often to such an
extent that the material crystallises/increases its crystallinity. Even in the absence of crystallisation,
such as in cold-drawn poly(ethylene terephthalate) (PET) and poly(ethylene naphthalate) (PEN), the
oxygen solubility and diffusivity, and consequently the permeability, decrease with increasing
orientation (Liu et al. 2002). Molecular close-packing in the absence of orientation and crystallisation
also leads to improved barrier properties. This is the reason why butyl rubber is a significantly better
gas barrier than most other rubbers/elastomers. It is fortunate that butyl rubber exists; it would be
necessary to fill car tires significantly more often than we do if they consisted of only
styrene–butadiene rubber or natural rubber. PEN is a better oxygen barrier than PET because of its
higher molecular stiffness imposed by the large naphthoic units. Chemical crosslinking is yet another
428 7 Transport Properties of Polymers

Fig. 7.47 Factors to consider for obtaining high barrier properties. (Drawn after Hedenqvist (2018))

approach for improving barrier properties. The lower molecular mobility imposed by the crosslinks
reduces the permeability, especially in initially flexible polymers. Examples include crosslinked
polyethylene and highly crosslinked natural rubber (compare the rubber and ebonite). Strong second-
ary bonds, which are grouped here into ‘polar’ factors (including dipoles, hydrogen bonding and ionic
interactions), are effective in blocking gas and non-polar molecules. This is the reason why ethylene-
vinyl alcohol (EVOH) is the most frequently used gas barrier polymer in food packaging. It should be
mentioned that the ‘equal-dissolves-equal’ phenomenon affects the barrier properties. The reason for
the good moisture barrier properties of polyethylene lies in its low moisture solubility, because of its
hydrophobic properties. For the same reason, polyamide 6,6 and nitrile rubber are good barriers, at
least in relative terms, against hydrophobic substances like petroleum/oil. Instead of focusing on
modifying the polymer phase, a way of improving the barrier is to incorporate impermeable fillers
such as clay particles and graphene. Here, the 2D shape of the fillers is used to effectively block
diffusing solutes. Another way is to blend the main polymer with a ‘tighter’ polymer, as in the case of
mixing a liquid crystalline polymer with polyethylene, creating a two-phase immiscible system. By
mixing the stiffer PEN with PET, the PET barrier properties increase. The system is ‘partially’
miscible showing only one glass transition. The best barrier properties are obtained by coating and
laminating/shielding the polymer with an impermeable material. Examples include aluminisation and
glass coating of food packaging.

7.10.2 Membrane Separation

Membrane separation can be divided into liquid and gas separation. It can also be divided based on
the size of the species separated out, e.g. in microfiltration, ultrafiltration and nanofiltration. Liquid
separation includes desalination processes based on hyperfiltration, ultrafiltration, pervaporation
and membrane distillation processes (Vieth 1991). Important mechanisms for liquid separation are
osmosis and reverse osmosis. Gas separation is divided into diffusivity and solubility selective
techniques (Freeman and Pinnau 1997). This section gives a brief description of the mechanisms of
some of these techniques. More comprehensive descriptions of most aspects and techniques of
membrane separation can be found in Galizia et al. (2017), Wang et al. (2016), Muntha, Kausar
and Siddiq (2016), Peterson (1993), Nagai (2001), Paul (2004), Gupta and Sethi (2018), Rezakazemi
et al. (2018), Su and Chung (2018) and Compan and del Castillo (2018).
7.10 Barriers and Membranes 429

Fig. 7.48 Schematic

illustrations of osmosis (1), P
1. 3. S
equilibrium (2) and reverse
osmosis (3). S is the salt HO S HO
solution and ΔP is the 2 2
overpressure applied to the
salt solution to obtain
reverse osmosis. (Drawn
after Vieth (1991)) 2. π
HO S
2

7.10.2.1 Liquid Separation

Mass transport across a membrane is described using the chemical potential or the partial molar free
energy (Vieth 1991):

∂G
μi ¼ ð7:164Þ
∂ni T,P,n j

where G, ni, T and P are the Gibbs free energy, moles of component i, temperature and pressure,
respectively. The chemical potential is a function of the activity (ai):
μi ¼ μoi þ RT ln ðai Þ ð7:165Þ

where μoi is the standard chemical potential, when ai ¼ 1, and it is a function, exclusively, of
temperature and pressure. Figure 7.48 shows a two-compartment cell containing pure water on one
side and a salt solution in water on the other side of a membrane, permeable only to water. The
chemical potential of water is:
μw,P1 ¼ μow,P1 ð7:166Þ

where μow,P1 is the standard chemical potential of water evaluated at pressure P1. The chemical
potential of water on the salt solution side is lower:

μ0w,P1 ¼ μow,P1 þ RT ln a0w,P ð7:167Þ

Water will diffuse towards the direction of lower activity, i.e. from the salt-free water side (with
solute impurities) to the salt solution (osmosis or forward osmosis; freshwater will be obtained by
reconcentrating the salt solution) (Su and Chung 2018). The forward osmosis will lead to an
increasing pressure on the solution side. Equilibrium will be attained when the chemical potential
of water is equal on both sides, μ0w,P2 =μow,P1 . The pressure dependence of the chemical potential in this
case is solved from:
ð μ0  ð2
P
w,P2 ∂μw 
dμ0w ¼ dP ¼ V w dP ð7:168Þ
μ0w,P ∂P T,n
1 P1

where Vw is the partial molar volume of water. This yields:

430 7 Transport Properties of Polymers

μ0w,P2  μ0w,P1 ¼ V w ðP2  P1 Þ ð7:169Þ

which is further rearranged:


RT ln a0w,P1 ¼ V w ðP2  P1 Þ ¼ V w π ð7:170Þ

where π is the osmotic pressure.

The osmotic pressure corresponds to the overpressure on the salt solution side that has to be
applied to prevent water from diffusing into the solution side. At this instant, the system is in
equilibrium without any mass flow. If the pressure is raised even further on the solution side, water
will diffuse through the membrane into the pure water side until the water chemical potential is again
identical on both sides. This process is referred to as reverse osmosis.
In ultrafiltration, pressure or concentration gradients force the liquid through a semipermeable
membrane (with small pores), which hinders contaminants, usually bigger than in reverse osmosis, to
pass through.
Pervaporation techniques are used for selective removal of species in liquid mixtures (Vieth 1991;
Rezakazemi, Sadrzadeh and Mohammadi 2018). It is based on selective sorption in the membrane on
the feed side, selective diffusion through the membrane and desorption on the low-pressure/permeate
side. The feed components permeate through the pervaporation membrane at different rates and are,
consequently, picked up by a carrier fluid on the low-pressure side of the membrane after different
times, Alternatively, the vapours may be collected by condensation on the low-pressure side.
Pervaporation has found use in, for example, solvent dehydration, removal of organic molecules
from water, organic/organic molecule separation and removal of ethanol from water. Polymers,
polymer blends and inorganic materials are used as pervaporation membranes. Membrane distillation
is similar to pervaporation, but here a porous membrane is used, and the removal of species relies on
vapour–liquid equilibrium.

7.10.2.2 Gas Separation

The permeability ratio of a binary gas mixture (A and B) is determined by their relative ratios of
diffusivity (D) and solubility (S):
PA D A S A
αA=B ¼ ¼ ð7:171Þ
PB D B S B

Hence, gas separation, or in other words permselectivity, occurs either by diffusivity selectivity or
solubility selectivity. Diffusivity selective membranes are used to separate smaller molecules from
larger molecules (Freeman and Pinnau 1997). This includes the separation of hydrogen gas from
nitrogen gas, air separation, removal of carbon dioxide from natural gas mixtures and in refineries
where hydrogen gas is removed from hydrocarbons. Glassy polymers have a large size sieving ability
and are, therefore, used in these cases. Some glassy polymers are 100-times more permeable to
hydrogen (kinetic diameter ¼ 2.89 Å) than to methane (kinetic diameter ¼ 3.8 Å), despite the fact that
methane has a higher solubility. It seems that a low polymer chain segment mobility coupled with a
high internal free volume favours selectivity based on the size of the solute. A major practical
problem is that the permselectivity decreases with increasing permeability of the membrane. Robeson
(1991) defined a ‘theoretical’ upper bound (Fig. 7.49) for the permselectivity as a function of
permeability for binary gas mixtures by plotting the values for a large number of polymer materials.
7.10 Barriers and Membranes 431

Fig. 7.49 Permselectivity of CO2 and CH4 as a function of CO2 permeability (obtained at 25–35  C). The bold and thin
lines correspond to upper bounds for glassy polymers as of 1991 and 2008. Note that only a few polymers are shown
here; the unfilled and filled circles represent poly(trimethylsilyl propyne) and poly(phenylene oxide), and the square
represents cellulose acetate. (Drawn after Galizia et al. (2017))

Great effort has been made to increase the upper bound through developments of new membrane
materials, and the upper bound as of 2008 is shown in Fig. 7.49.
Solubility selective membranes are used for the removal of organic vapours from supercritical
gases (i.e. gases with a critical temperature above room temperature), including separation of higher
hydrocarbons from refinery hydrogen purge streams and separation of higher hydrocarbons from
natural gas. The capture of post-combustion carbon is another application of solubility selective
membranes (CO2/N2). The separation of volatile organic compounds (VOCs) from process effluent
air streams is a process of increasing importance. EPA (Environmental Protection Agency) deter-
mined that in 2014 the emission of VOC from oil and gas production in the United States alone
amounted to 3 million metric tons (Montero-Montoya, López-Vargas and Arellano-Aguilar 2018).
Since hydrocarbons are generally the minor component in the purge streams, it is necessary, for
practical purposes, that the membrane is more permeable to the hydrocarbons than to the rest of the
purge gas. If a diffusivity selective membrane is used, it would imply that the large gas stream has to
be led through the membrane. For the process to be efficient, this would require a very large
membrane surface area. It is, therefore, better to choose a membrane that is more permeable to larger
molecules, e.g. higher hydrocarbons, and less permeable to smaller molecules. This is obtained by
using solubility selective materials. To obtain solubility selectivity, the diffusivity selectivity has to
be low. Elastomers are usually solubility selective materials, partly due to their low size-sieving
ability (Fig. 7.50).
The solubility of larger more condensable molecules is higher than smaller less condensable
molecules (Fig. 7.51). This results in a rubber material that is more permeable to condensable
molecules, i.e. larger molecules, than to smaller molecules.
Poly(dimethyl siloxane) is probably the most suitable rubber material for solubility selectivity
since it is among the most permeable of all rubbers and allows for relatively high selection rates.
‘Super-glassy’ polymers, i.e. polymers with a very low degree of chain packing, have shown to be
promising solubility selective alternatives to elastomers.
Poly(1-trimethylsilyl-1-propyne) (Fig. 7.52) is a bulky super-glassy polymer with high solubility
selectivity. It is suggested that the polymer forms open channels when solidified and that these
channels may extend from surface to surface (Fig. 7.53). When the polymer is exposed to a binary
432 7 Transport Properties of Polymers

Fig. 7.50 Effect of solute size on the diffusion coefficient in a rubbery polymer (○, cis-polyisoprene; T ¼ 50  C) and a
glassy polymer (●, polyvinylchloride, T ¼ 30  C). Critical volume is used as a measure of solute size. (Drawn after data
of Freeman and Pinnau (1997))

Fig. 7.51 Effect of solute size on the solubility coefficient in cis-polyisoprene. (Drawn after data of Freeman and
Pinnau (1997))

Fig. 7.52 Poly CH3

(1-trimethylsilyl-1-
propyne) (PTSMP) H 3C CH3
Si

CH3
7.11 Techniques for Measuring Permeability, Diffusivity and Solubility 433

Fig. 7.53 Suggested

mechanisms of solubility
selectivity for PTSMP

Supercritical gas

Condensable vapour

mixture of a supercritical gas and a condensable vapour, the condensable vapour condenses in the
channels and prohibits further permeation of the smaller gas through the channels.

7.11 Techniques for Measuring Permeability, Diffusivity and Solubility

The measuring techniques of mass uptake or loss and transfer can roughly be divided into those that
involve some type of gravimetric and volume/pressure measurement and those that typically use a
detector that relies on, for example, an electric signal or a chemical reaction. In some cases, the
gravimetric technique is combined with dilation measurements to enable the volume change to be
observed as a function of the mass uptake or loss. Note that this chapter is not a complete listing of all
different techniques. There exist numerous techniques for measuring P, D and S based on sorption or
permeation. From sorption and desorption techniques, the solubility and diffusivity are obtained and
the permeability may, in many cases, be calculated from them. The solute diffusivity is obtained from
sorption/desorption data by fitting the data with a suitable equation with appropriate initial and
boundary conditions (different cases are given in Crank 1986). As an example above, refer to
Eq. 7.87. Permeability techniques, where the flow of solute through a film is measured, give the
permeability directly. If the transient flow through the film is recorded, it is possible to obtain the
diffusivity and, since P is known, S can be calculated. These techniques do not directly yield the
solute profiles within the polymer. Imaging nuclear magnetic resonance, infrared, Raman and
UV-VIS spectroscopy can however be used to monitor solute concentration profiles, and from these
profiles, the transport parameters can be calculated. There are a number of techniques that may be
used to measure self-diffusion coefficients of small molecules and polymer chains within solid or
molten polymers based on, for example, echo techniques and the determination of concentration
profiles. Below, several of the techniques will be discussed, including some which are used for
determining diffusion of large molecules/macromolecules. Additional information on spectroscopic
methods is given in Chap. 3.
The quartz crystal microbalance (QCM) is a very sensitive gravimetric technique. The mass
change is recorded through the frequency change of a vibrating quartz crystal, which is covered by the
polymer cast/melted onto it. The crystal vibrates due to an applied voltage (www.thinksrs.com)
(Fig. 7.54). It is capable of detecting mass differences of the order of ng and less (Kulchytskyy,
Koconda and Xu 2007). The fact that the specimen has to be cast onto the crystal is a drawback. As
the solute enters the specimen, the mass of the specimen/quartz complex changes and, consequently,
the resonance frequency ( fR) is shifted. From the known crystal mass–resonance frequency relation-
ship, the mass uptake is quantified as a function of time, which allows for solubility and diffusivity to
434 7 Transport Properties of Polymers

Fig. 7.54 The principle of

the quartz crystal
microbalance

Fig. 7.55 The

measurement of solubility
according to van K
Amerongen (1946)

A C

be calculated. The frequency shift (Δf ) depends, however, on several factors, and the change due to
mass uptake in the sample must be extracted from the other factors (Kulchytskyy, Koconda and Xu
2007). This technique is useful when solute solubility is very low. It is also useful to detect slow
diffusion processes since it is possible to use very thin films (submicron).
Van Amerongen (1946) had already determined gas solubilities in rubber in the 40’s by using the
technique illustrated in 7.55. A flask A, filled with fine parts of a rubber, is connected to a gas volume
meter (BC), which is sealed with mercury. At first, vacuum is obtained in chambers A and B, and then
the cock K is closed. Chamber B is then filled with the test gas to a pressure of 1 atm, as determined by
the level of the moving mercury tube in C. Cock K is opened and the gas is transferred to flask A. At
the same time, the pressure is kept constant at 1 atm by letting more gas into the system at C. After
equilibrium is established, the gas volume is monitored. The gas sorbed is then the gas volume that
has entered flask A minus the ‘non-rubber’ volume of flask A (Fig. 7.55).
The quartz spring balance technique is common for measuring sorption and desorption of vapours
(Vieth 1991). It is also used for measuring uptakes of gases at high pressures. A limited amount of
vapour, corresponding to the desired activity, is allowed to enter a pre-evacuated cylinder containing
a quartz spring with a specimen hooked onto it. As the specimen sorbs vapour, the spring is elongated.
The elongation, corrected for buoyancy effects, yields the specimen mass uptake, provided the spring
constant is known. The accuracy is of the order of 0.1–1 μg (Hedenqvist and Doghieri 2002). By
measuring the length of the spring as a function of time using, for example, a CCD camera, the vapour
diffusivity is obtained. Different vapour pressures are obtained by regulating the amount of incoming
vapour or by changing the temperature of the evaporating liquid.
Magnetic suspension balances are sensitive equipment which allow a measuring sensitivity of tens
of μg’s and a sample mass, in contrast to normal TGA’s, of up to 15 g (Blasig et al. 2007). The
advantage of a magnetic suspension balance is also that the balance is not in direct contact with the
sample and the vapour, which allows for most vapours to be used as long as they are not aggressive to
7.11 Techniques for Measuring Permeability, Diffusivity and Solubility 435

the stainless steel in the sample cell. It should be pointed out that microbalances exist that can take
larger samples (5 g) and still have a resolution on a 0.1 μg scale (hidenisochema.com).
Some techniques are based on manometric measurements of the pressure drop (pressure decay
methods) that follows from sorption of the gas into the polymer specimen (Lee et al. 2017; Davis et al.
2004; Vieth 1991). Since the polymer is contained within a system of known volume, VT, and the
mass and density of the polymer are known, it is possible to calculate the void volume, VV, and thus
also the solubility coefficient S (cm3 (STP) cm3 atm1):
V V 273 Pi  Pe
S¼ ð7:172Þ
VP T Pe

where Vp is the polymer volume, Pi is the pressure immediately after introduction of the gas and Pe is
the equilibrium pressure. Desorption kinetics are obtained on the same system by evacuating the cell
containing the solute-saturated specimen using, for example, a cold trap.
Thermogravimetry instruments are useful in monitoring sorption and desorption in polymers down
to a sensitivity of 0.1 μg. In cases where the liquid swells the polymer, it is usually satisfactory to use a
balance of 10 μg precision. Sorption is performed by immersing the sample in the liquid and
intermittently weighing it on the balance after the surface has been dried.
The sorption of a swelling vapour is performed in the same way except that the specimen is placed
above the liquid. The vapour activity is varied by blending the liquid with a component that has a
different vapour pressure. Different relative humidities are obtained by mixing water with different
salts. The diffusivity is obtained from sorption and desorption techniques by fitting the sorption/
desorption curves, i.e. the mass increase or decrease with time, with a suitable analytical or numerical
method. In dynamic vapour sorption equipment, sorption and desorption isotherms are obtained by
stepwise changes in the activity of the vapour. Different vapour activities can be obtained by, for
example, varying the gas mixture in the vapour phase. By assuming a constant diffusivity between
each incremental change in activity, the concentration dependence of the diffusivity is obtained
(Alipour et al. 2019).
In integral sorption or desorption measurements, a complete sorption or desorption experiment is
performed with a given change in the environment. An example is a dry polymer sample immersed in
a liquid. Combined instruments are now available where a TGA is coupled to a DSC and a GCMS
and, in some cases, with FTIR equipment. The mass loss is detected in the TGA, and with direct
transfer of the vapour, or through intermediate storage in vials, to the GCMS, it is possible to reveal
what has been desorbed from the polymer. Also, the pressure decay method can analyse sorption of
gas mixtures.
Permeability techniques usually involve the measurement of the flow of a gas through a film
specimen. The film surfaces are flushed with a carrier gas, e.g. nitrogen, in order to degas the
specimen from the test gas. Subsequently, one side of the film is exposed to flowing test gas. The
gas will eventually penetrate the film, and the molecules are picked up on the opposite side of the film
by the carrier gas that transports the test gas molecules to a detector. A coulometric sensor detects
oxygen, and an infrared detector is used to monitor water or CO2. By connecting the permeability cell
with a mass spectrometer, the permeability of many different compounds can be studied. For further
descriptions on how to measure gas and vapour permeation, refer to the review by Giacinti Bachetti
and Minelli (2020). From the time lag (θ), i.e. the time until a steady-state gas flow through the film is
obtained, the solute diffusivity can be calculated (Fig. 7.56) (Vieth 1991):

l2
D¼ ð7:173Þ
6θ

where l is the film thickness.

436 7 Transport Properties of Polymers

Permeated amount of gas

Fig. 7.56 The definition of
the time lag

q Time

Fig. 7.57 The transient

period of solute flow

Solute flow rate

through a film, i.e. from the
instant when the specimen
is first exposed to the test
gas on one side until a
constant linear solute
concentration gradient
through the film is
established
Time

It is possible also to obtain the diffusivity from the transient part of the solute flow rate–time curve
by fitting the following equation to experimental data (Fig. 7.57):
 
Q 4l l2
¼ exp  ð7:174Þ
Q1 ð4πDtÞ0:5 4Dt

where Q and Q1 are the solute flow rate at time t and at infinite time (Pasternak, Schimscheimer and
Heller 1970; Webb et al. 1993). The solubility can subsequently be calculated from:
Q1 l
S¼ ð7:175Þ
Dp

where p is the partial pressure of the test gas on the high gas pressure side of the specimen ( p ¼ 0 on
the low-pressure side).
A low-cost and flexible technique for vapours (often used for water) is the cup method. The weight
change of the whole cup is measured on a balance. The mass transfer (uptake or loss) is measured by
recording the weight change of the whole cup. The sample is mounted with the sides tightly clamped
to the cup, as shown in Fig. 7.58. The permeation into the cup through the film can be performed by
placing an absorbent in the compartment below the film. In the case of water transport, phosphorus
pentoxide can be used. In the case of water, if the cup is placed in a room with controlled relative
humidity, for example, 50% RH as is often the case in climate rooms, the permeation over the film
will occur between 50% RH (outside) and 0% RH (inside). A permeation experiment with an
outflow of the vapour, e.g. moisture, is obtained by placing the liquid in the compartment below the
film. A different vapour pressure in the compartment can be achieved by using a salt solution (for
water) or mixing the liquid with a miscible non-volatile component. The cup is placed in an
environment of a lower vapour activity than in the compartment to achieve the outflow permeation.
It should be noted that the measured permeability depends on the vapour pressure at the film surfaces
7.11 Techniques for Measuring Permeability, Diffusivity and Solubility 437

Fig. 7.58 The cup test representing a case with a liquid in the compartment below the clamped sample film and with
the resulting outflow permeation of the vapour

( p2 and p3). These may not be the same as those directly over the pure liquid ( p1) and in the overall
environment ( p4). For p3 and p4 to be the same, it is important that there is a gas flow over the
specimen. The stagnant conditions below the film leading to the vapour pressure being lower at the
film surface than at the liquid surface can, however, be corrected for. Gennadios, Weller and Goodong
(1994) presented a correction method for water which is based on known water diffusivity in stagnant
air.
Infrared, Raman or UV/VIS spectroscopy can be used to obtain solute concentration profiles (for
absolute concentration values the data have to be calibrated with, e.g. a gravimetric method). The
penetration of solutes having characteristic absorption bands can be followed through the polymer
specimen by cutting slices of polymer along the thickness. For this to work, the solute must have a
low vapour pressure in order not to evaporate immediately. Further, since the resolution is of the order
of μm, the method is not suited for very low diffusivities (<109 cm2/s). In confocal techniques,
e.g. confocal Raman spectroscopy, no slicing is needed; the profile is obtained by focusing down-
wards through the sample (cf. Chap. 3).
Nuclear magnetic resonance (NMR) imaging can be used to establish solute concentration profiles
in polymers. In this technique, a magnetic field gradient (G) is applied in addition to the static field Bo.
The precessional frequency of the nuclear spins then becomes a function of their position in space (r)
and is given by ω(r) where:
ωðr Þ ¼¼ γ B0 þ γ G r ð7:176Þ

where γ is the magnetogyric ratio (Chap. 3). The signal dS from a volume element dV is:
dSðG, tÞ ¼ ρðr Þ dV exp ½iγ ðB0 þ G r Þ t ð7:177Þ

where ρ(r) is the spin density. The density of spins is thus resolved as a function of time. Harding et al.
(1997) measured concentration profiles of iso-octane and olive oil in polyethylene with this tech-
nique. Samples immersed in the liquids were taken out at different times, and after removal of liquid
from the surfaces, central sections of the films were cut and immediately inserted into the NMR for
imaging. They measured solute concentration profiles across a thickness of 2 mm, and the in-plane
resolution was 39.1 μm. The technique also enabled the calculation of the concentration dependence
of diffusion, and the method was especially powerful when combined with the pulsed field gradient
NMR technique. With today’s special techniques, the spatial resolution goes all the way down to
Angstrom level (Ma and Liu 2016).
In Elastic Recoil Detection (ERD), an incident beam of particles causes nuclei to recoil from the
target (Green 1996). The yield and energy of recoils allow the calculation of the solute concentration
versus depth/distance. In Nuclear Reaction Analysis (NRA), the particles in the incident beam are
carefully selected so as to cause a nuclear reaction when they collide with the nuclei in the specimen.
The yield and energy of the nuclear products give solute concentration profiles as a function of depth/
distance. ERD and NRA have very good resolutions, typically of the order of 200800 Å, and are,
438 7 Transport Properties of Polymers

A
z
S B
z=0

Fig. 7.59 Schematic representation of the ATR-IR diffusion experiment. IR radiation from source ‘S’ is internally
reflected in a crystal to a detector ‘D’, setting up an exponentially decaying evanescent field (in the z-direction) that
samples species at the material–crystal interface. Phase A is the solute solution and phase B is the polymer film. (Drawn
from Van Alsten (1995))

therefore, well suited for recording diffusivities in the midrange and down to very low diffusivities
(10101016 cm2/s).
Real-time solute diffusion can be observed optically by, for example, a UV/VIS spectrophotome-
ter. Krongauz, Mooney and Schmeizer (1994) doped a polymer film with electron acceptor molecules
and laminated the film with a polymer layer having electron-donating side chains. As the electron
acceptor molecules diffuse into the electron-donating regions, charge transfer complexes are formed
which have a distinct optical absorption band. By recording the change in optical density, the
diffusion coefficient of the charge transfer complexes was obtained over a thickness of approximately
50 μm. Diffusivities were of the order of 108109 cm2/s.
Attenuated total reflectance infrared spectroscopy (ATR-IR, Chap. 3) can be used in the determi-
nation of the solute diffusion coefficient (Fig. 7.59) (Balik and Smedinger 1998; Van Alsten 1995). A
polymer film is cast onto the ATR-crystal, and the solute is then put in contact with the polymer film.
The measured absorbance, A, as a function of time, t, from the instant when the polymer is first
exposed to the solute, is determined by the convolution of the solute concentration profile (C(z,t)) and
the exponentially weighted molar absorptivity:
ð
1

2z
AðtÞ ¼ α exp Cðz, tÞ S dz ð7:178Þ
d
0

where α is the characteristic absorptivity, d is the penetration depth (Chap. 3, Eq. (3.4)) and S is the
interfacial area over which the measurement takes place. The integration is performed from the
crystal surface to infinity. In practice, the crystal containing a cast polymer film is mounted in an ATR
cell suitable for liquids or gases, and the experiment starts when solute enters the cell. The solute
diffusivity is calculated from the time dependence of the absorbance (Eq. 7.178). C(z,t) is calculated
by knowing that the mass flux is proportional to the solute concentration gradient (Fig. 7.59).
In a Fluorescence Recovery after Pattern Photobleaching (FRAP) experiment, a small area of a
specimen containing a labelled compound, a fluorophore, is illuminated with a weak beam of exciting
light (Green 1996). The specimen is illuminated with a strong laser beam (photobleaching beam) to
yield an irreversible bleaching of the fluorophore. Following the bleaching, the fluorescence is again
monitored by the weak beam. The initial fluorescence is weak but increases gradually as fresh
fluorescent molecules diffuse into the bleached area. The diffusion coefficient is calculated from
the rate of fluorescence intensity recovery.
7.12 Heat Transfer 439

The concept of Forced Rayleigh Scattering (FRS) is similar to FRAP (Green 1996). In FRS, a
photochromatic dye is used instead of a fluorophore. Two intersecting laser beams produce spatially
periodic regions of high and low intensity, which yield a grid in the specimen due to photoisome-
risation (a reversible process). The time dependence of the decay of intensity of an incident laser
beam, diffracted from the grating, then yields the diffusion coefficient.
In the pulsed field gradient NMR technique, a sequence of radio-frequency (RF) pulses yield a
nuclear spin echo of a specific amplitude (Fleischer 1985). If the nuclei in the region exposed to the
RF pulses are undergoing a diffusion process, the amplitude is attenuated. The amplitude is given by:
h i
A ¼ A0 exp ðγδgÞ2 DΔ ð7:179Þ

where A0 is a constant determined from the nuclear magnetic relaxation times. The size of the RF
pulse is g and its duration is δ. The pulses are separated by a time Δ. It has been possible to measure
very fast diffusivities (D) with this technique (105109 cm2/s), and later developments have
enabled the determination of diffusivities as low as 1015 cm2/s.
Phosphorescence of singlet oxygen has also been used to obtain oxygen diffusivity (Klinger et al.
2009). The electron spin resonance (ESR) technique (Chap. 3) has been used to determine oxygen
diffusivity ‘indirectly’ in nanocellulose capsule materials using oxygen-sensitive spin probes (Svagan
et al. 2016). Other techniques that can be used to assess some kind of mass transport include neutron
scattering, inverse gas chromatography and light microscopy (Vesely 2008).

7.12 Heat Transfer

7.12.1 The Heat Equations

The heat flux, i.e. the flow of heat through unit cross section, in the case of heat conduction in a plate,
is described by Fourier’s law:
∂T
q00cond ¼ k ð7:180Þ
∂x

This is similar to Fick’s first law of diffusion (Carslaw and Jaeger 1959, Crank 1986). k is the
thermal conductivity with the units, for example, of J/(s m K). The rate of thermal equilibration is
described by:
 
∂T ∂ k ∂T
¼ ð7:181Þ
∂t ∂x ρ Cp ∂x

which is similar to Fick’s second law of diffusion. ρ and Cp are the density and specific heat of the
material. The ratio:
k
α¼ ð7:182Þ
ρ Cp

is defined as the thermal diffusivity and is basically equivalent to the solute diffusivity in the diffusion
equation with the same units: m2/s (cm2/s). The thermal conductivity determines the heat flux, and the
thermal diffusivity determines the rate at which thermal equilibrium is attained. Even if the thermal
440 7 Transport Properties of Polymers

Fig. 7.60 Schematic illustration of the temperature dependence of, from top to bottom, the thermal conductivity,
density, elongational sound velocity and specific heat for amorphous polymers. (Drawn after Van Krevelen (2009))

conductivity is low, as is the case for most gases, the rate of thermal equilibration may still be high.
The thermal diffusivity of a gas is normally high (Crank 1986).
Thermal energy is transported in solids through electrons and lattice vibrations. The latter in
quantified form are referred to as phonons. The contribution from electrons to the thermal conductiv-
ity of polymers is, in most cases, low, since the electron conductivity is normally low. The phonon-
based thermal conductivity in polymers is also normally low because phonons are scattered at
‘defects’ and polymers are rich in these. Besides being low, the thermal conductivity is not vastly
different between different polymers. At room temperature, the conductivity is within  0.1 and 0.2 J/
(s m K) for amorphous polymers (including thermosets) (Van Krevelen and Te Nijenhuis 2009).
Compare this with highly conductive metals like copper (400 J/(s m K), 0  C), steel with 0.5% carbon
(~50 J/(s m K), 20  C), glass (1 J/(s m K), 25  C), water (0.6 J/(s m K), 25  C) and air (0.03 J/(s m K),
25  C) (engineeringtoolbox.com). For amorphous polymers, the temperature dependence is essen-
tially the same, with an increase in conductivity with increasing temperature in the glassy state and a
decrease with increasing temperature above the glass transition. Hence, the conductivity is highest in
the glass transition region (Fig. 7.60) (Van Krevelen and Te Nijenhuis 2009). The conductivity can be
estimated from, for example, the following equation (Van Krevelen and Te Nijenhuis 2009):
k ¼ ρ Cp uL L ð7:183Þ

where uL is the ‘longitudinal wave’ sound velocity and L can be considered as the distance between
adjacent ‘isothermal layers’ in the molecular system (Van Krevelen and Te Nijenhuis 2009).
Figure 7.60 illustrates that the conductivity curve is a function of the temperature dependence of
the density, sound velocity and specific heat assuming also that L is temperature independent
(~0.40.7 pm (picometres)) for amorphous polymers (Van Krevelen and Te Nijenhuis 2009). The
increase in conductivity with increasing temperature below Tg is considered to be mainly due to the
increased molecular mobility with increasing temperature, which reduces the sound velocity and
increases the specific heat (Anderson 1966). Above the glass transition temperature, the decrease in
density with increasing temperature appears to play a greater role than the increase in specific heat.
A more ‘open’ structure seems to reduce the phonon transport. The thermal conductivity is faster
7.12 Heat Transfer 441

Fig. 7.61 Thermal conductivity of polyethylene with densities in kg/m3 noted next to the curves (Eiermann 1964). The
bottom two curves represent high-pressure polymerised polyethylene, the middle two curves represent Ziegler
catalysed polymerised polyethylene and the top two curves represent Phillips catalysed polymerised polyethylene

through a more rigid molecular system. This is reflected in the sound velocity, which increases with
the bulk modulus of the polymer. The conductivity is also higher along stronger molecular bonds.
Hence, the conductivity is higher along the polymer chain than transverse to it. This is clearly shown
for oriented materials. The conductivity increases in the orientation direction, illustrated, for exam-
ple, by oriented loosely crosslinked rubbers (Anderson 1966). The conductivity is almost five times
the value of the unstretched rubber at a draw ratio of nearly 3 (polyisobutylene, Tantz (1961)).
Naturally, the higher the crosslink density, the lower is this effect.
In semicrystalline polymers, there is concerted activity of the atoms (in the crystalline phase)
which increases the thermal conductivity relative to that of the amorphous polymer (Fig. 7.61). Note
the changes in the curve shape with changes in crystallinity. The actual curve shape depends on the
specific polymer and is a complex function of the amorphous conductivity and the conductivity
perpendicular to the chain axis in the crystal (Choy 1977). For polyethylene, the observed curve
shapes depend on the relative sizes of these two contributions, which vary with the crystallinity. The
amorphous conductivity increases with increasing temperature in the temperature interval in
Fig. 7.61, whereas the conductivity perpendicular to the chain axis in the crystal decreases with
increasing temperature. Also, for semicrystalline polymers, the conductivity increases with orienta-
tion, and the effects seem to be stronger than for amorphous polymers. In fact, a very high thermal
conductivity (9.1 J/(K m s) at 100 K) has been observed for polyethylene extruded to an extrusion
ratio of 25 (Choy 1977).

7.12.2 Surface Boundary Heat Conditions

Heat can be transferred through conduction and convection but also through radiation. For a radiator,
the heat flux from the surface is described by the Stefan–Boltzmann’s law (Welty et al. 2001):

q00rad ¼ ε σ T s 4 ð7:184Þ
442 7 Transport Properties of Polymers

where σ is the Stefan–Boltzmann constant and Ts is the surface temperature. ε is the emissivity which
takes values between 0 and 1 and expresses the non-ideality of the radiator. Additionally, if the
radiation is not perpendicular to the surface, only a portion of the radiation is absorbed. A surface
boundary condition, generally applicable to the net radiation between a small and a large surface,
separated by a gas, is:

q00rad ¼ ε σ T s 4  T sur 4 ð7:185Þ

where Tsur is the temperature of the surrounding gas. It is assumed here that the gas has little effect on
the magnitude of radiation and that the radiation occurs perpendicular to the surface.
Heat convection occurs from a surface in contact with a fluid. The total convection is a result of
heat conduction and bulk fluid motion and is proportional to the temperature difference between the
surface (Ts) and the liquid (T1):

q00conv ¼ hðT s  T 1 Þ ð7:186Þ

where h is the heat convection transfer coefficient.

7.13 Summary

Polymers are, because of the relatively open structure, permeable, and this has implications in many
applications. It can be an advantage, as in the case of membrane separation, and a disadvantage, for
example, where gas-tight packaging is required. The transport properties are governed by the
behaviour of the solute solubility and diffusivity, and these depend on the specific solute, its size
and its chemistry, as well as on the specific polymer and its chemistry and structure. In addition, the
concentration (vapour pressure) and aggregational state (gas/vapour or liquid) of the solute
surrounding the polymer and the temperature play important roles. At low vapour pressure, the
concentration in the polymer is low, and the system is ‘ideal Fickian’; Fick’s laws are applicable in
their ‘standard form’ with a constant diffusivity, and the solubility follows Henry’s law for a polymer
above the glass transition. For glassy polymers, the system follows a dual mode mass transport where
the solute is both ‘dissolved’ in the polymer and also filling ‘voids’. At larger uptake, the diffusivity
becomes solute concentration dependent, and Henry’s law is no longer applicable. For glassy
polymers, anomalous diffusion and Case II behaviour occur where, in its most extreme case, the
sharp solute front and the associated internal stresses cause crazing. In semicrystalline polymers, the
crystals represent obstacles for the penetrating solute, giving improved barrier properties as compared
to the fully amorphous analogue. In rubbery polymers, crystals also constrain the amorphous phase,
leading to further improvement in barrier properties. On the other hand, for semicrystalline glassy
polymers, the rigid amorphous phase has different features due to poorer packing in the vicinity of the
crystals. Other factors that affect the mass transport properties include molecular orientation, chain
stiffness and polarity (of both solute and polymer). In cases when solute uptake is higher, its
interactions with the polymer can lead to several phenomena which are not observed in other types
of material (metals and ceramics), including antiplasticisation, crystallisation, extraction of species,
time/history-dependent effects, swelling and shape deformation.
7.14 Exercises 443

7.14 Exercises

7.1 Derive the well-known equation (Crank 1986) where the solute diffusivity D (considered a
constant) can be obtained from the time (t0.5) when half of the uptake until equilibrium has
been reached in a sorption experiment using a plate sample (assume one-dimensional mass
transport through the plate with thickness l ):

0:049
D¼ t0:5 ð7:187Þ
l2

7.2 The diffusivity can also be obtained from the initial slope of the sorption curve in the above case
using the following equation (Crank 1986):

2
π ∂ðMt =M1 Þ
D¼ ð7:188Þ
16 ∂ðt0:5 =lÞ

Determine, using the ‘half-time method’ and the initial slope, the diffusivity from the following
sorption curve of a solute in a plate with a thickness of 2 mm (Fig. 7.62).
7.3 Consider the common case when a membrane is exposed on the left-hand side at time zero to a
solute, yielding a constant surface concentration C1. On the right side, the permeating solute is
effectively removed, keeping the surface concentration on that side constant at C2 ¼ 0. If the
accumulated loss of solute on the right-hand side is measured as a function of time, a straight total
loss versus time curve will develop at long times when steady state is reached. By extrapolating
the straight curve down to where it crosses the time axis (at θ), the diffusivity can be obtained
from Eq. 7.173. Determine the solute diffusivity from the following graph, where the
accumulated loss of solute, per unit cross section, through a plate with thickness l of 100 μm
was measured. Derive also the equation used. The equation to start with can be found in Crank
(1986) (Fig. 7.63).
7.4 In the plot below, the zero concentration diffusivities of five substances in natural rubber are
plotted as a function of their molar volumes. They correspond to three linear alkanes (n-hexane
(C6), n-decane (C10) and tetradecane (C14)) and two non-linear alkanes (cyclohexane and
2,2-dimethylbutane). Determine what substance is behind each number in the figure (Fig. 7.64).
7.5 Diffusion of small molecules in elastomers is frequently concentration dependent. There are
various ways to deal with the concentration dependence, but the most popular methods are based

Fig. 7.62 Solute sorption

curve
444 7 Transport Properties of Polymers

Fig. 7.63 Accumulated

loss of solute as a function
of time

Fig. 7.64 Zero concentration diffusivity versus molar volume for five hydrocarbons in natural rubber. (Data from data
of Ritums et al. (2007))

on free volume concepts. It is here considered that the free space or volume in the polymer
increases with the addition of solutes since the latter are usually associated with higher free
volumes than the pure polymer. Calculate and plot solute (benzene) concentration profiles within
a sheet of poly(ethyl acrylate) at 30  C. You can use any stable intrinsic numerical integration/
iteration method (e.g. a backward Euler method). Its glass transition temperature is 22  C
(Kishimoto and Enda 1963). Plot also the diffusion coefficient as a function of the volume
fraction of solute (Eq. 7.121):

D ¼ A exp ðB=f Þ ð7:189Þ


f ¼ f Tg þ α T  T g v2 þ f 1 v1 ð7:190Þ

Indices 1 and 2 refer to the penetrant and polymer. The parameters are as follows:
7.14 Exercises 445

Thickness ¼ 0.1 cm. Initial condition: v1(x,t ¼ 0) ¼ 0, 0 x 0.1, where v1 is the volume
fraction of solute (v2 that of the polymer), x is the space coordinate and t is time. Boundary
condition: v1(x ¼ 0, x ¼ 0.1,t) ¼ 0.20. A (1 cm2/s) is a pre-exponential factor normally
proportional to temperature and varies with the size and shape of the penetrant. B ¼ 1 is a
constant proportional to the relative size of the jumping unit of the penetrant and the size of
available free volume holes. fTg (¼0.025) is the fractional free volume of the polymer at the glass
transition. α (¼4.8·104 1/ C) is the polymer thermal expansion coefficient. f1(30  C) (¼ 0.12) is
the fractional free volume of the penetrant.
7.6 Suppose the diffusivity of oxygen and argon in a fully amorphous elastomer is 2·106 and
1.5·106 cm2 s1, respectively. Calculate the diffusivity of oxygen and argon in a polyethylene
which has a volume crystallinity of 40%. The effects of the crystals present can be estimated
using Eqs. (7.132) and (7.134) assuming a random orientation of crystals with the same aspect
ratio. From TEM images, the average crystal width and thickness have been estimated to be
0.5 μm and 10 nm, respectively. The immobilisation factor ß is estimated to be 1.5 and 1.7 for
oxygen and argon.
7.7 Derive Fick’s second law for the dual solubility model. The model only considers diffusion in the
dissolved state, and it is possible to start with the expression:

2
∂C ∂ CD
ðCD þ CH Þ ¼ DD ð7:191Þ
∂t ∂x2

Assume that DD is a constant.

7.8 Diffusion in glassy polymers at low and intermediate concentrations can be described by the dual
mobility concept where it is considered that there exist two solute populations and both can diffuse.
The first population is dissolved in the polymer matrix according to Henry’s law, and the other
population is adsorbed at the walls of voids, which are considered to be present in the polymer.
Fick’s second law with an effective diffusivity can be formulated (Koros, Paul and Rocha 1976):
( )
2
∂C ∂ FK þ ð1 þ α CD Þ ∂C
¼ DD ð7:192Þ
∂t ∂x K þ ð1 þ α CD Þ2 ∂x

where C is the total concentration of solute in the sample. F ¼ DH/DD. K is the ratio: b C’H/kD. α is
the ratio: b/kD. CD, the concentration of solute dissolved in the matrix, can be calculated from
(derived from Eq. (7.149)):
 0:5
ð1 þ K  αCÞ þ ð1 þ K  αCÞ2 þ 4αC
CD ¼ ð7:193Þ
2α

Consider now a sheet of a glassy polymer which is exposed to CO2. The following data holds
for this solute–polymer system: DD ¼ 1.108 cm2/s, Co ¼ 60 cm3 (STP)/cm3 polymer (total
saturation concentration), b ¼ 0.365 1/atm, C’H ¼ 17 cm3 (STP)/cm3 polymer, kD ¼ 0.63 cm3
(STP)/(cm3 polymer atm) and the thickness ¼ 100 μm.
Calculate sorption curves, i.e. Mt/M1, as a function of the square root of time where M is mass
increase of the sample in the presence of solute. Consider the cases where the mobility (diffusiv-
ity) of the adsorbed species is zero, equal to DD and ten times DD. The three cases correspond,
respectively, to when the adsorbed species are immobilised, have the same mobility as the
dissolved species or have significantly higher mobility. Calculate also CD/C and the effective
446 7 Transport Properties of Polymers

diffusivity (DH ¼ 0 and DH ¼ 10 DD) as a function of concentration C. Present the results in

graphs and explain the trends. Again, use a stable integration/iteration method for the sorption
curves.
7.9 Solute diffusion in Case II cases can be described with Eq. (7.162). Plot D as a function of
concentration for D0 ¼ 1.109 cm2/s, D1 ¼ 1.106 cm2/s and C01 ¼ 0.5 in the interval C ¼ 0-1.
Use L ¼ 1, 10 and 20. Calculate and plot normalised mass increase (M/M1) as a function of time
for L ¼ 20, C01 ¼ 0.5 and C01 ¼ 0.9. Consider a plate with a thickness of 0.1 cm which is
exposed to a solute, leading to a surface concentration equal to 1 for all times. Use a stable
numerical integration/iteration method. Alternatively, M/M1 can be drawn schematically.
Discuss the results.
7.10 What are the main differences between the theories of Barrer and Brandt and between the
theories of DiBenedetto/Paul and Pace/Datyner?
7.11 There are numerous examples where mass transport is the limiting step in chemical reactions.
When oxidation dominates the degradation of polymers, diffusion-limited oxidation effects are
commonly observed. This occurs when the consumption of oxygen is faster than the rate at
which oxygen is resupplied from the surrounding atmosphere. This results in spatially hetero-
geneous oxidation where ‘equilibrium’ oxidation occurs at the surface and a lower degree of
oxidation occurs in the interior of the sample. Suppose the oxygen concentration profiles in a
sheet are described by a combination of Fick’s second law and the oxygen consumption rate
according to:
 
∂C ∂ ∂C C1 C
¼ D  ð7:194Þ
∂t ∂x ∂x 1 þ C2 C

where C is the oxygen concentration and C1 and C2 are constants. In the figure below,
calculated/simulated oxygen concentration profiles for a specific set of C1 and C2 values are
shown (Fig. 7.65).
Plot the oxidation rate profiles given by the second term in Eq. (7.194) as a function of time
and the accumulated amount of oxygen consumed through oxidation, i.e. the degree of oxida-
tion, as a function of time and depth from the surface. Use the data given in the figure above.
Assume that there is no limit to the extent of oxidation.

Fig. 7.65 Relative oxygen

concentration profiles
normalised to the surface
boundary concentration as
a function of time for a
sheet exposed to oxygen
calculated with Eq. (7.194).
Model parameters are:
D ¼ 1.104 cm2/s,
C1 ¼ 0.03 1/s and
C2 ¼ 100. Sheet thickness
is 1 cm. The arrow
indicates the profile
development with
increasing time
7.14 Exercises 447

Fig. 7.66 Arrangement of

clay sheets in a polymer
matrix

Fig. 7.67 The transient

period of oxygen flow
through the sheet, i.e. from
the instant when the
specimen is first exposed to
the test gas on one side
until a constant solute
concentration gradient
through the sheet has
established

7.12 Derive a model that describes the geometrical impedance to a diffusing gas molecule in a film
which contains clay sheets (Fig. 7.66).
The sheets are oriented along the film thickness in an overlapping way as shown in Fig. 7.65.
Consider only the two-dimensional case shown in the figure. The model should contain the clay
width (W ) and thickness (L ). Consider the diffusivity and the permeability in the case when the
penetrant diffusivity and solubility in the sheets are zero.
7.13 Show the analytic solution to the diffusion equation in the case of one-dimensional diffusion
through a sheet of thickness 2 l. The solution should describe the solute concentration as a
function of time and position in terms of an orthogonal Fourier series expression. Assume that
the diffusivity is constant. The sheet is initially at a uniform concentration C0 but is, at time zero,
exposed to zero solute environment on both sides, which yields the sheet surface concentration
0 and a desorption process. Plot the solute concentration profiles as a function of time.
D ¼ 1.104 cm2 s1, thickness ¼ 1 cm.
7.14 The permeability of oxygen was measured through a low-density polyethylene 0.5 mm thick
sheet. The permeability was recorded from the instant when one side of the sheet was first
exposed to oxygen until steady state was attained. Calculate the diffusivity of oxygen by fitting
the data in the figure below to Eq. (7.174) (Fig. 7.67).
7.15 The CO2 diffusivity is lower than for that of O2 in PET. Why is then the permeability of CO2
higher than that of O2 through a PET film?
7.16 Calculate the self-diffusivity of a solute ‘1’ in a polymer ‘2’ and the mutual diffusivity as a
function of solute concentration using the Vrentas–Duda free volume model at 60  C. The
polymer has a glass transition temperature (Tg2) of 25  C, and the solute glass transition
temperature is Tg1 ¼ 143  C. The densities of the solute and polymer are, respectively, 0.7
and 1.1 g/cm3 at the actual temperature. The self-diffusion coefficient can be written (Zielinski
and Duda 1992) as:
448 7 Transport Properties of Polymers

0  ∗ ∗ 1

E  w1 Vb1 þ ξ w2 Vb2
D1 ¼ D0 exp  exp @   A ð7:195Þ
RT w K11 K 21  T g1 þ T þ w2 Kγ12 K 22  T g2 þ T
1 γ1 2

and the mutual binary diffusion coefficient is:

D ¼ D1 ð1  ϕ1 Þ2 ð1  2χϕ1 Þ ð7:196Þ
∗ ∗
Vb1 and Vb2 are defined in Sect. 7.6.2.2 (refer to Eqs. (7.123) and (7.124)). They can be
estimated from the molar volumes of the solute and polymer at 0 K using group contribution
∗ ∗
methods. These are here Vb1 ¼ 0.70 and Vb2 ¼ 0.89 cm3/g. χ is the Flory–Huggins solubility
parameter and is here 0.4 at the actual temperature. γ is defined in Sect. 7.6.2.2. E is the critical
energy that a molecule has to overcome to avoid the attractive forces that hold it to its
neighbours. It can often be neglected and here it is, therefore, zero. wi is weight fraction of the
solute (1) and polymer (2). K11 and K21 are the free volume parameters for the solute, and K12 and
K22 are those for the polymer. D0 is a pre-exponential factor and is a property of the solute only
(denoted D01 when E is taken as 0).
K12/γ and K22-Tg2 can be obtained from polymer viscosity–temperature data. These are here
0.0003 cm3/(g K) and 178 K, respectively. K11/γ, K21-Tg1 and D01 can be obtained from specific
volume and viscosity data of the solute. These are here 0.0015 cm3/(g K), 60 K and 0.0006 cm2/s,
respectively. The solute moves as a single unit which means that the ratio in the size of the solute and
polymer jumping unit can be obtained as:

∗
Vb1 M1
ξ¼ ∗ ð7:197Þ
Vb M2j
2

where M1 and M2j are the molecular weight of solvent and polymer jumping unit. The polymer
jumping unit can be considered a property of the polymer and independent of the solute and can be
estimated from K12/γ and a polymer-specific proportionality constant (β), the latter available in
literature for a number of polymers. M2j seems to correlate with the polymer repeating unit but is
1.5 times bigger for several polymers. Here it is 120 g/mol (polymer repeat unit is 80 g/mol).

References

Aifantis, E. C. (1980). Acta Mechanica, 37, 265.

Alipour, N., Vinnerås, B., Gouanve, F., Espuche, E., & Hedenqvist, M. S. (2019). Polymers, 11, 287.
Anderson, D. R. (1966). Chemical Reviews, 6, 677.
Angellier-Coussy, H., Gastaldi, E., Gontard, N., & Guillard, V. (2011). Industrial Crops and Products, 33, 457.
Aziz, S. B., Woo, T. J., Kadir, M. F. Z., & Ahmed, H. M. (2018). Journal of Science: Advanced Materials and Devices,
3, 1.
Balik, C. M., & Simendinger, W. H., III. (1998). Polymer, 39, 4723.
Bargmann, S., McBride, A. T., & Steinmann, P. (2011). Transactions of the ASME, 64, 010803–010801.
Barrer, R. M. (1957). Journal of Physical Chemistry, 61, 178.
Giacinti Bachetti, M., & Minelli, M. (2020). Polymer Testing, 89, 106606.
Berens, A. R., & Hopfenberg, H. B. (1982). Journal of Membrane Science, 10, 283.
Bian, X., Kim, C., & Karniadakis, G. E. (2016). Soft Matter, 12, 6331.
Blackadder, D. A., & Keniry, J. S. (1972). Journal of Applied Polymer Science, 16, 2141.
Blackadder, D. A., & Keniry, J. S. (1973). Journal of Applied Polymer Science, 17, 351.
References 449

Blackadder, D. A., & Keniry, J. S. (1974). Journal of Applied Polymer Science, 18, 699.
Blasig, A., Tang, J., Hu, X., Shen, Y., & Radosz, M. (2007). Fluid Phase Equilibria, 256, 75.
Bondi, A. (1964). Journal of Physical Chemistry, 68, 441.
Boyd, R. H. (1983). Journal of Polymer Science, Polymer Physics Edition, 21, 505.
Brandt, W. W. (1959). Journal of Physical Chemistry, 63, 1080.
Brandt, W. W., & Anysas, G. A. (1963). Journal of Applied Polymer Science, 7, 1919.
Bueche, F. (1953). Journal of Chemical Physics, 21, 1850.
Bueche, F. (1956). Journal of Chemical Physics, 24, 418.
Camera-Roda, G., & Sarti, G. C. (1990). AICHE Journal, 36, 851.
Carslaw, H. S., & Jaeger, J. C. (1959). Conduction of heat in solids. Oxford: Clarendon Press.
Challa, V. V., & Visco, D. P., Jr. (2005). Journal of Cellular Plastics, 41, 563.
Chan, A. H., & Paul, D. R. (1980). Journal of Applied Polymer Science, 25, 971.
Chandrasekhar, S. (1943). Reviews of Modern Physics, 15, 1.
Choy, C. L. (1977). Polymer, 18, 984.
Cohen, M. H., & Turnbull, D. (1959). Journal of Chemical Physics, 31, 1164.
Compan, V., & del Castillo, L. F. (2018). Chapter 26, Activation entropy for diffusion of gases through mixed matrix
membranes. In S. Thomas, R. Wilson, A. Kumar, & S. C. George (Eds.), Transport properties of polymeric
membranes (p. 547). Amsterdam: Elsevier.
Crank, J. (1986). The mathematics of diffusion. Oxford: Clarendon Press.
Crank, J. & Park, G.S. (1968) Diffusion in Polymers, 1st Ed. Academic Press, London & New York.
Cussler, E. L. (1975). Chemical engineering monographs, Vol. 3: Multicomponent diffusion. New York: Elsevier.
Davis, P. K., Lundy, G. D., Palamara, J. E., Duda, J. L., & Danner, R. P. (2004). Industrial Engineering & Chemical
Research, 43, 1537.
Davis, E. M., Minelli, M., Giacinti-Baschetta, M., Sarti, G. C., & Elabd, Y. A. (2012). Macromolecules, 45, 7486.
De Angelis, M. G., & Sarti, G. C. (2011). Annual Review of Chemical and Biomolecular Engineering, 2, 97.
De Angelis, M. G., Merkel, T. C., Bondar, V. I., Freeman, B. D., Doghieri, F., & Sarti, G. C. (1999). Journal of Polymer
Science: Part: Polymer Physics, 37, 3011.
De Angelis, M. G., Sarti, G. C., & Doghieri, F. (2007). Industrial Engineering and Chemical Research, 46, 7645.
DiBenedetto, A. T., & Paul, D. R. (1964). Journal of Polymer Science, Part A, 2, 1001.
DiBenedetto, A. T., & Paul, D. R. (1965). Journal of Polymer Science, Part C, 10, 17.
DiBenedetto, A. T. (1963a). Journal of Polymer Science, Part A, 1, 3459.
DiBenedetto, A. T. (1963b). Journal of Polymer Science, Part A, 1, 3477.
Doghieri, F., De Angelis, M. G., Baschetti, M. G., & Sarti, G. C. (2006). Fluid Phase Equilibria, 241, 300.
Duda, J. L., Ni, Y. C., & Vrentas, J. S. (1979). Journal of Applied Polymer Science, 23, 947.
Edsberg, L., & Wedin, P.-Å. (1995). Optimization Methods and Software, 6, 193.
Eiermann, K. (1964). Kollold-Zeltschrift und Zeitchrift für Polymere, 201, 3.
Fang, X., & Vitrac, O. (2017). Critical Review on Food Science and Nutrition, 57, 275.
Fernandes Nassar, S., Guinault, A., Delpouve, N., Divry, V., Ducruet, V., Sollogoub, C., & Domenek, S. (2017).
Polymer, 108, 163.
Fick, A. (1855a). Poggendorff’s. Annalen der Physik und Chemie, 95, 59.
Fick, A. (1855b). Philosophical Magazine, 10, 30.
Fick, A. (1995). Journal of Membrane Science, 100, 33.
Fleischer, G. (1984). Colloid and Polymer Science, 264, 919.
Fleischer, G. (1985). Polymer, 26, 1677.
Freeman, B. D., & Pinnau, I. (1997). Trends in Polymer Science, 5, 167.
Fricke, H. (1924). Physical Review, 24, 575.
Fujita, H., Kishimoto, A., & Matsumoto, K. (1960). Transactions of the Faraday Society, 56, 424.
Galindo, A., Davies, L. A., Gil-Villegas, A., & Jackson, G. (1998). Molecular Physics, 93, 241.
Galizia, M., Seok Chi, W., Smith, Z. P., Merkel, T. C., Baker, R. W., & Freeman, B. D. (2017). Macromolecules, 50,
7809.
Gedde, U. W., & Hedenqvist, M. S. (2019a). Polymer solutions (Chap. 4) In Fudamental Polymer Science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Morphology of semicrystalline polymers. In Fundamental Polymer
Science (Chap. 7) Cham: Springer Nature Switzerland AG.
Geise, G. M., Paul, D. R., & Freeman, B. D. (2014). Progress in Polymer Science, 39, 1.
Gennadios, A., Weller, C. L., & Gooding, C. H. (1994). Journal of Food Engineering, 21, 395.
Giacinti Baschetti, M., Doghieri, F., & Sarti, G. C. (2001). Industrial Engineering and Chemical Research, 40, 3027.
Gil-Villegas, A., Galindo, A., Whitehead, P. J., Mills, S. J., Jackson, G., & Burgess, A. N. (1997). Journal of Chemical
Physics, 106, 4168.
450 7 Transport Properties of Polymers

Green, P. F. (1996). Translational dynamics of macromolecules in melts. In P. Neogi (Ed.), Diffusion in polymers.
New York: Marcel Dekker.
Greenberg, M. D. (1988). Advanced engineering mathematics. Englewood Cliffs: Prentice-Hall.
Greenfield, M. L., & Theodorou, D. N. (1993). Macromolecules, 26, 5461.
Gross, J., & Sadowski, G. (2001). Industrial Engineering and Chemical Research, 40, 1244.
Grulke, E. A. (1989). Polymer handbook (J. Brandrup, & E. H. Immergut, Eds., p. 519). New York: Wiley.
Guinault, A., Sollogoub, C., Ducruet, V., & Domenek, S. (2012). European Polymer Journal, 48, 779.
Gupta, V. K., & Sethi, B. S. (2018). Chapter 12, Liquid-liquid separation through polymeric membranes. In S. Thomas,
R. Wilson, A. Kumar, & S. C. George (Eds.), Transport properties of polymeric membranes (p. 217). Amsterdam:
Elsevier.
Han, J., & Boyd, R. H. (1996). Polymer, 37, 1797.
Hansen, C. M. (2007). Solubility parameters: A User’s handbook (2nd ed.). Boca Raton: CRC Press.
Harding, S. G., Johns, M. L., Pugh, S. R., Fryer, P. J., & Gladden, L. F. (1997). Food Additives and Contaminants, 14,
583.
Hedenqvist, M. S. (2018). Barrier packaging materials. Chapter 26. In M. Kutz (Ed.), Environmental degradation of
materials (3rd ed., pp. 559–581). Oxford: Elsevier/William Andrew Publ.
Hedenqvist, M. S., & Doghieri, F. (2002). Polymer, 43, 223.
Hedenqvist, M. S., Johnsson, G., Tr€ankner, T., & Gedde, U. W. (1996a). Polymer Engineering and Science, 36, 271.
Hedenqvist, M. S., Krook, M., & Gedde, U. W. (2002). Polymer, 43, 3061.
Hedenqvist, M. S., Yousefi, H., Malmstr€om, E., Johansson, M., Hult, A., Gedde, U. W., Trollsås, M., & Hedrick, J. L.
(2000). Polymer, 41, 1827.
Hedenqvist, M., & Gedde, U. W. (1996). Progress in Polymer Science, 21, 299.
Hedenqvist, M.S. & Gedde, U.W. (1999). Polymer, 40, 2381.
Hedenqvist, M., Tr€ankner, T., Varkalis, A., Johnsson, G., & Gedde, U. W. (1993). Thermochimica Acta, 214, 111.
Hedenqvist, M. S., Angelstok, A., Edsberg, L., Larsson, P. T., & Gedde, U. W. (1996b). Polymer, 37, 2887.
Hopfenberg, H. B., & Frisch, H. L. (1969). Journal of Polymer Science: Polymer Physics Edition, 7, 405.
Jaiswal, M., & Menon, R. (2006). Polymer International, 55, 1371.
Kahaner, D., Moler, C., & Nash, S. (1989). Numerical methods and software. Englewood Cliffs: Prentice-Hall.
Kanehashi, S., Kusakabe, A., Sato, S., & Nagai, K. (2010). Journal of Membrane Science, 365, 40.
K€argel, J., Rurhven, D., & Theodorou, D. N. (2012). Diffusion in nanoporous materials. Weinheim: Wiley-VCH.
Kimball, J. C., & Frisch, M. L. (1991). Physical Review A, 43, 1840.
Kishimoto, A., & Enda, Y. (1963). Journal of Polymer Science: Part A., 1, 1799.
Klinger, M., Poulsen Tolbud, L., Gothelf, K. V., & Ogilby, P. R. (2009). ACS Applied Materials and Interfaces, 1, 661.
Kobayashi, Y., Haraya, K., Hattori, S., & Sasuga, T. (1994). Polymer, 35, 925.
Kokes, R. J., & Long, F. A. (1953). Journal of the American Chemical Society, 75, 6142.
Koros, W. J. (1980). Journal of Polymer Science, Polymer Physics Edition, 18, 981.
Koros, W. J., Paul, D. R., & Rocha, A. A. (1976). Journal of Polymer Science, Polymer Physics Edition, 14, 687.
Koros, W. J., & Paul, D. R. (1978). Journal of Polymer Science, Polymer Physics Edition, 16, 1947.
Koros, W. J., Burgess, S. K., & Chen, Z. (2015). Encyclopedia of polymer science and technology (pp. 1–96).
New York: Wiley.
Krongauz, V. V., Mooney, W. F. R., & Schmelzer, E. (1994). Polymer, 35, 929.
Kulchytskyy, I., Kocanda, M. G., & Xu, T. (2007). Applied Physics Letters, 91, 113507.
Kulkarni, S. S., & Stern, S. A. (1983). Journal of Polymer Science: Polymer Physics Edition, 21, 441.
Kumins, C. A., & Roteman, J. (1961). Journal of Polymer Science, 1955, 699.
Lacombe, R. H., & Sanchez, I. C. (1976). Journal of Physical Chemistry, 80, 2568.
Lange, J., Nicolas, B., Galy, J., & Gerard, J.-F. (2002). Polymer, 43, 5985.
Lee, J. K., Yao, S. X., Li, G., Jun, M. B. G., & Lee, P. C. (2017). Polymer Reviews, 57, 695.
Lin, J., Shenogin, S., & Nazarenko, S. (2002). Polymer, 43, 4733.
Liu, R. Y. F., Hu, Y. S., Schiraldi, D. A., Hiltner, A., & Baer, E. (2004). Journal of Applied Polymer Science, 40, 862.
Liu, R. Y. F., Schiraldi, D. A., Hiltner, A., & Baer, E. (2002). Journal of Polymer Science: Polymer Physics Edition, 40,
862.
Long, F. A., & Richman, D. (1960). Journal of American Chemical Society, 82, 513.
Losaski, S. W., & Cobbs, W. H. (1959). Journal of Polymer Science, 36, 21.
Ma, W.-L., & Liu, R.-B. (2016). Physical Review Applied, 6, 024019.
Masclaux, C., Gouanvé, F., & Espuche, E. (2010). Journal of Membrane Science, 363, 221.
Matteucci, S., Yampolskii, Y., Freeman, B. D., & Pinnau, I. (2006). Transport of glassy and rubbery polymers. In
Y. Yampolskii, I. Pinnau, & B. D. Freeman (Eds.), Materials science of membranes for gas and vapor separation.
New York: Wiley.
Mattozzi, A., Minelli, M., Hedenqvist, M. S., & Gedde, U. W. (2007). Polymer, 48, 2453.
References 451

Mattozzi, A., Neway, B., Hedenqvist, M. S., & Gedde, U. W. (2005). Polymer, 46, 929.
McCabe, C., Galindo, A., Garcia-Lisbon, M. N., & Jackson, G. (2001). Industrial Engineering and Chemical Research,
40, 3835.
Meares, P. (1954). Journal of the American Chemical Society, 76, 3415.
Michaels, A. S., & Bixler, H. J. (1961a). Journal of Polymer Science, 50, 393.
Michaels, A. S., & Bixler, H. J. (1961b). Journal of Polymer Science, 50, 413.
Michaels, A. S., & Parker, R. B. (1959). Journal of Polymer Science, 41, 53.
Michaels, A. S., Vieth, W. R., & Barrie, J. A. (1963). Journal of Applied Physics, 34, 13.
Michaels, A. S., Bixler, H. J., & Fein, H. L. (1964). Journal of Applied Physics, 35, 3165.
Minelli, M., De Angelis, M. G., & Sarti, G. (2017). Frontiers of Chemical Science and Engineering, 11, 405.
Minelli, M., & Sarti, G. C. (2020). Industrial & Engineering Chemistry Research, 59, 341.
Montero-Montoya, R., López-Vargas, R., & Arellano-Aguilar, O. (2018). Annuals of Global Health, 84, 225.
Muntha, S. T., Kausar, A., & Siddiq, M. (2016). Polymer-Plastics Technology and Engineering, 56, 18.
Myers, A. W., Rogers, C. E., Stannett, V., & Szwarc, M. (1958). Tappi Journal, 4, 716.
Nagai, K., Masuda, T., Nakagawa, T., Freeman, B. D., & Pinnau, I. (2001). Progress in Polymer Science, 26, 721.
Neway, B., Hedenqvist, M. S., & Gedde, U. W. (2003). Polymer, 44, 4003.
Nilsson, F., Gedde, U. W., & Hedenqvist, M. S. (2009). European Polymer Journal, 45, 3409.
Nilsson, F., & Hedenqvist, M. S. (2011). Mass transport and high barrier properties of food packaging polymers,
Chapter 6. In J.-M. Lagaron (Ed.), Multifunctional and nanoreinforced polymers for food packaging (pp. 129–151).
Cambridge: Woodhead Publishing Ltd.
Nilsson, F., Hallstensson, K., Johansson, K., Umar, Z., & Hedenqvist, M. S. (2013). Industrial Engineering and
Chemical Research, 52, 8655.
Nisizawa, M. (1969). Journal of Applied Polymer Science, 13, 1621.
Ohashi, H., Ito, T., & Yamaguchi, T. (2010). Industrial and Engineering Chemical Research, 49, 11676.
Pace, R. J., & Datyner, A. (1979a). Journal of Polymer Science: Polymer Physics Edition, 17, 437.
Pace, R. J., & Datyner, A. (1979b). Journal of Polymer Science: Polymer Physics Edition, 17, 453.
Pace, R. J., & Datyner, A. (1980a). Journal of Polymer Science: Polymer Physics Edition, 18, 1103.
Pace, R. J., & Datyner, A. (1980b). Journal of Polymer Science: Polymer Physics Edition, 18, 1169.
Pant, P. V. K., & Boyd, R. (1992). Macromolecules, 25, 494.
Park, J. Y., & Paul, D. R. (1997). Journal of Membrane Science, 125, 23.
Pasternak, R. A., Schimscheimer, J. F., & Heller, J. (1970). Journal of Polymer Science, A2, 467.
Paul, D. R. (2004). Journal of Membrane Science, 241, 371.
Peterlin, A. (1975). Journal of Macromolecular Science-Physics, B11, 57.
Peterlin, A. (1984). Material Science Monographs, 21, 585.
Peterson, R. (1993). Journal of Membrane Science, 83, 81.
Pilon, L., Fedorov, A. G., & Viskanta, R. (2000). Journal of Cellular Plastics, 36, 451.
Radfarnia, H. R., Ghotbi, C., Taghikhani, V., & Kontogeorgis, G. M. (2005). Fluid Phase Equilibria, 234, 94.
Rezakazemi, M., Sadrzadeh, M., & Mohammadi, T. (2018). Chapter 13, Separation via pervaporation techniques
through polymeric membranes. In S. Thomas, R. Wilson, A. Kumar, & S. C. George (Eds.), Transport properties of
polymeric membranes (p. 243). Amsterdam: Elsevier.
Ritums, J. E., Hedenqvist, M. S., Bergman, G., Prodan, T., & Emri, I. (2005). Polymer Engineering and Science, 45,
1194.
Ritums, J. E., Neway, B., Doghieri, F., Bergman, G., Gedde, U. W., & Hedenqvist, M. S. (2007). Journal of Polymer
Science, Part B: Polymer Physics, 45, 723.
Robeson, L. M. (1991). Journal of Membrane Science, 62, 165.
Rogers, C. E., Stannett, V., & Szwarc, M. (1960). Journal of Polymer Science, 45, 61.
Rogers, C. E. (1988). Permeation of gases and vapours in polymers. In J. Comyn (Ed.), Polymer permeability (p. 11).
Barking: Elsevier.
Rolker, J., Seiler, M., Mokrushina, L., & Arlt, W. (2007). Industrial Engineering and Chemistry Research, 46, 6572.
R€omhild, S., Bergman, G., & Hedenqvist, M. S. (2009). Journal of Applied Polymer Science, 116, 1057.
Rossi, G. (1996). Trends in Polymer Science, 4, 337.
Sada, W., Kumazawa, H., Mizutani, S., & Ando, T. (1987). Journal of Applied Polymer Science, 33, 305.
Sadat-Shojaei, A., Movaghar, M. R. K., & Dehghani, S. A. M. (2015). The Canadian Journal of Chemical Engineering,
93, 1483.
Samus, M. A., & Rossi, G. (1996). Macromolecules, 29, 2275.
Sanchez, I. C., & Lacombe, R. H. (1976). Journal of Physical Chemistry, 80, 2352.
Sanchez, I. C., & Lacombe, R. H. (1977). Journal of Polymer Science Polymer Letters Edition, 15, 71.
Sanchez, I. C., & Lacombe, R. H. (1978). Macromolecules, 11, 1145.
452 7 Transport Properties of Polymers

Serna, L. V., Becker, J. L., Galdámez, J. R., Danner, R. P., & Duda, J. L. (2008). Journal of Applied Polymer Science,
107, 138.
Shah, K., Ling, M. T. K., Woo, L., Nebgen, G., Edwards, S., & Zakarija, L. (1998, September). Plastics and
Biomaterials Magazine, 52.
Sharma, J., Tewari, K., & Arya, R. K. (2017). Progress in Organic Coatings, 111, 83.
Su, J., & Chung, T.-S. (2018). Chapter 16, Membrane distillation, forward osmosis, and pressure-retarded osmosis
through polymer membranes. In S. Thomas, R. Wilson, A. Kumar, & S. C. George (Eds.), Transport properties of
polymeric membranes (p. 323). Amsterdam: Elsevier.
Sturm, D. R., Danner, R. P., Moser, J. D., & Chiu, S.-W. (2019). 47351.
Svagan, A. J., Bender Koch, C., Hedenqvist, M. S., Nilsson, F., Glasser, G., Baluschev, S., & Andersen, M. L. (2016).
Carbohydrate Polymers, 136, 292.
Takagi, Y. (1965). Journal of Applied Polymer Science, 9, 3887.
Takagi, Y., & Huttori, H. (1965). Journal of Applied Polymer Science, 9, 2167.
Tantz, H. (1961). Kolloid-Zeitschrift, 174, 128.
Tartakovsky, D., & Dentz, M. (2019). Transport in Porous Media, 130, 105.
Thomas, N. L., & Windle, A. H. (1980). Polymer, 21, 613.
Thomas, S., Wilson, R., Kumar, A., & George, S. C. (2018). Transport properties of polymeric membranes.
Amsterdam: Elsevier.
Thornton, A. W., Nairna, K. M., Hill, A. J., & Hill, J. M. (2009). Journal of Membrane Science, 338, 29.
Tihic, A., Kontogeorgis, G. M., von Solms, N., & Michelsen, M. L. (2008). Industrial Engineering and Chemical
Research, 47, 5092.
Tr€ankner, T., Hedenqvist, M., & Gedde, U. W. (1994). Polymer Engineering and Science, 34, 1581.
Vahdat, N. (1991). Journal of Applied Polymer Science, 42, 3165.
Van Alsten, J. G. (1995). Trends in Polymer Science, 3, 272.
Van Amerongen, G. J. (1946). Journal of Applied Physics, 17, 972.
Van Amerongen, G. J. (1964). Rubber Chemistry and Technology, 37, 1065.
Van Krevelen, D. W. & Te Nijenhuis, K. (2009). Properties of polymers (4th ed.). Amsterdam: Elsevier.
Venkatram, S., Kim, C., Chandrasekaran, A., & Ramprasad, R. (2019). Journal of Chemical Information & Modelling,
59, 4188.
Vesely, D. (2008). International Materials Reviews, 53, 299.
Vieth, W. R., & Wuerth, W. F. (1969). Journal of Applied Polymer Science, 13, 685.
Vieth, W. R. (1991). Diffusion in and through polymers. Munich: Hanser Verlag.
Vittoria, V., DeCandia, F., Capodanno, V., & Peterlin, A. (1986). Journal of Polymer Science, Polymer Physics Edition,
24, 1009.
Von Meerwall, E., Grigsby, J., Tomich, D., & Van Antwerp, R. (1982). Journal of Polymer Science: Polymer Physics
Edition, 20, 1037.
Von Solms, N., Michelsen, M. L., & Kontogeorgis, G. M. (2005). Industrial Engineering and Chemical Research, 44,
3330.
Vrentas, J. S., & Chu, C.-H. (1989). Journal of Polymer Science: Polymer Physics Edition, 27, 1179.
Vrentas, J. S., Duda, J. L., & Lau, M. K. (1982). Journal of Applied Polymer Science, 27, 3987.
Vrentas, J. S., Duda, J. L., & Ling, H.-C. (1989). Journal of Membrane Science, 40, 101.
Vrentas, J. S., & Vrentas, C. M. (1992). Journal of Polymer Science: Part B: Polymer Physics, 30, 1005.
Wang, L.-S. (2007). Fluid Phase Equilibria, 260, 105.
Wang, S., Li, X., Wu, H., Tian, Z., Xin, Q., He, G., Peng, D., Chen, S., Yin, Y., Jiang, Z., & Guivier, M. D. (2016).
Energy & Environmental Science, 9, 1863.
Webb, J. A., Bauer, D. I., Ward, I. M., & Carden, P. T. (1993). Journal of Polymer Science, Polymer Physics Edition,
31, 743.
Weber, A. Z., Borup, R. L., Darling, R. M., Das, P. K., Dursch, T. J., Gu, W., Harvey, D., Kusoglu, A., Litster, S.,
Mench, M. M., Mukundan, R., Owejan, J. P., Pharoah, J. G., Secanell, M., & Zenyuka, I. V. (2014). Journal of the
Electrochemical Society, 161, F1254.
Wei, Y. S., & Sadus, R. J. (2000). AICHE Journal, 46, 169.
Welty, J. R., Wicks, C. E., Wilson, R. E., & Rorrer, G. (2001). Fundamentals of momentum, heat, and mass transfer.
New York: Wiley.
Wibawa, G., & Widyastuti, A. (2009). Fluid Phase Equilibria, 285, 105.
Williams, J. L., & Peterlin, A. (1971). Journal of Polymer Science, Part A, 9, 1483.
Zekriardehani, S., Jabarin, S. A., Gidley, D. R., & Coleman, M. R. (2017). Macromolecules, 50, 2845.
Zeppa, C., Gouanvé, F., & Espuche, E. (2009). Journal of Applied Polymer Science, 112, 2044.
Zielinski, J. M., & Duda, J. L. (1992). AICHE Journal, 38, 405.
Chapter 8
Processing of Polymeric Materials

8.1 Introduction

Polymeric materials can be readily processed to complex shapes with precise dimensions with a
minimum of energy and environmental effect, which to a large extent is due to their thermal and
rheological properties. Polymers in general have a low melting temperature, which for semicrystal-
line polymers is the crystal melting point (Tm) and for amorphous polymers is the glass transition
temperature (Tg) (cf. Gedde and Hedenqvist 2019a, b). Thermoplastic processing is carried out at
temperatures between the melting temperature (Tm or Tg) and the decomposition temperature, the
latter being related to the ceiling temperature (Snow and Frey 1943). In some melt processing
methods, the hot polymer melt has the access to oxygen, and in such cases, the polymer needs to
be protected by antioxidant to avoid thermal oxidation. Most polymers are processed at 200
50  C,
which is at a much lower temperature than the processing of competing materials, such as inorganic
glass (1000–1200  C), steel (1400–1600  C), aluminium (>700  C) and brass (1000  C). Processing
of polymers is more energy and environmentally (CO2) friendly than processing of the alternative
materials (cf. Chap. 9), even though the heat capacities of polymeric materials are generally higher
than those of the competitive materials. There are three basic steps in polymer processing: heating,
shaping (involving shear and extensional flow) and cooling. Heating is usually accomplished exter-
nally from electric heaters or infrared radiation, but it occurs also by the heat released from the
viscous flow of the polymer melt and the friction between the melt and the surrounding materials. The
latter is more prevalent in certain processing methods, such as injection moulding and extrusion.
Polymer processing is either carried out as a continuous process (e.g. extrusion) or through repeated
cycles (e.g. injection moulding), and in both cases automation is possible.
The stress–strain behaviour (rheology) of polymer melts is an important knowledge part of
polymer processing. Recommended texts about polymer rheology are Rubinstein and Colby
(2003), Graessley (2004) and Gedde and Hedenqvist (2019c). The melts of commercial polymers
show viscoelastic behaviour, because they are based on melts with chain entangled molecules
(Fig. 8.1). The pristine glassy behaviour appears at extremely short time scales, <1010 s, followed
by a glass transition (midpoint at 108 s) and a rubber plateau for a high molar mass sample; the latter
are due to the presence of chain entanglements which act as crosslink junctions at these time scales.
The final drop in relaxation modules is due to a relaxation process involving reptation in a high molar
mass polymer and flow of an unentangled melt in a low molar mass polymer. In a creep experiment
(constant stress), the effect of the glass transition is revealed as the first stepwise increase in the creep
compliance, and this is followed by the rubber plateau during which the chain entanglements act as
effective junction points in the polymer network and finally the constant rate increase in J (i.e. in

# Springer Nature Switzerland AG 2021 453

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_8
454 8 Processing of Polymeric Materials

Fig. 8.1 (a) Relaxation modulus as a function of time (t, logarithmic scale) of molten polymers of different molar mass
(schematic curves). From Gedde and Hedenqvist (2019c). (b) Creep compliance (J ) of a high molar mass polymer as a
function of time (t) showing steady-state-recoverable shear compliance (Je0) and the shear viscosity (η). (From Gedde
and Hedenqvist 2019c)

Fig. 8.2 Effect of

extensional flow on the
chain orientation and the
die swell occurring on the
relaxation of melt exiting
the die. (Drawn after Gedde
and Hedenqvist 2019c)

strain at constant stress), where the slope is inversely proportional to the viscosity; during this time
period, the chains undergo reptation through their molecular tubes (cf. Gedde and Hedenqvist 2019c).
The rubber elastic response to stress of entangled polymer melts is also revealed by a phenomenon
called die swell (Fig. 8.2). If a melt is pushed through a converging channel, the linear flow rate along
the symmetry axis is constantly increasing, i.e. the melt is stretched, which is referred to as
extensional flow and the molecules align along the flow direction (Fig. 8.2). When the melt is leaving
the die and the stress is removed, the oriented molecules relax to an anisotropic state which causes a
lateral expansion of the melt, the die swell. A low molar mass, unentangled molten polymer shows no
die swell.
Molten polymers with chain entanglements are pseudoplastic, i.e. the apparent viscosity (the ratio
of the shear stress and the shear rate) decreases with increasing shear rate (Fig. 8.3a). The decrease in
apparent melt viscosity in the linear regime is by a factor 4 for a ten factor in shear rate. The data also
suggest that the shear viscosity approaches a constant value at low shear rates; this value is referred to
as the zero-shear-rate viscosity, η0. The molar mass (m) dependence of the zero-shear-rate viscosity
(η0) is very pronounced, η0 / M 3.4 (Fig. 8.3b). Note the molar mass values for the onset molar mass
for this regime, which holds for polymer melts with chain entanglements: 3800 g mol1 (PE) and
8.1 Introduction 455

Fig. 8.3 (a) Apparent melt viscosity (η) as function of shear rate at 285  C for poly(butylene terephthalate). Drawn
after data of Engberg et al. (1994). (b) The logarithm of the zero-shear-rate viscosity (η0) as a function of the logarithm
of the molar mass; schematic representation

36,000 g mol1 (PS), corresponding to 136 and 346 repeating units, respectively (Graessley 1984).
Most important, however, is the fact that commercial polymers have melts with a great many chain
entanglement showing a 2000 factorial increase in η0 for each ten-factor increase in molar mass
(slope ¼ 3.4 in Fig. 8.3b).
The implication of shear viscosity (apparent shear viscosity, η) on the flow of matter (volume flow
rate ¼ Q) through a channel with radius r and length l is described by the Poiseuille equation (also
known as the Hagen–Poiseuille equation after Jean Leonard Marie Poiseuille and Gotthilf Heinrich
Ludvig Hagen, who based the equation on experimental data; the equation was later theoretically
derived by George Stokes 1845):

πpr 4
η¼ ð8:1Þ
8lQ

where p is the pressure drop along the channel. The required pressure to push the melt through the
channel is thus given by:

η8lQ
p¼ ð8:2Þ
πr 4

The required pressure is thus proportional to three quantities, η, l and Q, and very strongly
dependent on the radius of the channel: p / 1/r4. A thin channel requires a very high pressure. The
volume flow (Q) depends on the apparent viscosity according to:

πpr 4
Q¼ ð8:3Þ
η8l

which implies that flow rate through a channel is very strongly dependent on the molar mass (M ):

1
η / M3:4 ∧Q / ) Q / M3:4 ð8:4Þ
η
456 8 Processing of Polymeric Materials

Fig. 8.4 Sketch illustrating how to solve the problem of establishing an optimum material based on the consideration of
both processing and the properties of the final solidified material

Melt flow index measurements assess the flow rate (unit: melt mass in g/10 min, which is
proportional to Q) through a standardised channel using a standardised mass (i.e. pressure p) at a
given temperature. The MFI value is thus proportional to M–3.4. A low MFI means a high molar mass.
The Poiseuille equation is generally applicable to the flow in any channel in an extruder or mould.
The strong molar mass dependence for a given set of geometry (r and l ) and pressure ( p) is evident
from Eqs. (8.3) and (8.4). In order to keep Q constant for a given geometry, the ratio p/η has to be
constant. Hence, for a change in molar mass by an order of magnitude, the required pressure has to be
increased by a factor of 2000 to keep Q constant! This makes it almost impossible to melt process
ultra-high molar mass polymers using extrusion or injection moulding.
A steady increase in molar mass leads at some point to an unacceptable high viscosity
(cf. Eqs. (8.3) and (8.4)); the polymer is not any more processable using a given processing method.
The gain obtained by increasing the molar mass in the solid-state properties (e.g. in the fracture
toughness; cf. Chap. 6) is much weaker, and the best molar mass is an intermediately high molar mass
(Fig. 8.4)
Some products can be made using different processing methods. Figure 8.5 shows a list of
advantages and disadvantages for moulding methods that can produce similar types of objects.
Injection moulding is very efficient with short cycle time and is capable of producing objects with
complex and precise shapes and dimensions including many different functions. It is also applicable
to a wide range of different thermoplastic polymers. Injection moulding utilises very high pressure
during moulding, and the drawback is the requirement of the use of very expensive equipment (both
machine and mould). A short series of products is not economically feasible. The cost for mould and
machine is low for thermoforming, but the cycle time is long. Furthermore, the method is only
applicable to relatively few polymers and the dimensions cannot be controlled to a high tolerance and
the design options are few. If the quality of the product is sufficiently good with both injection
moulding and thermoforming, injection moulding is the preferred choice for a long series, and
thermoforming is best for a short series.
Blow moulding and rotational moulding can produce hollow objects (e.g. bottles); injection
moulding can only produce two halves which in a subsequent process are welded forming a hollow
object. Rotational moulding is a low-pressure process, and the low-cost moulds are sufficient to
produce fine products. The product can be very large (containers); the drawback is the long cycle
times. Rotational moulding is thus a perfect method to make a large hollow object in a low number
(short series). Figure 8.6 shows a comparison of the cost per product for injection moulding, blow
moulding and thermoforming. Injection moulding is best for products produced in great numbers
8.1 Introduction 457

Fig. 8.5 Comparing different moulding methods

Fig. 8.6 Cost per product

sketched as a function of
the number of products
made by three different
moulding methods

(>10,000 products), whereas thermoforming is best for short series. Blow moulding can be the best
alternative for intermediate long series.
Large objects like the interior of a fridge or panels in cars are traditionally made by
thermoforming, often in ABS plastic. Injection moulding has more recently been used to make
production of such large objects more efficient (short cycle time), but very high pressures are required
and very expensive equipment (machine and mould) is required.
This chapter presents in a brief form the thermoplastic processing methods. Injection moulding is
probably the most used processing method, and it is indeed complex and versatile. Emphases on the
different design aspects of injection moulding are presented. Also continuous processes, extrusion
and calendering, are part of this chapter. 3D printing is an emerging method, which is also included.
458 8 Processing of Polymeric Materials

Methods used to produce fibres are dealt with in Gedde and Hedenqvist (2019e). A recommended
textbook on thermoset processing also including processing of polymer composites is that by
Åstr€om (1997).

8.2 Polymer Processing: A Complex Applied Polymer Science Discipline

Figure 8.7 shows a list of the fundamental sciences that are part of the applied discipline ‘polymer
processing’. Rheology is obviously fundamental, in particular in view of the complex viscoelastic
properties of molten polymers (cf. Sect. 8.1, Rubinstein and Colby 2003, Graessley (2004) and Gedde
and Hedenqvist 2019c). The rheology is partly expressed in constitutive equations and equations
expressing the structural dependence of constitutional parameters, e.g. η ¼ f (M ), which are based on
fundamental polymer physics (cf. de Gennes 1979, Doi and Edwards 1986, Rubinstein and Colby
2003, Graessley 2004 and Gedde and Hedenqvist 2019c). The temperature dependence of the
apparent shear viscosity is revealed from experimental shear stress–shear rate data taken at different
temperatures and by shifting the data along the shear rate axis, data from the different temperatures
overlap (temperature–shear rate superposition). The shift factor log aT is plotted versus reciprocal
temperature (1/T ) which allows interpolation or even extrapolation of the master curve. Details from
this procedure have been presented by Mendelson (1968). A general description of the
time–temperature superposition principle is presented in Chap. 6. Data for the temperature depen-
dence of the apparent shear viscosity at 1000 s1 are shown in Morton-Jones (1989), who have taken
them from an original publication of Barrie (1978): log η shows a linear decrease with temperature for
LDPE, a ten-factor decrease over a 200  C temperature range and for PVC, a ten-factor decrease over
a 100  C temperature range. Some processing (e.g. injection moulding) is carried out at very high
pressures. Pressure compresses the melt, and the viscosity increases with increasing pressure.
However, the effect is fairly small. The pressure effect is often expressed in a temperature equivalent;
1 MPa pressure drop corresponds approximately to 1  C temperature increase (Morton-Jones 1989).
High shear rates as during injection moulding cause a temperature rise by the viscous flow which
counteracts the high pressure effect on the viscosity, and it is a good approximation to ignore both
effects on the viscosity (Morton-Jones 1989).

Fig. 8.7 Schematic representation of the fundamental science which forms the basis of polymer processing
8.2 Polymer Processing: A Complex Applied Polymer Science Discipline 459

Fig. 8.8 Two melt fronts

are approaching, each of
them having a flow
indicated by the arrows
(fountain flow) which after
the merging results in
molecular alignment along
the central ‘cold’ knit line

Polymer physics is also relevant in another part of the processing chain, specifically the solidifica-
tion, which occurs by either crystallisation or glass formation (cf. Gedde and Hedenqvist 2019a, b).
The kinetics of crystallisation has a useful simple mathematical form in the original Avrami equation
and in the further developed versions applicable to non-isothermal conditions. The simplicity of these
equations makes them useful in moulding simulation involving also the solidification stage.
Recommended texts about this topic are Gedde and Hedenqvist (2019d), Lorenzo et al. (2007),
Ozawa (1965, 1971), Nakamura et al. (1972), Mo (2008) and Jarecki and Pecherski (2018). Another
important polymer physics topic concerns the process that occurs when two molten polymer surfaces
have established physical contact. Initially there are no molecules bridging from one side to the other
of the contact surface, but with time, molecules diffuse by reptation across the boundary and after a
time the boundary is annihilated. This process is sometimes referred to as healing, and it is relevant
for polymer processing for knit lines which are formed when two melt fronts are impinging as shown
in Fig. 8.8. The so-called fountain flow at the frontier of the propagating melts results in orientation of
the molecules in the knit lines plane with very few molecular connections across the knit line, and this
makes the knit lines weak regions. Also in extrusion of a pipe, the melt passes through the spider
holding the mandrel, and the continuous melt stream is divided into several melt streams that are
brought together after leaving the extruder die. During compression moulding the molten polymer
powder or polymer pellets are coming into physical contact by the applied pressure, and the
establishment of full mechanical strength is obtained after a time period sufficient for full healing,
i.e. molecular diffusion across all boundaries in the component. Recommended texts about healing
are Jud, Kausch and Williams (1981), Wool and Connor (1981) and Kim and Wool (1983).
The basic laws of thermodynamics (cf. Gedde (2020)), the transport property science describing
the temperature distribution in the melt during different instances based heat conduction and
convective transport are essential parts of polymer processing regarding both in the heating and
cooling phases (recommended reading is the classical text by Carslaw and Jaeger 1959; another
classical text but more recently published as a fifth edition is Lienhard 2019). Diffusion of molecules
is another relevant topic; cf. Chap. 7 and Crank (1979). Solid mechanics of polymer, anelasticity and
viscoelasticity and the ultimate properties are dealt with in Chap. 6, whereas the general solid
mechanics are treated in many textbooks, Bertram and Glüge (2015) and Lubliner and Papadopoulos
(2017) are recommended. Moulding simulation including simulation of the solidification is an
important tool for deciding which polymer should be used, positions of the gates and the general
design of the mould. This is a topic dealt with in this chapter. This kind of modelling is based on finite
element modelling algorithm, a topic which is covered in Chap. 5.
Mould shrinkage (measured by standards: ASTM D955; ISO 294–4 and ISO 2577) is caused by
the volumetric change accompanying solidification as well as the normal thermal shrinkage, and it is
460 8 Processing of Polymeric Materials

expected to be much larger for semicrystalline polymers than for glassy polymers. Expressed in the
linear shrinkage, these groups show the following values: 0.2–0.9% (glassy polymers) and 1–5%
(semicrystalline polymers); the linear shrinkage increases with increasing crystallinity (Seyler et al.
2003). By adding 30% glass fibre to glassy polymers, the linear shrinkage is reduced to only 0.1 to
0.2%, and a significant reduction is also obtained for the semicrystalline polymers: 0.3–0.8%.
Injection moulded composites show anisotropic mould shrinkage with almost no shrinkage in
major flow direction (Folkes 1982). Liquid crystalline polymers show almost no mould shrinkage
(Ticona GmbH 2001).
Data of melting points and melting enthalpies for a wide range of polymers are found in Polymer
Handbook (Brandrup, Immergut and Grulke 2003) and Wunderlich (1980). A comprehensive collec-
tion of glass transition temperature data are found in Polymer Handbook (Brandrup, Immergut and
Grulke 2003). Ceiling temperature data are also found in Polymer Handbook. Further information
about thermo-oxidative degradation is found in textbooks by Halim Hamid (2000) and
Crompton (2010).
The price of the polymer is important factor for its possible use in competition with other
polymers. Information about the current prices are provided by Plastics Information Europe (PIE),
the so-called PIE Polymer Price Index, which covers commodity polymers, engineering polymers and
specialty polymers. A rich source about various properties of polymers is the book by van Krevelen
and Te Nijenhuis (2009).

8.3 Compounding

Compounding or mixing of components precedes the final shaping methods, e.g. injection moulding
or extrusion. The aim of compounding is to mix the polymer(s) with a few or a greater number of
additives. An additive can be in a solid form (powder, fibres or flakes) or it can be a liquid.
Compounding is divided into two different mixing types: (i) Distributive mixing is essentially to
distribute different powders evenly or, if some components are liquids, to cover the particles with a
thin liquid layer. (ii) Dispersive mixing is a much more intensive mixing; the particles are introduced
into the polymer matrix, by melting the polymer and the application of significant shear forces to
disperse the particles or liquids uniformly in the polymer matrix. Figure 8.9 shows equipment aimed
for distributive mixing: drum tumbler, ribbon blender and ball mill mixer. The drum tumbler uses two
axes of rotations to mix powders. It is possible to heat the drum and to melt additives in order to cover
surfaces on a powder with the molten component. The ball mill is capable of disintegrating
agglomerates of particles for a better dispersion of the different solid particles. High-speed mixer
(Henschel mixer) is a faster and more efficient mixer of powders.

Fig. 8.9 Examples of equipment for distributive mixing

8.4 Injection Moulding 461

Fig. 8.10 Equipment for dispersive mixing

In order to obtain a uniform distribution of additives and polymers (polymer blends) on finer scale,
dispersive mixing is required involving melting of the polymer(s) and high shearing of the melt; three
such compounding equipment are sketched in Fig. 8.10. The two-roll mill has been extensively used
for rubbers. The high-shear zone is quite confined to the region between the rolls. The machine is
good in dispersive mixing but poor in distributive mixing. Mixing with a two-roll mill is slow. The
Banbury mixer is more efficient, and the chamber can be heated, although the strong shear produces
internal heat. Banbury mixer is much more efficient than the two-roll mill, 350 kg in less than 15 min,
whereas a 200 kg batch takes 2 h in two-roll mill. Extrusion is commonly used for commodity
polymers and engineering polymers. This continuous process allows degassing and drying and
excellent dispersive mixing; the distributive mixing is sometimes carried out with Henschel mixer.
In essence, the materials to be mixed are introduced in a barrel containing an auger screw or twin-
screws or two parallel screws. The material is heated and transported forwards by the pumping action
of the flight of the screws, which generate effective mixing since the polymer melts under the
influence of heat. It is possible to have zones which generate extension flow, which is very efficient
in order to break up agglomerates of nanoparticles. Volatile species including dissolved water can be
removed by having a special zone with reduced pressure equipped with an outlet. The molten
polymeric material with fillers/reinforcements is forced through a die having multiple holes that
create spaghetti-like shapes, which are eventually cooled and chopped into pellets for further
processing. Figure 8.11 shows the exit of polymeric material from an extruder die. A detailed
description of extrusion is provided in Sect. 8.5.
Recommended texts about mixing technology and compounding are Morton-Jones (1989),
Rauwendaal (1998) and Manus-Zloczower (2005).

8.4 Injection Moulding

Injection moulding is one of the most common and important processing methods for polymeric
products. Injection-moulded plastic products are ubiquitous in today’s world and can be found in
homes, offices, hospitals and industries. Some common injection-moulded products are medical
syringes, hair combs, phone casings, toys, and machinery casings. Injection moulding is a relatively
simple process wherein a thermoplastic is ‘injected’ at high pressure into a mould of particular shape
and is ejected when the part has cooled off. Injection moulding can also be used to process
thermosetting plastics and elastomers; however, contrary to thermoplastic processing, the mould is
heated to cause curing and vulcanisation, respectively (Rubin 1973). Injection moulding is attractive
because it can produce products having complex shapes with relative high speed in a process that can
462 8 Processing of Polymeric Materials

Fig. 8.11 Melt blending of polymeric materials through an extruder

Fig. 8.12 The schematic of an injection moulding machine. The polymer (in powdered/granular form) is introduced in
the hopper. Then the screw turns forcing the polymer through a heater, thereby melting it. When the entire polymer has
melted, the screw then acts as a battering ram and forces the polymer into mould, where it cools, solidifies and finally is
ejected

be automated. The process is also quite versatile, and the final products do not need extensive
finishing.
Figure 8.12 shows a sketch of an injection moulding machine. In simple terms, a typical injection
moulding machine has four parts: injection, mould-clamping, hydraulic and control unit. The
processed polymer undergoes heating and shearing to form a melt in the injection unit. The hydraulic
unit controls the injection unit. The molten polymer that is injected into the cavity of the mould is held
together in place by the mould-clamping unit. The overall moulding process can be organised and
sequenced by the control unit. The overarching process involves the introduction of the polymer, to
be processed, in granulated form through the hopper. The polymer granules then enter the barrel
having auger screws where they are heated to transform into a fluid state. The heating occurs both
8.4 Injection Moulding 463

Fig. 8.13 The several stages of injection moulding cycle

through the conduction from the inner wall of the barrel and from the viscous flow of the polymer
melt. By the action of the screw, the homogenised polymer melt is forced through a narrow nozzle at
very high pressure into a mould having lower temperature than that of the barrel. The same screw that
transported the polymer melt functions as a plunger to perform the injection, after which it moves
backwards. The mould is then clamped tightly with the transport channel being closed where the melt
solidifies by the aid of a cooling agent and pressure. Once the process is complete, the mould is
opened to eject the final product with a subsequent repetition of the same process.
The cycle of the injection moulding is illustrated in Fig. 8.13. After the closing of the mould, the
screw, now acting as a plunger, injects the polymer melt into the mould. At this point, the air that was
inside the mould cavity is expelled from the outer regions of the path of the polymer melt flow. Upon
filling of the mould cavity, a holding pressure is created by the screw, which continues to move
forward. This action introduces extra polymer melt into the cavity and mitigates the shrinkage effect
of the cooling polymer. When the gate freezes, any additional introduction of polymer melt into the
cavity is not possible. At this stage, the screw starts moving backwards to plasticise more polymers
for the next shot by collecting more polymeric granules from the hopper. The collected polymer is
carried at the front of the screw, where it awaits the next cycle of injection. During this time, the
machine can go under a pause phase because the previously injected mould is being solidified.
Finally, the mould cools to a temperature where the polymeric material will retain the desired shape.
The mould is opened to extract the final product. The mould is closed again for the commencement of
the next cycle. The process described above is for reciprocating screw injection moulding, and this
machine has several important parts that are significant in ensuring an efficient processing of
polymers. Some of these parts are described below.
The screws used in injection moulding are essentially same as the ones used in extruder, the details
of which will be provided in the subsequent section. However, the screws used in injection moulding
possess a back-flow check valve located at the screw tip. The function of the back-flow check valve is
to ‘check’ or prevent the backward motion of polymer melt over the screw flights when the screw is
acting as a plunger to perform the injection. This valve remains open when the screw is transporting
polymer melt forwards towards the mould through rotational motion. Screw length to diameter (L/D)
464 8 Processing of Polymeric Materials

ratio (commonly in the range of 15–20) is critical for having uniform polymer melt at higher output
and for having a uniform melt temperature profile.
The barrel holds the screw and is heated by electric heater through the means of conduction.
Heating of the barrel is usually done in zones (feed, rear, middle and front), and it is important to have
an optimised temperature profile to mitigate the loss of polymeric materials due to dimensional
defects, warping, degradation, etc. Some barrels can also be vented to allow moisture to escape during
processing. Excess moisture in the polymeric material is detrimental for the final product quality.
The nozzle is positioned at the end of the barrel and acts as the transition point from which the
polymer melt enters the relatively cold mould from the heated barrel. At the nozzle, the polymer
blend is heated due to internal friction as well as barrel heaters. However, when the nozzle encounters
the cold mould, heat transfer occurs, i.e. the nozzle loses temperature. When the phenomenon of this
heat transfer is excessive, the nozzle should be in retrograde to avoid the polymer melt freezing off
within the nozzle. In practice, there are different types of nozzles used in injection moulding process,
which are open, needle shut-off and external shut-off nozzles. Further details about nozzles are found
in Crawford (1998).
The mould in injection moulding is one of the most important and complicated parts. Usually the
mould has two halves, a fixed part that is attached to the stationary platen and a moving part, which is
affixed to the moving platen. The end product requirement can be complicated with complex
geometries, and the flow path of the polymeric melt can be rather long. In such cases, the polymer
melt needs to be introduced into the mould at various injection points through several channels. In
order to ensure that the polymer melt does not solidify before entering the mould cavity, the channels
are usually heated. The polymer melt should be at proper temperature and pressure at the injection
points to guarantee the required flow pattern in the mould cavity. During the injection, the mould
remains closed facilitated by a hydraulic mechanism. Usually, the size of the injection moulding
machines is determined by the extent of this closing force. After the filling of the cavity, the melt is
cooled by flowing water through the thick walls of the mould. In case of thermosetting plastics and
elastomers, instead of cooling water, these channels are used to heat the mould to cause curing and
vulcanisation. The mould is equipped with an ejection mechanism to remove the product once the
final shape of the polymer is conserved.
Injection is very efficient (short cycle times) and very little material is lost as scrap. One
disadvantage is the requirement of high pressure, which means that the mould has to be built in
high-quality materials (usually steel) and that the cost to build the mould is high. What is perhaps
most impressive is the freedom in the design possibilities, and injection moulding can produce
components with great many built-in functions. Figure 8.14 illustrates the way to make the design

Fig. 8.14 Illustration of design philosophy suitable for injection-moulded components

8.4 Injection Moulding 465

Fig. 8.15 The effect of the cooling on the internal stress distribution. (a) The same cooling from both sides causes a
symmetric stress distribution (no warpage). (b) Different cooling rates cause an asymmetric internal stress distribution
and warpage (bending)

Fig. 8.16 How to design a corner

of a complex shape. The aimed shape is shown in the left sketch. It should be noticed that the material
thickness shows extensive variation, and this cause defects during the cooling phase: valleys, bent
(warped) sections and internal cavities. The reason for the warpage is explained in Fig. 8.15; if the
cooling is different from the two sides, the internal stress distribution becomes asymmetric and the
section will be bent (warpage). The surface is cooled at a higher rate than the inner part, and this
causes a larger volumetric shrinkage of the core parts, and hence they will be stretched slightly, thus
subjected to a tensile stress; the outer parts, on the hand, are subjected to compressive stresses. The
solution to these problems is to make the design a uniform wall thickness (Fig. 8.14, right-hand
sketch). This makes the cooling more symmetric which in turn will solve the warpage problem. The
valley and internal problem is a direct effect of the decrease in volume associated with the solidifica-
tion; crystallisation involved more extensive changes in the specific volume than glass formation.
In addition, to use the design to solve these problems, also the injection moulding parameters can
be helpful. The phenomenon of shrinkage causes the solidified melt to have lower dimensions than
that of the mould. However, shrinkage can be minimised by providing an after-pressure that
maintains the injection pressure post-injection of the polymer melt. This consequently introduces
some extra melt into the mould cavity that counterbalances the shrinkage occurring at the extremities
of the solidifying mould.
Figure 8.16 shows how corners should be designed to keep the 90 angle. The solution is to design
so that the wall thickness remains constant through the corner.
466 8 Processing of Polymeric Materials

Fig. 8.17 Special functions added to moulded products: (a) stiffness by a rib, (b) stable tap, (c) hole for screw
connection and (d) two types of knit lines

Figure 8.17 shows a few other functions added to a mould product. Stiffness should not be
obtained by a thicker wall, but instead by using ribs (a design is shown in Fig. 8.17a). The sloping
sides of the rib are important to enable the opening of the mould for the release of the component. A
small valley may be formed on the backside of the rib. Taps and hole are included; note the stabilising
ribs surrounding the tap (Fig. 8.17a–c). The design is realised with a uniform thickness of the different
sections. A metal insert makes the screw connection strong. Knit lines are formed whenever two melt
fronts are joining; the cold knot lines are formed as a result of colinear impingement, whereas a warm
knit line is formed by splitting and joining the melt at a very low angle (Fig. 8.17d). The strength of a
cold knit line is reduced significantly with reference to the normal (full) strength, to 50% for brittle
amorphous polymers such PS and SAN (Criens and Moslé 1986) and to 80–100% for ductile
polymers such as iPP (Criens and Moslé 1986; Malguarnera and Manisali 1981). The strength is
much higher for the warm knit lines with, e.g. 75% of the full strength for the brittle amorphous
polymers according to Criens and Moslé (1986). An even stronger depression in strength was
reported for liquid crystalline polymers (Vectra) by Engberg et al. (1990), which showed only
10–20% (cold knit line) and 40–50% (warm knit line) of the full strength. Short-fibre composites
show a depression in strength similar to that of the liquid crystalline polymers (Waxman et al. 1991).
Injection moulding adds a few other unique functions which are used in everyday products: plastic
hinges and snap connections; both are often made of iPP or POM.
Although injection moulding is an effective processing method, it has some disadvantages. If the
injection pressure becomes excessively high, the closing mechanism of the mould fails to retain all
the polymeric melt within the cavity. This causes some of the polymeric melt to escape the cavity and
then solidifies as thin extensions of the final product. These thin extensions are called ‘flash’, and they
are usually trimmed off during the final product quality control stage. Nevertheless, injection
moulding is quite precise that can manufacture plastic products having attractive finishes and with
consistent quality. Recommended modern textbooks on injection moulding are Zheng, Tanner and
Fan (2011), Schiller (2018), Turng and Chen (2019) and Dangel (2020).
8.5 Extrusion and Associated Techniques 467

8.5 Extrusion and Associated Techniques

Extrusion is a major polymer processing method where materials of various viscosities and types are
conveyed, molten, mixed, vented and homogenised that significantly alter the properties of the
finished plastic product. A sketch of the extruder with commonly used accessories is shown in
Fig. 8.18. The extruder consists of a screw (or two screw) that is contained within a cylinder or
barrel fitted with heating elements and water-cooling channels. This is to ensure the set-up of a
required temperature profile along the length of the barrel. At the one end of the screw(s), the
polymeric materials in the form of granules, pellets or flakes are introduced through a hopper. The
hopper itself can be fitted with a mini-extruder that can be used to control the material flow rate. At
the other end of the screw, the molten polymer exits the barrel through a forming die. The entire
process starts by feeding the polymer, to be processed, through a hopper, which then falls on a
rotating screw(s). The polymeric material is conveyed across the length of the barrel where it is
subjected to conductive high temperature from the external heating elements and shear resulting from
the transport along the screw flights. Along the length of the screw(s), the depth of the channel
gradually reduces in order to compress the polymeric material, which then finally departs the barrel
through a die. The extrudate is conveyed over a belt where it is cooled to retain the particular shape.
The die can be designed to create various shapes, and the extrudate retains the desired shape as per
product specifications.
The screw is a vital part of the extruder. Geometric characteristics that classify an extruder are the
screw diameter (D) and length to diameter ratio (L/D). Based on these aforementioned parameters, the
required power and output for a particular rotating speed are determined (Rauwendaal 2001). The
following are some of the functions of the screw(s) in an extruder:
• Conveying of polymeric feedstock
• Enhancing both dispersive and distributive mixing
• Responsible for causing melting and degassing of the polymer
• Creating enough pressure at its end to force out the polymer through a die
The type of screw varies between extruders. Broadly, they can be categorised as single-screw and
twin-screw extruders (Fig. 8.19a). Single-screw extruders contain only one auger-shaped screw,
whereas twin-screw extruders have two parallel auger-shaped screws that are housed in a single
barrel. Twin-screw extruders are further divided into several subclasses: co-rotating, counter-rotating,
intermeshing and non-intermeshing (Fig. 8.19b). Intermeshing type of screws can be applied for both
co- and counter-rotating screws. In co-rotating screws, both the screws rotate in the same angular
direction, and in counter-rotating, they rotate in opposite angular directions. In intermeshing screws,
the inter-axis distance (l ) is shorter than the sum of the radii of the two screws. In other words, in
intermeshing screws, the flights of one of the screws infiltrate the channel of the other. For
non-intermeshing screws, the inter-axis distance is the same as the sum of the radii of the two screws,
meaning the flight tips of screws are in very close proximity to each other.

Fig. 8.18 Sketch of the extrusion process

468 8 Processing of Polymeric Materials

Fig. 8.19 (a) Differences between single- and twin-screw extruders shown in a schematic manner. (b) Pictorial
representation of different types of twin-screws (l ¼ distance between the centre of the screws, d ¼ screw diameter,
h ¼ depth of screw channel)

The conveying mechanism of the polymeric material differs between single- and twin-screw
extruders. In single-screw extruders, the extrusion process is heavily reliant on the inherent polymer
material properties of viscosity and friction. In general, twin-screw extruders, owing to the greater
extent of attrition between the screws, cause efficient mixing and homogenisation. Twin-screw
extruders are independent of the frictional and viscous forces for the transport of the polymer. Due
to this reason, most modern extruders are twin-screw since they are able to manufacture higher-
quality plastic products at a faster rate.
The virtue of being a twin-screw may not necessarily grant the extruder enhanced homogenising
power. If the extent of intermeshing is high, blending will be less, although with high forward
pumping. On the contrary, if the intermeshing is low, blending will be more with reduced pumping.
Basically, low intermeshing allows polymeric material to be trapped between the screws and is then
subjected to greater homogenising. In spite of this, twin-screw extruders, as stated earlier, have
greater ability than single-screw extruder for homogenisation, and as such, they are the desired screw
type in the plastics industry.
The extruder has four zones that are shown in Fig. 8.20a. The apparent lengths of the following
zones are dependent on the type of polymer to be processed (Fig. 8.20b). The nylon screw has a very
short compression zone, which has been believed to be due to the sharp melting of the polyamides.
The nylon screw is usually also utilised for iPP and POM. However, according to Rauwendaal (2014),
these polymers (PA, iPP and POM) perform well in continuously compressing screws. PVC is more
difficult to extrude due its broad melting, and therefore the continuously compressing screw works
best. The polyethylene screw is intermediate between the nylon and the PVC screws.
In the feed zone, the polymeric material is preheated and transported to the subsequent zones. The
screw depth in this zone is constant to ensure that appropriate amount of material is conveyed to the
subsequent zones. The melting in this zone is a function of the frictional properties of the polymeric
material, which is caused by the interaction of the polymer granules between each other, between the
granules and the inner barrel and between the screw and the polymer granules.
In the compression zone, the polymeric material is compacted since the depth of the screw
gradually reduces. This is done simultaneously to expel trapped air pockets back into the feed zone
8.5 Extrusion and Associated Techniques 469

Fig. 8.20 (a) The four zones in the extruder. (b) Three different extruder designs. (c) A decompression zone to enable
removal of volatiles. (d) A method to calibrate (i.e. control the size) the outer diameter of an extruded pipe

and enhance the conductive heat transfer from the now-reduced thickness of the polymeric material
(owing to being in close proximity to the barrel inner wall).
In metering zone, the depth of the screw is constant, as it is at the end of compression zone. The
constant depth of screws facilitates the homogenisation of the polymer, which is then supplied at
unvarying pressure and temperature to the die at a constant rate.
The die zone is the last zone in the barrel. It contains a perforated breaker plate and sieves that filter
out foreign and unwanted particles. The die zone enables head pressure to develop and eliminates the
‘turning memory’ of the melt. The turning memory of the melt is the tendency to be blended in a
rotating fashion by the screws. Several factors determine the die pressure: the polymer viscosity, die
temperature/cross-sectional area and the rate of flow of the melt. A proper control of the die pressure
enables high rates of production and an efficient operation of the extruder.
A decompression zone creates a locally reduced pressure, which makes possible the transfer of
dissolved volatiles (e.g. moisture) to the gas phase, which exits through a hole (Fig. 8.20c). The start-
up of the extrusion is carried out in steps wherein initially the barrel and the die are preheated to the
standard operating temperature of the polymeric material. The screw speed is gradually increased to
the operating level to avoid excessive load on the motor and to discourage high pressure build-up
within the barrel. Usually, before the polymeric material is introduced in the extruder, purging is
done. This is to confirm that any contaminated and degraded materials that are remaining inside the
barrel and within the screws from the previous operation are expelled. Purging is a cleaning process of
the barrel and the screw(s), and it is interrupted when clear extruded polymer (without any specs of
dirt/contamination) leaves the die. The temperature in the feed area should be relatively high that
causes preheating of the polymer; however, the temperature should not be too high that can cause
gaps in the feed throat. The temperature in the barrel is usually set somewhat below the melt
temperature of the particular polymer being processed. In this case, the heat generated from the
shear of screw speed and channel depth controls the temperature. On the other hand, the temperature
of the die is set according to polymeric material processing specification, because the aforementioned
phenomenon is non-existent at that section of the extruder.
470 8 Processing of Polymeric Materials

The melting of the polymeric material occurs by first forming a thin layer or film on the inner wall
of the barrel. The rotating action of the screw(s) abrades this layer from the wall and conveys it to the
front of the screw flight. The melt upon reaching the screw core rises up again, thereby creating a
rotational motion in the forward face of the screw flight. At the start of the screw flight, solid granules
are present, but they are transported to the molten state of the polymer through the above-mentioned
rotational action. This process continues until no solid polymeric material remains between the screw
flights. Nevertheless, the transport of the polymeric material is also dependent on its adherence with
the screw and/or the barrel. The undesired scenario is the polymeric melt adhering only to the screw
leading to no output of the extrudate. On the other hand, if the polymer melt does not stick to the
screw and resists the rotational action, an axial melt movement is created that is ideal. However, in
reality, a combination of the aforementioned scenarios occurs because the polymeric melt has the
tendency to stick to both the screw and the inner wall of the barrel. The output from an extrusion
process can have certain fluctuations that are generated e.g. by non-uniform feeding, melting and
die–flow instabilities. If this variability in extrudate is significant, then undesirable characteristics are
manifested on the product, e.g. sharkskin, spurt flow, and bamboo fracture. All these lead to poor
finish of the plastic product.
An array of products can be made by extrusion having constant cross-sectional shapes like pipes,
rods, sheets, films, etc. at high production rates. Pipes are one of the major products obtained by
extrusion; prominent pipe materials are polyolefins and PVC. Very large pipes with diameters in
metre range are usually thick-walled, and there is clear risk that the inner wall material is oxidised
which leads to inferior quality (Gedde, Terselius and Jansson 1981). By using inner atmosphere
which is inert (e.g. nitrogen), oxidation can be avoided. Figure 8.20d shows design of the die with a
mandrel held by a spider and the calibration of the outer diameter with an internal pressurised gas.
Another large volume, high-quality product is electrical cables, which have several layers
surrounding the conductive metal core. Figure 8.21 shows film blowing. The melt with a tubular
shape is exciting the extruder and is expanded in the radial direction by an internal gas pressure, and
simultaneously the tube is stretched axially by two rollers. Thus, the film is biaxially stretched,
primarily in the region below the freezing line (cf. Fig. 8.21). Film blowing is rheologically
demanding which requires a material with uniform structure and material that shows strain hardening
while exposed to extensional flow (Münstedt 2020). Low-density polyethylene with long-chain
branching is one such strain-hardening polymer. A recent highly recommended text is by Münstedt
(2020). The relatively recent general textbooks on extrusion by Rauwendaal (2014) and Wagner et al.
(2014) are also recommended.

Fig. 8.21 Film blowing by pressurising tube of molten polymer exiting from extruder
8.6 Thermoforming 471

8.6 Thermoforming

Thermoforming is a cost-effective polymer processing method for small-sized to larger products of

relatively simple shapes wherein a thermoplastic sheet is initially heated and then shaped by
decreasing the atmospheric pressure between it and a mould or by applying a higher pressure on
one of the sides by pressurised gas or by a solid plug. The starting material for thermoforming usually
is a sheet manufactured by extrusion, which is then shaped by application of pressure on one side to
generate a three-dimensional final shape.
Not all polymers can be thermoformed with ease. It is critical that the polymer being processed
possesses a rubbery behaviour above its melting point when subjected to high-temperature profiles.
This makes amorphous polymers like poly vinyl chloride, polystyrene and polycarbonate apt for
thermoforming processing above their glass transition temperature. Semicrystalline polymers like
polypropylene and polyethylene have a glass transition temperature well below the crystal melting
temperature, which makes their processing by thermoforming challenging. A disadvantage of the
method is the uneven wall thickness and the larger percentage waste especially compared to injection
moulding. The cheap moulds, which are sufficient in view of the low pressures involved, make the
method economic for short series. For long series and for finished product with high demands on the
tolerances, injection moulding is a better method.
The thermoforming process is schematically shown in Fig. 8.22. Initially, a sheet of thermoplastic
is clamped at the edges over an open mould. Upon heating, the polymer sheet expands and softens,
which cause it to sag below under the influence of gravity. The heating is removed when appropriate
softening has taken place, and pressure is applied either on the top surface of the sheet or by sucking
out air between the sheet and the mould. At this point, the heated sheet lies over the mould, and
deformation occurs as the sheet takes the form of the mould. The pressure is retained until the sheet
forms a thin-walled structure around the mould. After cooling, the pressure is removed; however, the
shaped sheet rests over the mould. The final plastic product is obtained by trimming it from the rest of
the sheet having a uniform shape.

Fig. 8.22 Schematic representation of the thermoforming process

472 8 Processing of Polymeric Materials

The softening of the polymer is done to ease the stretching without the possibility for tearing. The
thermomechanical properties of the polymer determine its softening performance. Infrared heaters
provide the required temperature in the thermoforming heating process. It is important that the
temperature over the sheet is uniform. Since air pressure is used, the moulds used in thermoforming
do not need to be overtly strong. Moulds can thus be made of wood and epoxy resins; however, the
preferred material for thermoforming is aluminium since it can be polished into a smooth shape and
exhibits acceptable thermal conductivity. These moulds can be male (convex structure) or female
(concave structure) with the female mould facilitating an easy product ejection and defined peripheral
areas, whereas the male mould can be economic to produce. Thermoforming moulds are much
cheaper than injection mounding moulds. When the sheet is formed over the male mould with the
inner surface of the sheet replicating in direct contact with the mould, the process is called ‘positive
forming’. On the contrary, if the sheet is over the female mould with the outer surface replicating, the
process is termed as ‘negative forming’. In negative forming, the wall thickness near to the clamp is
higher than the corners, whereas in positive forming, a better distribution can be achieved. Neverthe-
less, positive forming has challenges to release the formed part.
During the thermoforming process, care must be taken so that the heating does not cause
degradation in the polymer. Usually, a lower heat applied uniformly for a long time achieves the
best results. The definition and finish of the product are also reliant on the extent of pressure applied,
and the mould should have a uniform and high thermal conductivity. If the polymer sheet is thick and
the final shape complex, a combination of air pressure and vacuum is simultaneously applied.
Regardless, thermoforming is widely used for food packaging applications like cups and trays and
other products having large surface area like refrigerator liners, switch panels, automobile bumpers,
bathtubs and small rowing boats. A modern account for the development in this field is the textbook
of Engelmann (2012).

8.7 Producing Hollow Objects: Rotational Moulding and Blow Moulding

Rotational moulding is an economic polymer processing method that can be employed to produce
hollow medium-to-large-sized plastic products in short series. In simple terms, the processing
involves heating a hollow mould to a temperature above the melting point of the polymer to be
processed; the polymer is usually initially in form of a fine powder which by biaxial rotation of the
mould using a Polhem knot (due to Swedish Christopher Polhem) is spread in the containment and
melts when it reaches the mould wall. After cooling, the mould is opened to collect the finished
product.
The polymeric particles soften when they encounter the heated wall of the mould, and the solid
particles subsequently follow the shaping process until a compact layer of polymer is formed on the
inner mould surface. The outer surface of the finished hollow products is smooth, whereas the inner
wall surface is thus much rougher. The mould used is cheap since no pressure is involved in the
process, which is good for the production of short series. Rotational moulding is apt for the
manufacture of medium-to-large-sized products like boats, containers, garbage boxes as well as
open-surface products like automobile dashboards and furniture.
The different stages of the rotational moulding cycle are sketched in Fig. 8.23. Initially, a certain
mass of polymer powder is introduced inside the metallic mould. Then the two halves of the mould
are clamped together and are heated with the aid of a convection oven. In most cases, the oven is
preheated to shorten the cycle time. While the mould is being heated, it is rotated about two axes
perpendicular to each other; the rotation velocity is typically 20 revolutions per min. The combined
heating and rotation cause the plastic to soften and form a homogeneous layer on the inner surface of
the mould. The mould cooled is by air, fan or water shower while still rotating when all powder has
8.7 Producing Hollow Objects: Rotational Moulding and Blow Moulding 473

Fig. 8.23 Step-by-step schematic representation of rotational moulding

been transferred to molten layer at the inner wall surface of the mould cooled. This causes the
polymer to attain acceptable rigidity. At this point, the rotation is ceased, and the mould is opened to
collect the final finished plastic product.
The melt viscosity of the polymer is important for rotational moulding. If the viscosity is high, the
polymeric melt mobility is curbed, and this makes sintering slow and makes the cycle time long.
Incomplete sintering creates products that possess holes (bad for a hollow product) with low fracture
toughness. For some designs it may be that the centre of the rotation is not equally distant from all the
mould inner surfaces, which may cause a non-uniform wall thickness when centrifugal force is
applied.
The moulds used in this process are generally made of aluminium or steel. The mould wall
thickness can be relatively small (ca. 1.5 mm) since the rotational forces are not high. The cycle
times are based on the wall thickness of the finished product and are usually between 3 and 40 min
(Saldı́var-Guerra and Vivaldo-Lima 2013). In rotational moulding process, the cooling rate of the
mould is of paramount importance. Although faster cooling rate can be set for economical reason, the
finished product can receive defects from warping. Due to this reason, the mould is gradually cooled
by first subjecting it to air and then spraying it with water.
There are several advantages of the rotational moulding method: (i) no other single processing
method can make very large hollow products; (ii) the mould is cheap; (iii) products show zero chain
orientation with resulting excellent dimensional stability; and (iv) the wall thickness shows less
variation than competing methods, blow moulding and thermoforming. However, the disadvantage of
474 8 Processing of Polymeric Materials

Fig. 8.24 Sketch of extrusion blow moulding equipment also showing the different stages in bottle blowing. (Drawn
after Morton-Jones 1989)

the method is the long cycle time. Recommended further readings about rotational moulding are
textbooks by Beall (1998) and Crawford and Kearns (2012).
Blow moulding can be conducted in combination with an extruder as shown in Fig. 8.24. The melt
exiting the die is meeting the parison variator and the melt turns vertical and changes shape to a tube.
The parison variator can be adjusted to obtain tubes with different wall thickness. The tube enters the
mould according to the illustration in Fig. 8.24. The feeding of tube is stopped and the mould is
closed. The pressurised gas blows the hollow product until it touches the inner walls of the mould.
The aim is to stabilise the process by the strain-hardening characteristic of elastic extensional flow.
This can be accomplished by reducing the time scale of the process, fast blowing and/or low
temperature. Too fast blowing leads however to rupture of the hollow melt. The process has to be
adjusted by running a number of experiments in order to find the optimum conditions. In practice the
finished product usually shows some variation in wall thickness.
A related process is injection blow moulding by which first forms a preform instead of the extruded
parison. The preform is cold to a solid state and at a later stage is heated above the glass transition
temperature or crystal melting temperature. The formation of the finished bottle-shaped product is
obtained by axial deformation of the preform by the so-called blow pin, and this accessory also
manages the air blowing which makes the preform to expand in the radial direction. This is how PET
is blown to the well-known bottles storing soft drinks.
Further facts about blow moulding can be found in Lee (2006) and Brandau (2016).

8.8 Compression Moulding

Compression moulding is the oldest polymer processing method. Commercially, it is currently mostly
used for thermosets but is indeed a useful method in research laboratories to make thermoplastic
specimens for mechanical testing. In essence, the polymeric material, in the form of powders,
granules or flakes, is placed between two halves of a mould, which is then heated and pressed
8.8 Compression Moulding 475

Fig. 8.25 A sketch of the compression moulding process

together to manufacture the finished plastic product. Thermoplastics, thermosetting plastics and
elastomers can be processed using this method.
The step-by-step processing of polymeric materials in compression moulding is depicted in
Fig. 8.25. In the process, the mould is preheated, and then a pre-weighed amount of polymeric
material is placed in the lower part of a heated mould. Non-sticky sheets, such as PTFE or PET, are
placed between the polymeric material and the mould inner surfaces. This is to ensure that the molten
polymer does not stick to the mould walls after the completion of processing. After this, the upper part
of the mould is pressed down using hydraulic power, or it can be done manually too. The combined
effect of heat and pressure shapes the polymer (it causes crosslinking reactions in thermosetting
polymers). Once the polymeric material is shaped, the mould is cooled by water channels in case of
thermoplastics) or opened while still hot in case of thermosetting plastics. The optimum temperature
and pressure to be used depend on the type of polymer. It is critical that the temperature set is below
the degradation temperature of the polymer that is being processed. The polymeric material is
generally preheated before the actual compression moulding operation because it reduces the
temperature difference between the feed and the mould. The amount of polymeric material to be
added is determined by calculating its mass based on the density of the polymer and the volume of the
mould. It is also advisable that the pressure is applied onto the mould gradually to allow the polymer
to soften and flow in all the corners of the mould. This will also ensure that the heating is uniform and
the finish is of high quality.
The processing of thermoplastic polymers by compression moulding is not desirable since the
mould has to be cooled to allow the plastic product to shape before it can be taken out of the mould.
Thermosetting polymers, on the other hand, are easy to process since the polymeric material along
with fillers and curing agent can be applied directly onto the mould where the heat and the pressure
cause the curing reaction. Thermosets can be taken out of the mould before it has cooled. Elastomers
are also processed similarly to thermosetting polymers, where the rubber along with vulcanisation
agent and fillers is subjected to the process and the temperature is held until the vulcanisation reaction
is complete, after which the finished product is taken out.
476 8 Processing of Polymeric Materials

In summary, the advantages with compression moulding are low tool costs (including moulds),
low maintenance costs and low scrap. The structure of moulded products has also favourable features:
low orientation and low internal stress. However, it is not possible to create intricate shapes by this
method. Numerous products are manufactured by compression moulding, e.g. light fittings, electric
switches and other several other household articles. Davis et al. (2003) is a recommended textbook on
the subject.

8.9 Calendering

The term calender has originated from the Greek word ‘kylindros’ meaning ‘cylinder’. As per the
Webster’s International Dictionary, it means pressing in between rollers to produce glassy and
smooth thin sheets. In the calendering process, continuous plastic films or sheets are formed by
pressing the melted polymer by two or more rotating cylinders or rollers. The concept of forming
plastic into thin sheet or film through the calendering process was adopted from the manufacturing
process involved in making paper, metal and rubber in industries. When the method was first
developed (nineteenth century), it was mainly used for the processing of rubber, but at present, it is
used for multipurpose applications, such as thermoplastic sheets, coatings and films. At the time of
development, the calendering of polymers was not a complete success due to the lack of capability in
the process to regulate the speed of the rollers and the gap (nip) between the rollers. However, in the
early 1930s, the process began to develop with changes in the calendering process for the processing
of new polymers, which also enabled modifications in machines that are capable of controlling
individual roller speeds and gaps between them at a controlled temperature. Such changes are still the
key features in the present calendering machines. The calendering process is used in various
manufacturing industries for the production of thin sheets of plastics having required size and
specification. By means of calendering, high-quality plastics can be produced as films and sheets at
a relatively high volume. The typical arrangement of rollers is shown in Fig. 8.26.
The number of rotating rollers in the calender machine varies depending on the type of machine
used. Usually, the industrial-type machine consists of three to six rollers. Based on the configuration
of the roller, the calenders are classified as superimposed calenders (I-Type), offset calenders (L or

Fig. 8.26 Typical

arrangement of calender
rolls
8.10 Additive Manufacturing 477

inverted L type) and the Z-type calender. The “I type” calender has one more roller in the stack and
has an advantage of taking up the minimum floor space. However, this design is not ideal, since the
external force acting on the roller causes the nip to deviate. Calenders of type L are used for the
processing of rigid polymers, whereas the inverted type L is used for flexible polymers. These
calenders are popular because the first pair of horizontal rollers hold the input material that can be
easily passed to the next two nips and ordered to provide size and measurement control. In the L-type,
the rollers are positioned perpendicular to each other, which reduces the roll-separating forces acting
on the adjacent rollers. In the Z-type, the rollers are positioned at right angles to the next pair of
rollers, thereby avoiding any effect of roll-separating forces on any roller, which offer improved
thickness accuracy. Generally, the size of the calender ranges from a diameter of 90 cm to a width of
250 cm. The rotational speed of the rollers varies with respect to the polymers, and typically for thin
flexible polymers, the rotational speed is greater than 2 m s1. Several additional equipment are
needed in real time to prepare and process the polymer material before it is rolled. Rigid polymers as
well as plasticised compounds can be processed in the calendering process, such as polyvinyl chloride
(PVC), polypropylene and acrylonitrile butadiene styrene (ABS) plastic. However, PVC is a highly
used material that needs to be blended, mixed and compounded with appropriate chemical additives
prior to processing. Before processing, the polymers are pre-mixed and blended with other additives
using a ribbon blender or a Henschel mixer (cf. Sect. 8.3). Premixing is done at a temperature up to
80  C in order to achieve good absorption and soaking of polymers. The blended mixture is fed to a
fluxing machine where the gelation is done in a Banbury mixture or a continuous extruder operating at
150  C. During these operations and at other stages, the foreign materials present in the polymer are
removed by means of a metal detector. After the preliminary processes, the processed fluxed polymer
melt having a dough-like consistency is fed into the calender roll at around 140 to 160  C, to be
converted into film or sheet.
Calender rolls apply heat and pressure to the bypassing polymer to obtain a flat, glossy and smooth
surface. Calender rollers are maintained at a temperature of 150  C to 200  C by steam or hot oil.
Fluxed material is then passed through a series of rollers of varying nips, which reduces the thickness.
For example, the fluxed material, when passed to the first roller, forms a 15 cm roller bank, which,
when passed to the next set of rollers, converts to a 10 cm roller bank, so that the final roller achieves
the required thickness. The high temperature of the sheet may cause it to shrink. This is avoided by
passing the hot sheet through the cooling rolls and then through the thickness gauge. After proper
cooling, the sheet is trimmed to the required product specification. The sheets with thicknesses
ranging from 0.1 to 1.0 mm or thicker can be formed through calendering process, and they may
be transparent, coloured, embossed, printed or laminated A tolerance of
0.005 mm can be
maintained; however, to achieve this, an enhanced control of roller temperature/speed and nip gap
is necessary. Most importantly, the diameters of the rollers should be accurate and have a uniform
surface finish. The leftover material can also be recycled. Plastic products made from the calendering
process are used in various applications such as in automobiles, packaging and medical fields. Some
examples of plastic products are luggage bags, toys, stationery items, files, folders and raincoats.

8.10 Additive Manufacturing

Additive manufacturing, also known as 3D printing, is used for the manufacturing of 3D models and
products. Additive manufacturing is a computer-controlled process involving layer-by-layer addition
of material that enables the production of objects with complex shapes. Additive manufacturing
eliminates certain drawbacks of conventional manufacturing processing. Based on the type of
materials used, deposition technique and solidifying process, ISO/TC 261 and ASTM F42 classified
the additive manufacturing process into seven major groups, which are photopolymerisation,
478 8 Processing of Polymeric Materials

material jetting, binder jetting, material extrusion, powder bed fusion, sheet lamination and direct
energy deposition. The process selection depends on the specification of the product and the intended
end use. Photopolymerisation, material jetting, powder bed fusion and material extrusion are the most
widely used methods for polymer printing. On the other hand, direct energy deposition is commonly
used for metal printing. The common element in all of these processes is the use of computer
software. The digital model of the product is required for all of these processes to print the 3D
model. The digital model is developed using a computer-aided design (CAD) package, a 3D scanner
or a digital camera and a photogrammetric software package. Due to design specifications and
material properties, not all types of polymers can be printed using all the 3D printing methods. In
some cases, polymers must be specifically processed for additive manufacturing.

8.10.1 Steps in the Development of Digital Model

In the first step, the digital model of the part to be printed is created using the available design
software, such as AutoCAD, Creo Parametric, SOLIDWORKS , CATIA and Unigraphics. These
softwares enable the operator to modify the design model in the event of errors, which also facilitate
design verification prior to the printing. Design faults can also be analysed using software such as
ANSYS, COMSOL and ABAQUS. The design model is produced according to the material type and
manufacturing process, which follows the design guide provided by the 3D printing machine
manufacturer. The stages involved in the 3D printing from the designing to the 3D modelling are
shown in Fig. 8.27.
After the generation of the digital model of the part, it is converted to the Standard Triangle
Language (STL) – file format. The .STL file describes the surface geometry of the 3D model. The .
STL file format uses a sequence of connected triangles to replicate a surface geometry of the model to
be printed. The .STL file is critical for dividing the 3D model into multiple layers.
In order to divide the 3D model into several layers, a slicing operation is performed using a slicer
software. The slicer software uses the .STL file and generates the coordinates (G-code), and
accordingly, the printer head moves. Printing factors such as number of layers, layer height and
print orientation are considered during slicing. The slicing process is crucial as it determines the
quality of the printed 3D model. CURA, MakerBot, Netfabb, PrusaSlicer, OctoPrint and Simplify3D

Fig. 8.27 The steps in 3D printing process

8.10 Additive Manufacturing 479

are some of the slicing software available in the additive manufacturing market. After the successful
conversion of the STL file to G-codes, the 3D printer is able to print the model. The printing process
varies depending on the type of AM operation involved.

8.10.2 3D Printing Methods for Polymers

In the photopolymerisation technique, the pre-deposited photopolymer liquid contained in the vat
(tank) is converted to solid or cured by heat. The heat is selectively applied layer by layer until the
complete 3D model is printed. Generally, photopolymerisation uses laser for curing. In addition to the
conventional laser-curing source, there are different types of curing sources available. Digital light
projectors and liquid crystal display (LCD) screens are popular materials for photopolymerisation due
to their high resolution and are also economically viable. These two techniques are capable of curing
the full layer of photopolymer, whereas in the laser-curing technique, the layer can only be cured by
illuminating the laser over the entire surface. Stereolithography (SLA), digital light processing
(DLP), continuous liquid interface production (CLIP) and daylight polymer printing (DPP) by
Photocentric are the widely used photopolymerisation methods. Radiation-curing acrylics and acrylic
hybrids are the commonly used polymeric materials in the photopolymerisation method. The
polymerisation method is capable of producing models with the high precision and good surface
finish. Figure 8.28 shows the process diagram of vat photopolymerisation 3D printing. The steps in
the photopolymerisation method are:
• The build platform at the top of the printer set-up is lowered to the photopolymer liquid tank
according to the thickness of the layer.
• By passing the UV light beam, the first layer is formed by curing, and the build platform moves
down to create the second layer. Thus, the process continues until the entire layer is printed into a
3D model.
• Some photopolymerisation machine consists of a blade in between the layers to ensure smoothness
for next layer to build upon. The remaining liquid polymer that is not cured by laser remains inside
the tank, which can be reused.
• After the entire layer is printed, the build platform rises, allowing the excess liquid polymer to
drain into the tank. The model is then washed to eliminate any liquid polymer in the cured model.

Fig. 8.28 Vat photopolymerisation printing process

480 8 Processing of Polymeric Materials

• The printed model is then placed inside the oven for final curing, which bestows strength and
stability for the models.

Stereolithography (SLA) is a widely used photopolymerisation method in additive manufacturing,

which belongs to the resin 3D printing method. In this method, the laser is used as a curing medium to
harden the liquid photopolymer in the reservoir. During the process, the surface of the liquid
photopolymer is heated layer by layer by focusing using a concentrated laser beam. As a result, the
polymer gets crosslinked creating a 3D printed model. SLA is considered to be a relatively fast
method for building a 3D model; however it is not sufficiently fast for mass production. To achieve
this, different photopolymerisation technology, such as continuous liquid interface production
(CLIP), has been developed, which is 25–100 times faster than the traditional 3D printing.
The material jetting process is similar to the conventional 2D ink jetting process, wherein the
material is jetted onto the platform and allowed to solidify. Solidification is done by the application of
UV light source akin to the photopolymerisation process. Thus, each layer is built by jetting the
material layer by layer. Jetting is done by two different approaches: continuous or drop-on-demand
(DOD) approach. Figure 8.29 shows the process diagram of material jetting 3D printing. The material
is jetted from the small pores of the nozzles that move horizontally over the surface of the build
platform. Since the material used in the jetting-type-3D printing process is to be deposited in drops,
only polymers and waxes are preferred to be used. The material jetting a 3D printed model exhibits
good surface finish and high precision with excellent detail. However, this type of printing technology
is costly, and due to the UV curing, the model exhibits poor durability and may become brittle.
In binder jetting type of the 3D printing process, a bonding adhesive material is deposited over a
thin layer of powder material (Fig. 8.30). In this process, two types of material are used: the adhesive
binder material and the powdered build material. In general, the binder material used is in the form of
liquid, and build material is in the form of solid powder. The binder adhesive is deposited as a droplet
by horizontally moving the print head analogous to a 2D printer that prints ink on paper. After
deposition of the adhesive over one layer, the build platform moves downwards, and the new layer of
the build material, in powder form, spreads over the build-up area by the use of roller or blade. Then
the adhesive material is deposited as the next layer, and the process is repeated until all parts are

Fig. 8.29 Material jetting 3D printing process

8.10 Additive Manufacturing 481

Fig. 8.30 Binder jetting 3D printing process

Fig. 8.31 Material extrusion (FDM) 3D printing process

finished. However, an additional post-treatment process is required to increase the strength of the
printed part. Binder jetting process is commonly used for metal and ceramic printing. However,
polymeric materials such as epoxy, acrylic, ABS, PA, PC and PLA can also be printed. The process is
simple, is fast and produces relatively low waste; however, the printed part often features a poor
surface finish.
In material extrusion type 3D printing method, thermoplastic materials are printed through the
extrusion of semi-melt thermoplastic build material. In this process, Fused Deposition Modelling
(FDM) is the most popular method. FDM printing machine consists of the various parts such as
feedstock feeder, extrusion head and build platform. Figure 8.31 shows the process diagram of
material extrusion-FDM 3D printing. The feedstock material is in the form of long continuous
filament. During the printing process, the thermoplastic filament is first loaded into the printer.
When the nozzle has achieved the target temperature, the filament is fed to the extrusion head and
to the nozzle where it is converted into a semi-melt form. The extrusion head is attached to a three-
482 8 Processing of Polymeric Materials

Fig. 8.32 Process diagram of powder bed fusion 3D printing

axis system that allows the extrusion head to move in x, y and z directions. The partially molten
polymer is extruded in the form of thin threads and placed layer by layer in defined places where it
cools and solidifies. The cooling of the printed part is enhanced with the usage of cooling fans
connected to the extrusion end. When one layer is finished, the build platform moves down and the
next layer is deposited. The process will continue until the complete model is printed. Various types
of thermoplastics can be printed using this process, such as ABS, PLA, PP, PEEK, PA, PC and PP. By
this process, it is possible to print composite type 3D model. The extrusion type 3D printing is
economical with low waste. However, the process is slow, requires a strong filament and its operating
temperature is high.
Power bed fusion type additive manufacturing process works similarly to the binder jetting
process; however, here the build material powder is fused layer by layer by the application of heat
produced through a laser or electron beam. Based on the working process, this method can be further
classified into direct metal laser sintering (DMLS), electron beam melting (EBM), selective heat
sintering (SHS), selective laser melting (SLM) and selective laser sintering (SLS). In powder bed
fusion process, a laser or electron beam is used to melt and fuse the build material. Figure 8.32 shows
the process diagram of powder bed fusion 3D printing. The build material is in the form of a powder
that is stored in a hopper or in a container located near the build platform. Once the first layer is fused,
the build platform is lowered according to the thickness of next layer. Following this, the fresh
powder is spread over the previously fused layer by the use of a roller or blade. The process of fusion
is then repeated. The entire process is repeated until the final layer is finished. Selective laser sintering
(SLS) 3D printing technology is the most preferred powder bed fusion method for building polymeric
materials. This process uses a laser source to fuse the polymer powder, which can be either
amorphous or crystalline thermoplastic. Polyamide, filled polyamide, polyether ether ketone
(PEEK) and polystyrene are widely used polymeric materials for this method.
In sheet lamination type 3D printing process, the material is added through a feed roller system
and bonded together layer by layer and cut to the required shape with a knife or laser. Plastic sheet
lamination, laminated object manufacturing (LOM) and ultrasonic consolidation (UC) are the
examples for the 3D printing process of sheet lamination. Materials like paper, metals and polymers
can be used in the sheet lamination process; however, each of them needs a specific method for
bonding the material together. In the case of polymers and paper, the material is bonded together by
the application of heat and pressure. However, in the case of metal, bonding between the layers is
created by applying ultrasonic vibrations under pressure. One of the main advantages of this process
is that 3D fibre composite material can also be formed using this process. This process is fast,
8.11 Summary 483

Fig. 8.33 Process diagram of sheet lamination 3D printing

economical and easy to operate. However, for polymers, post-treatment is needed to improve the
bonding between layers. Figure 8.33 shows the process diagram of sheet lamination 3D printing.
The sheet lamination process is performed as follows: (i) The building material is initially stacked
on the base plate. The stacked layer may or may not bond to the previously laid layer depending on
the type of the process. In the case of the plastic sheet lamination process, the layer is bonded by the
adhesive, whereas in the case of the laminated object manufacturing, the layers are bonded by fusing
polymer that is fed by the roller. (ii) The layer is then cut by using laser heat or a knife according to
the geometry of the product specification. (iii)The next layer of build material is stacked, and the
process repeats until the entire layer is finished. (iv) Finally, the printed model is removed and
trimmed to remove unwanted materials.

8.11 Summary

Polymer processing is a multifaceted engineering science because it is based on so many different

fundamental disciplines: the rheology and physics of polymers, polymer chemistry, thermodynamics,
transport properties concerning both heat and matter (Chap. 7), solid mechanics (Chap. 6), mechani-
cal engineering and modelling and simulation (Chap. 5). Essentially all of these topics are thoroughly
treated by our recent textbooks, Gedde and Hedenqvist (2019f) and Gedde (2020), and in other
chapters of the current volume. In the two first sections, the connections between the fundamental
fields and the polymer processing field are highlighted. The different processing methods are
compared presenting both the advantages and the disadvantages of each method and also a ranking
of the different methods as applied to different types of products. The environmental impact of
making and using products based on different material types is best learnt by combining a study of the
current chapter and Chap. 9. The main thermoplastic processing methods are covered in the following
sections: compounding, injection moulding, extrusion (including film blowing), thermoforming,
rotational moulding, compression moulding and calendering. The last part deals with the modern
484 8 Processing of Polymeric Materials

additive manufacturing using a 3D printing technology. It is evident that ca. 40 pages is a very short
account for this huge field, and therefore most sections contain references labelled ‘recommended
further reading’.

8.12 Exercises

8.1. Compounding is an important start of polymer processing. It is often carried out by the company
selling the raw materials. Explain the different stages in obtaining a good compound. How is it
possible to distribute evenly a number of different powders? Can heating be a good means to
obtain a uniform distribution? How can you validate a good dispersion and good homogeneity of
the additives in the compound?
8.2. Compare injection moulding, rotational moulding and blow moulding with regard to object size,
dimensional tolerance and number of products (e.g. 100, 1000 and 106).
8.3. Use Fig. 8.3 and Eq. (8.3) to calculate the relative melt flow index (MFI) for a series of polymers
with different molar masses: M ¼ 1000, 10,000, 50,000, 300,000 and 1,000,000 g mol1.
Normalise with regard to the sample with M ¼ 10,000 g mol1. The British Standard dimensions
of the MFI equipment: channel length (l ) ¼ 8.00 mm and channel radius (r) ¼ 1.05 mm. How is
it possible to assess MFI for polymers with very high molar mass?
8.4. A detail with hole can be obtained by injection moulding. Why is the material on backside from
the gate sometimes brittle? Which categories of polymeric materials are sensitive to this design?
Suppose you have a real problem with your product; the fracture toughness is only 20% of the
full strength behind the hole. How can you improve the fracture toughness (you are not allowed
to remove the hole)?
8.5. Provide examples of extensional flow in different polymer processing methods? What rheologi-
cal properties need to be assessed in order to control this type of processing?
8.6. This plastic object is made of isotactic PP and it has three valuable functions. Please try to
identify them (and tell more about their functions) from the photographs shown in Fig. 8.34.
Which processing method has been used to make this product?

Fig. 8.34 Plastic product with three different built-in functions

References 485

Fig. 8.35 Two well-functioning household plastic products

8.7. Plastic products are used in the household in food contact. The bowl with cover to the left is
used for processing of food mixtures (Fig. 8.35). The bowl is made of melamine thermoset, and
the cover is made of a thermoplastic polyolefin. How were they made and what was critical in
the processing? What functions of the different components were important? The right-hand
container (bowl and cover) is used for storage in the fridge of processed food (Fig. 8.35). This
particular item is 20 years old, and it is still fresh and well-functioning. Both parts are made of
thermoplastic polyolefins. Suggest best processing methods, and pinpoint what is critical with
the processing. What are the requirements that the used material should fulfil?
8.8. Rotational moulding is a useful method to produce large hollow thermoplastic objects. The
moulding is rotating in a biaxial fashion. (a) The thickness of the plastic wall is very uniform.
How is that possible in view of the large size of the finished product? (b) Why is the mould
cooled while it is still rotating?
8.9. Before starting an extrusion session, what do you need to do to make sure the product is of good
quality?
8.10. Extrusion of thick-walled plastic pipes of polyethylene is problematic. Explain why! How can
you find out if your hypothesis is correct? What experiments need to be carried out? Finally,
how can you solve the problem?

References

om, B. T. (1997). Manufacturing of polymer composites. London: Chapman & Hall.

Åstr€
Barrie, I. (1978). In R. S. Lenk (Ed.), Polymer rheology. Barking: Applied Science Publisher.
Beall, G. L. (1998). Rotational moulding: Design, materials, tooling, and processing. Munich: Carl Hanser-Verlag.
Bertram, A., & Glüge, R. (2015). Solid mechanics: Theory, modeling, and problems. Cham: Springer.
Brandau, O. (2016). Stretch blow molding (3rd ed.). Oxford: Elsevier.
Brandrup, J., Immergut, E. H., & Grulke, E. A. (2003). Polymer handbook (4th ed.). New York: Wiley–Interscience.
Carslaw, H. S., & Jaeger, J. C. (1959). Conduction of heat in solids. Oxford: Clarendon Press.
Crank, J. C. (1979). The mathematics of diffusion. Oxford: Clarendon Press.
Crawford, R. J. (1998). Plastics engineering (3rd ed.). Oxford: Butterworth-Heinemann.
Crawford, R. J., & Kearns, M. P. (2012). Practical guide to rotational moulding (2nd ed.). Shawbury: Smithers Rapra
Technology.
Criens, R. M., & Moslé, H.-G. (1986). In W. Bristow & R. D. Corneliussen (Eds.), Failure of plastics. Vienna: Carl
Hanser-Verlag.
Crompton, T. R. (2010). Thermo-oxidative degradation of polymers. Shawbury: Smithers Rapra Publishing.
Dangel, R. (2020). Injection molds for beginners (2nd ed.). Munich: Carl Hanser-Verlag.
Davis, B., Gramann, P. J., Osswald, T. A., & Rios, A. C. (2003). Compression moulding. Munich: Carl Hanser-Verlag.
de Gennes, P. G. (1979). Scaling concepts in polymer physics. Ithaca/London: Cornell University Press.
486 8 Processing of Polymeric Materials

Doi, M. & Edwards, S. F. (1986). The theory of polymer dynamics. Oxford: clarendon Press.
Engberg, K., Knutsson, A., Werner, P.-E., & Gedde, U. W. (1990). Polymer Engineering and Science, 30, 1621.
Engberg, K., Ekblad, M., Werner, P.-E., & Gedde, U. W. (1994). Polymer Engineering and Science, 34, 1346.
Engelmann, S. (2012). Advanced thermoforming: Methods, machines and materials, applications and automation.
Hoboken: Wiley.
Folkes, M. J. (1982). Short fibre reinforced composites. Hoboken: Wiley.
Gedde, U. W. (2020). Essential classical thermodynamics. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019a). Chapter 7: Morphology of semicrystalline polymers. In Fundamental
polymer science. Cham: Springer.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Chapter 5: The glassy amorphous state. In Fundamental polymer science.
Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019c). Chapter 6: The molten state. In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019d). Chapter 8: Crystallization kinetics. In Fundamental polymer science.
Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019e). Chapter 9: Chain orientation. In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019f). Fundamental polymer science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., Terselius, B., & Jansson, J.-F. (1981). Polymer Testing, 2, 85.
Graessley, W. W. (1984). Viscoelasticity and flow in polymer melts and concentrated solutions. In J. E. Mark,
A. Eisenberg, W. S. Graessley, L. Mandelkern, & J. L. Koenig (Eds.), Physical properties of polymers.
Washington, DC: American Chemical Society.
Graessley, W. W. (2004). Polymeric liquids and networks: Structure and properties. London: Garland Science.
Halim Hamid, S. (2000). Handbook of polymer degradation. Boca Raton: CRC Press.
Jarecki, L., & Pecherski, R. B. (2018). Express Polymer Letters, 12, 330.
Jud, K., Kausch, H. H., & Williams, J. G. (1981). Journal of Materials Science, 16, 204.
Kim, Y. H., & Wool, R. P. (1983). Macromolecules, 16, 1115.
Lee, N. C. (2006). Practical guide to blow moulding. Shawbury: Smithers Rapra Publishing.
Lienhard, J. H. (2019). A heat transfer textbook (5th ed.). Mineola: Dover Publ.
Lorenzo, A. T., Arnal, M. L., Albuerne, J., & Müller, A. J. (2007). Polymer Testing, 26, 222.
Lubliner, J., & Papadopoulos, P. (2017). Introduction to solid mechanics: An integrated approach (2nd ed.). Cham:
Springer.
Malguarnera, S. C., & Manisali, A. (1981). Society of Plastics Engineers Annual Technical Conference Technical
Papers, 27, 775.
Manus-Zloczower, I. (2005). Mixing and compounding of polymers. Munich: Hanser Publishers.
Mendelson, R. A. (1968). Polymer Engineering and Science, 8, 235.
Mo, Z. (2008). Acta Polymerica Sinica, 7, 656.
Morton-Jones, D. H. (1989). Polymer processing. London/New York: Chapman and Hall.
Münstedt, H. (2020). Polymers, 12, 1512.
Nakamura, K., Watanabe, T., Katayama, K., & Amano, T. (1972). Journal of Applied Polymer Science, 16, 1077.
Ozawa, T. (1965). Bulletin of the Chemical Society of Japan, 38, 1881.
Ozawa, T. (1971). Polymer, 12, 150.
Rauwendaal, C. (1998). Polymer mixing: A self-study guide. Munich: Hanser Publishers.
Rauwendaal, C. (2001). Polymer extrusion (4th ed.). Munich: Hanser Publishers.
Rauwendaal, C. (2014). Polymer extrusion (5th ed.). Munich: Hanser Publishers.
Rubin, I. I. (1973). Injection molding: Theory and practice. Hoboken: Wiley.
Rubinstein, M., & Colby, R. H. (2003). Polymer physics. Oxford: Oxford University Press.
Saldı́var-Guerra, E., & Vivaldo-Lima, E. (Eds.). (2013). Handbook of polymer synthesis, characterization, and
processing. New York: Wiley.
Schiller, G. F. (2018). A practical approach to scientific molding. Munich: Carl Hanser-Verlag.
Seyler, R., Erie, P., & Schenck, A. (2003). Proceedings of the 61st annual meeting of the Society of Plastic Engineers
(pp. 3373–3377). USA: Nashville.
Snow, R. D., & Frey, F. E. (1943). Journal of the American Chemical Society, 65, 2417.
Stokes, G. G. (1845). Transactions of the Cambridge philosophical. Society, 8, 287.
Ticona. (2001). Vectra – Liquid crystal polymer (LCP). Summit: Ticona GmbH.
Turng, L., & Chen, S. (2019). Advanced injection molding technologies. Munich: Carl Hanser-Verlag.
Van Krevelen, D. W., & Te Nijenhuis, K. (2009). Properties of polymers: Their correlation with chemical structure;
their numerical estimation and prediction from additive group contributions (4th ed.). Amsterdam: Elsevier.
References 487

Wagner, J. R., Mount, E. M., III, & Giles, H. F. (2014). Extrusion: The definitive processing guide and handbook (2nd
ed.). Oxford: Elsevier.
Waxman, A., Narkis, M., Siegmann, A., & Kenig, S. (1991). Polymer Composites, 12, 161.
Wool, R. P., & Connor, K. H. (1981). Journal of Applied Physics, 52, 5953.
Wunderlich, B. (1980). Macromolecular physics: 3. Crystal melting. New York/London: Academic.
Zheng, R., Tanner, R. I., & Fan, X.-J. (2011). Injection molding: Integration of theory and modeling methods. Berlin/
Heidelberg: Springer.
Chapter 9
Plastics and Sustainability

9.1 Introduction

The annual global plastic production in 2018 was greater than 360 million tons (megatons; Mt). This
is a more than 20-fold increase from the plastic production in the 1960s. The production volumes
keep increasing, and the plastic production is anticipated to further double in the next 20 years.
Plastics and other polymeric materials are an essential part of our daily lives and modern society with
applications ranging from packaging to agriculture, transportation, buildings and construction, energy
production and healthcare (Fig. 9.1).
Plastics are often the most environmentally benign and sustainable alternative during production
and service life, and they contribute in several ways to a decreased greenhouse gas emission (cf. Sect.
9.2). At the same time, the way some plastics are designed, used and discarded causes serious
environmental problems. The accumulation of plastic waste is a threat to our environment and
especially to the oceans. A sizeable part of plastic products is aimed for only short-term use, which
ranges from days to a few years. The drawback is that only a few percent of the end-of-life plastics are
recycled. The large and increased production volumes in combination with low price and the
undeveloped waste management all contribute to the global plastic waste problem. A calculation
showed that 8300 million tons (Mt) of plastics had been produced up to 2015 and that 75% of the
plastics ever produced (ca. 6300 Mt) had already become waste. At the same time, only a small
fraction had been recycled to materials with a similar value as that of the original product (Geyer et al.
2017). The current largely linear production and consumption modes are not sustainable, and it
discards the value of the end-of-life plastics as a resource for both material and chemical production.
This chapter deals with the sustainability aspects of plastics and two related major challenges:
(i) the transformation from fossil resource-based to bio-based materials and (ii) the change from
linear to circular materials (Fig. 9.2). These changes of the material paradigm have an impact on the
plastic waste problem and the potential of polymeric materials as being fully sustainable (Billiet and
Trenor 2020). Although this chapter highlights the benefits and drawbacks of plastics, it should be
emphasized that all materials, also including metals, ceramics, paper and cardboard among others,
have their pros and cons. All material processing and use benefit from applying the circular model.

# Springer Nature Switzerland AG 2021 489

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_9
490 9 Plastics and Sustainability

Fig. 9.1 The versatile contribution of plastics to sustainable society

Fig. 9.2 Two challenges for turning plastics into sustainable materials; note the concepts circular and linear, which
refer to (i) circular economy; a systematic approach to develop businesses, society and environment (e.g. by minimizing
the waste which includes also ‘useless energy’ and CO2). (ii) This is different from the linear model take-make-waste

9.2 Contribution of Plastics to Sustainable Society

Plastics in present form contribute to a sustainable society, quality of life and reduced global warming
(Andrady 2015). However, plastic production and design and our consumption must be adjusted to a
circular economy and unnecessary single-use products should be avoided to contribute to a fully
sustainable society. Production of plastic products uses far less energy compared to traditional
materials, and plastics are in general more energy efficient in the use phase than the alternative
materials. Several large studies show that substitution of plastics by second best alternatives would
cause a substantial increase in energy consumption and thus contribute to further global warming
(Billiet and Trenor 2020; Pilz et al. 2005, 2010).
A large European study calculated that broad replacement of plastics with other materials would
increase the energy consumption and greenhouse gas emissions by, respectively, 2140 million GJ and
110 Mt CO2 equivalents per year (Pilz et al. 2005). Furthermore, the plastic products on the market
9.2 Contribution of Plastics to Sustainable Society 491

today enable energy savings corresponding to 53 Mt of crude oil per year (Pilz et al. 2010). An
investigation by the Plastics Division of the American Chemistry Council revealed that replacing
common plastic packaging by aluminium, steel, glass or paper in the United States would lead to 69%
higher CO2 emission, 90% higher energy and 480% higher water consumption (Franklin Associates
2018). A life cycle assessment of grocery carrier bags performed by the Danish Environmental
Protection Aagency evidenced plastic bags as by far the most sustainable alternative (Ministry of
Environment and Food of Denmark 2018). An unbleached paper bag needed to be used 43 times and
organic cotton bag 20,000 times to come down to same impact as plain LDPE bag. The production of
plastic bags especially required less fresh water and released less CO2 than the alternative
grocery bags.
Plastic packaging is crucial for increasing the shelf life of food and reduces the food waste
(Andrady 2015). A cucumber wrapped in plastic packaging can stay fresh for 14 days, whereas an
unwrapped cucumber only holds for approximately 5 days. Similar effects on the shelf life have been
reported for other vegetables, dairy products, meat and fish. Plastic food packaging has been
estimated to contribute to 20% reduction in food losses, which corresponds to reduction of CO2
emissions by 190 Mt per year. A case study of egg packaging prepared from 100% post-consumer-
recycled PET had 15% lower carbon footprint than pulp egg packs, and, in addition, the plastic
solution yielded savings including 40% less lorries, 40% less warehouse space and no fresh water
consumption (Davis et al. 2011). Agricultural plastics can significantly increase the crop production
per square meter, improve the food quality, decrease water consumption and reduce the release of
pesticides or fertilizers. Furthermore, vegetables can be grown out of the season and outside the
natural regions. Plastic packaging allows hygienic transport and helps to preserve the already
produced food.
Only 4% of fossil resources are used for plastic production. At the same time, plastics significantly
contribute to saving, transporting and producing energy. Plastic packaging is light, which saves
energy during the transportation. On average, only 1–3% of the mass of a packaged product comes
from the plastic packaging. As an example, a glass jar for coffee constitutes 36% of the mass of the
product, whereas a plastic pouch constitutes less than 4% of the total product mass (BPF 2019). Cars
and planes with plastic parts are lighter and consume less fuel. Replacing traditional materials in cars
by plastics gives an average cut in fuel consumption of 750 L per lifespan of 150,000 km. The annual
fuel consumption of cars is reduced by 12 Mt in Europe and CO2 emissions by 30 Mt thanks to
plastics (BPF 2019; PlasticsEurope 2020).
The energy used for heating and cooling of buildings is 40% of the total energy distributed. Pipes,
isolation, energy-efficient windows and doors, solar panels, wind turbine rotors and electrical cables
are made of plastics. Effective plastic insulation materials significantly lower the energy loss from
buildings. The energy to produce plastic insulation can be recouped after only 1 year of use, and over
its lifetime, plastic insulation saves energy between 150 and 200 times the energy used in its
manufacture (PlasticsEurope 2020). Altogether it has been estimated that the produced plastic
insulations have resulted in 9500–19,900 million GJ of energy savings over their lifetime,
corresponding to reduced emissions of 536–1120 Mt of CO2 equivalents (Pilz et al. 2010). Plastic
products are also essential in the distribution of energy (e.g. as insulations of HVDC cables) and for
the generation of clean energy by wind turbines, solar cells, batteries and fuel cells.
Both simple and advanced plastic products used in the healthcare, inside and outside the human
body, save and improve lives of millions of people on a daily basis. Plastics are found everywhere and
are crucial for the diagnostics and treatment of patients and the protection of healthcare workers
(e.g. gloves and face shields). Examples of products are sterile syringes, catheters, blood bags, heart
valves, prosthesis, hearing aids, sutures and drug capsules to advanced products such as temporary
scaffolds for tissue engineering and artificial organs. Plastics are also an essential part of our leisure
and sport activities leading to increased comfort and safety.
492 9 Plastics and Sustainability

To maximize the benefits and minimize the negative impacts of plastics, it is important to consider
the taxonomy reduce, reuse and recycle (A European Strategy for Plastic in Circular Strategy 2018)
or eliminate, innovate and circulate (Ellen MacArthur Foundation 2017). These strategies aim at
eliminating the plastics that are unnecessary, reusing or recycling to keep the material in the economy
and not out in the environment. Furthermore, innovation is needed to make all plastics reusable,
recyclable or compostable. This will be further discussed in following sections.

9.3 Bio-based Materials

One step towards more sustainable materials is the development of plastics from renewable resources
and the efficient use of waste and by-products as a resource. Bio-based plastics include materials
produced synthetically from biomass-derived monomers, vegetable oils, carbon dioxide or
biopolymers (Isikgor and Becer 2015; Mohanty et al. 2000). Different carbohydrates, proteins and
lignin are promising resources for production of new plastic materials. They can be used as such or as
chemically modified or catalytically degraded forming a platform of green chemicals (Isikgor and
Becer 2015). Essentially all traditional plastic products including thermoplastics, thermosets,
elastomers and hydrogels can be produced from renewable resources. However, it is not enough
just to use bio-based resources to fully reach a sustainable production and use of plastics. Simulta-
neously with the production of new materials, the sustainability of these new products and production
processes must be ensured (Zhu et al. 2016). A brief introduction of the different types of bio-based
materials follows.
The concept bio-based is not synonymous to biodegradable. It is the chemical structure and not the
origin that determines the susceptibility of a material to degrade. Biodegradable plastics are produced
from both bio-based resources and petroleum. More generally, all plastics can be divided into the
following four groups (Fig. 9.3): (i) traditional petroleum-based durable plastics; (ii) petroleum-based
biodegradable plastics; (iii) bio-based durable plastics, including ‘drop-in’ plastics that have identical

Fig. 9.3 Classification of plastics based on origin and durability and examples of common polymers for each group
9.3 Bio-based Materials 493

technical properties with petroleum-based counterparts; and (iv) bio-based and biodegradable
plastics.
A bio-based material is not automatically more sustainable than a petroleum-based material. It is
important to consider the whole life cycle from raw material to consumption of energy and fresh
water during different production stages and finally, but not least, the possible recycling routes. A
biodegradable material is not automatically more environmentally friendly than the durable (stable)
material. The optimum service life and end-of-life management option depends on the application. A
stable material has a potentially long lifetime, which benefits sustainability. Some applications
require stable materials with long service life, whereas some products are used for only short periods
of time but can be recycled and used in new products. For some products, biodegradability is part of
the aimed function. End-of-life management and degradation of plastics are further discussed in
Sects. 9.4 and 9.5.

9.3.1 Thermoplastic Materials Derived from Bio-based Monomers

Several commercial thermoplastic materials, such as the aliphatic polyester, polylactide (PLA) and
drop-in plastics such as bio-poly(ethylene terephthalate) (bio-PET) and bio-polyethylene (bio-PE),
are produced from plant-based monomers. For drop-in plastics, the market and applications are
already present. Furthermore, after production of the monomers, existing infrastructure for synthesis
and processing of petroleum-based products can be used. PLA, bio-PE and bio-PET are commercially
produced from starch or sugar-rich plants, such as sugar cane, sugar beet, wheat or corn. The
production is based on a multistep process starting from the production of glucose. Glucose is then
turned into, for example, lactic acid for the production of cyclic lactide monomer which is
polymerized into PLA, ethene for bio-PE and ethylene glycol for bio-PET. The production and the
properties of PLA have been reviewed by Inkinen et al. (2011), and Siracusa and Blanco (2020) have
presented a similar text for bio-PET and bio-PE.
PLA is an example of bio-based and biodegradable plastic that can be mechanically or chemically
recycled or alternatively disposed by industrial composting. The potential applications range from
packaging to biomedical bioresorbable products to textile fibres (Auras et al. 2004; Lim et al. 2008;
Ulery et al. 2011). Bio-PE and bio-PET basically have same chemical structure, properties and
applications as the petroleum-based counterparts. Bio-PE and bio-PET are stable polymers and can be
mechanically recycled together with petroleum-based PE and PET, incinerated or chemically
recycled to original monomers (bio-PET) or to fuels (bio-PE). Bio-PE is commercially produced
from sugar cane-derived ethanol that is dehydrated to ethene (Zhang and Yu 2013). Current bio-PET
is produced from bio-based ethylene glycol that is polymerized with petroleum-based terephthalic
acid. Life cycle assessment has shown 20–50% reduction in greenhouse gases compared to fully
petroleum-based PET (Pang et al. 2016). Current research aims at developing methods to produce
also terephthalic acid from bio-based resources (Pang et al. 2016). A potential bio-based alternative to
PET is polyethylene furanoate (PEF), where the terephthalic acid is replaced with bio-based
2,5-furandicarboxylic acid. PEF shows lower gas permeability than PET, which makes it more
suitable for barrier application than PET. The existing infrastructure for polymerization and recycling
of PET can potentially be used for production and recycling of PEF, a possible fact which could
advance the commercialization (Burgess et al. 2014). A life cycle assessment showed a reduction of
greenhouse gas emissions by 55% if petroleum-based PET is replaced by fully bio-based PEF
(Earhart et al. 2012). Another life cycle assessment showed that the replacement of PET by PLA
could reduce the greenhouse gas emissions by 40% (Shen et al. 2012).
Succinic acid is produced commercially in relatively large quantities by fermentation of glucose
and can be used for the synthesis of poly(butylene succinate) (PBS) and poly(butylene succinate-co-
494 9 Plastics and Sustainability

butylene adipate) (PBSA), both being aliphatic biodegradable polyesters (Fujimaki 1998). PBS can
be processed by melt extrusion, injection moulding and film blowing, and its current commercial
applications include barrier films and biodegradable packaging.

9.3.2 Biopolymer-Based or Biopolymer-Derived Thermoplastics

Chemically modified cellulose in the form cellulose acetate was first prepared in 1865 and is still a
large-scale commercial material (Buchanan et al. 1993). The most known application is as material
for cigarette filters, but cellulose acetate is also used as in fabrics, transparent films, coatings and
many other applications. There are several other large-scale cellulose derivatives, such as cellulose
ethers and water-soluble products such as carboxymethyl cellulose and hydroxyethyl cellulose
(Isikgor and Becer 2015). Thermoplastic starch-based materials are among the most successful
commercial bio-based and biodegradable materials. Thermoplastic starch can be produced by extru-
sion of starch in the presence of water and plasticizers (Bastioli 1998), which destructs the starch
granules and yields a melt-processable gelatinized starch material. Thermoplastic starch can be
processed by the conventional polymer processing methods (cf. Chap. 8). The poor mechanical
properties and moisture resistance of thermoplastic starch is a drawback, and thermoplastic starch is
therefore commonly blended with other biodegradable polymers such as aliphatic polyesters to gain a
useful property profile while retaining biodegradability and compostability (Averous 2004). In
addition to thermoplastic starch materials (Wu et al. 2017), starch granules can be blended with
thermoplastic polymers (Xu et al. 2016) or can also be covalently modified (Wang et al. 2020; Yang
et al. 2013). Proteins, such as wheat gluten, are an interesting resource for production of plastics
(G€allstedt et al. 2004).
Other interesting biopolymers include bacterially produced polyhydroxyalkanoates (PHA), such
as poly(3-hydroxybutyrate) (PHB) and bacterial cellulose. PHA is produced by several
microorganisms as an energy reserve (Amass et al. 1998). The PHA materials are biodegradable
and suitable for thermoplastic applications potentially replacing commodity polymers such as
polypropylene (PP), but the production is currently not cost-effective for volume applications.
However, in several biomedical applications, PHA is potentially very useful (Michalak et al. 2017).

9.3.3 Bio-based Thermosets and Coatings

The interest for bio-based thermosets has rapidly increased during the last years. Vegetable oils,
which can be recovered from plants, such as soybean and sunflower, at industrial scale are among the
forerunners for the bio-based thermoset and coating materials (McConnell 2008; Campanella et al.
2011; Dai et al. 2015; Nameer et al. 2019). Vegetable oils for the production of resins, coatings and
paints have a long history. In addition to recovering vegetable oils from plants, the production of
UV-curable resins from used cooking oil has been demonstrated by Wu et al. (2020). Vegetable oils
are also potential raw materials for production of thermoplastic elastomers (Cordier et al. 2008).
Some commercial polyamides, polyamide 11, polyamide 4,10 and polyamide 6,10, are produced
from castor oil. Potential future vegetable oil-based fatty acid-derived products include elastomers,
epoxy resins, polyamides and long-chain aliphatic polyesters. The long-chain aliphatic polyesters are
also called for polyethylene-like polyesters, because of the long aliphatic hydrocarbon segments
between the ester bonds giving polyethylene-like properties. To avoid competition with food produc-
tion, interest has turned towards engineering algae for the generation of unsaturated fatty acids.
9.3 Bio-based Materials 495

Recent studies present a number of fully bio-based unsaturated polyester thermosets (Dai et al.
2015; Panic et al. 2017; Xu et al. 2018). Isosorbide and itaconic acid are examples of promising bio-
based monomers for production of unsaturated polyesters (Xu et al. 2018; Trotta et al. 2019). Lignin
and lignin-derived aromatic compounds are also rendering interest as raw materials for the production
of aromatic thermosets (Jawerth et al. 2017; Gioia et al. 2018; Xu et al. 2019; Moreno and Sipponen
2020).

9.3.4 Production of Polymers from Waste–Greenhouse Gases

Synthesis of plastic materials from waste–greenhouse gases, e.g. carbon dioxide, is very promising
and can provide both economic and environmental benefits, in accordance with a circular modelled
society (Grignard et al. 2019). The main principal is that CO2 is copolymerized with epoxides, such as
propylene oxide. Depending on the catalyst, this can yield different polymers ranging from
alternating copolymers to block copolymers with ester, ether and carbonate units. An interesting
product is a low molar mass polyol for further fabrication of polyurethane (Lee et al. 2012) or high
molar mass polycarbonates (Hauenstein et al. 2016) for potential applications as a binder or as a rigid
plastic. The process for carbon dioxide-based materials is close to commercialization and has the
potential to generate new materials with 30–50% of mass coming from CO2. Some promising results
demonstrated copolymerization of propylene oxide and carbon dioxide from power station in the
United Kingdom, illustrating the potential for the recycling of industrial carbon dioxide emissions
(Chapman et al. 2015). A life cycle assessment showed that partial replacement of propylene oxide
with carbon dioxide reduced the emissions of greenhouse gases and fossil fuel consumption (Von der
Assen and Bardow 2014).

9.3.5 Challenges and Opportunities for Bio-based Materials

One fear with the production of bio-based materials is the competition with food production. Instead
of using editable biomass as raw material, processes are needed to more effectively upcycle mixed
waste products to monomers for polymer production. Current land use for production of bio-based
plastics is still small. It was calculated that in the European Union, replacing 1% of petroleum-based
materials with bio-based materials would require ca. 1% of the land currently used for wheat
production (Patel et al. 2005). A sustainable solution to this problem can be the use of cellulose,
which has the potential to replace starch as a raw material for bio-based material production.
The new bio-based materials may represent totally new chemical structures and materials or they
may be so-called drop-in plastics. For drop-in plastics, it is necessary to reach similar properties and
price levels as those of the petroleum-based counterparts. The existing industrial infrastructure and
the current applications are a big help to reach commercialization for the bio-based polymers. For
truly new materials, the demands can be even higher, and the presence of improved properties is
likely to be a requirement in order to be commercialized. A large diversity of materials could, on the
other hand, counteract the recycling efforts. As the price of bio-based materials is generally higher, it
could be easier to start by developing materials for high-value applications rather than focusing on
large-volume applications such as packaging. It should also be kept in mind that bio-based materials
face the same recycling challenges as their petroleum-derived counterparts (Zhu et al. 2016; Zhang
et al. 2018).
496 9 Plastics and Sustainability

9.4 End-of-Life Management: From Waste to Resource

More than 8 Mt of plastics end up in oceans each year (Jambeck et al. 2015). This equals to one
garbage truck per minute, and the rate is accelerating. Responsible and innovative actions are needed
to change this wasteful and environmentally harmful situation. Following hierarchy has been
suggested for waste management:
(1) Prevention including manufacturing techniques that reduce waste and production of thinner and
lighter products.
(2) Reuse.
(3) Material recycling.
(4) Other recovery such as energy recovery. Organic recycling or composting is possible for
biodegradable materials classified as compostable.
(5) Disposal in landfills as the very last option, which should be avoided.
In 2015, the European Commission adopted the European Action Plan for a circular economy,
where plastics were identified as a key priority. A goal was also set to ensure recyclability of all
plastic packaging by 2030. This was followed by a European Strategy for Plastics in a Circular
Economy (2018) and many national plastic strategies. Numerous reports and investigations show the
environmental and economic benefits of plastic recycling. In addition to solving the problem with
plastic pollution, material recycling reduces the need for new raw materials and has the potential to
save significant amounts of energy and to reduce the CO2 emission. A recycling efficiency of 50% for
manufacturing a product with polyamide-66 fibres yields a CO2 reduction of 48% (SOU 2018). The
production of 1 kg petroleum-based PE requires 2 kg fossil fuel, whereas the production of 1 kg
recycled PE requires only 0.5 kg fossil fuel (SOU 2018). Cradle-to-grave life cycle analysis showed
that greenhouse gas emissions and fossil fuel consumption can be reduced by 13% and 56% for PET
bottles produced from recycled material with reference to that of virgin PET (Benavides et al. 2018).
In addition to the resource savings and the economic benefits, an increased plastic recycling also has
an impact on the solution of the serious problems with plastic pollution; the current situation is that
several million tons of plastic waste ends up in oceans and in other natural environments every year.
In 2016, the Ellen MacArthur Foundation launched the New Plastics Economy initiative, which
has inspired many businesses across the plastic value chain (Ellen MacArthur 2016). This in
combination with different national and international directives is starting to make a real difference
(Ellen MacArthur 2017). However, plastics are a highly versatile group of materials used in all areas
of our lives. This is a strength of plastic materials from the application point of view, but on the other
hand limitation concerning the recycling (Garcia and Robertson 2017). Multiple solutions from
closed-loop recycling to open-loop recycling, where plastics are utilized as raw materials for design
of products differing from the original ones, are thus required (Ragaert et al. 2017).
Recycling has been on the agenda for many years, but only a fraction of plastics is recycled to
products with at least the same value as in the original product. Plastic recycling is more problematic
than recycling of, e.g. metals and glass. An obstacle for the recycling of plastic through ‘simple’
re-processing is the need for pure and sorted materials with good quality and without presence of
harmful additives. As an example, polyethylene is a polymer with a simple repeating unit structure (
CH2 –), but yet it still comes in a wide variety of qualities and different molar mass, molecular
architecture (linear, short-chain or long-chain branched and crosslinked) and additives, formulated
for different manufacturing processes and for different applications. There is also a limit to how many
times a material can be reprocessed due to the deterioration quantum of each cycle. This approach is
also less feasible for contaminated, partly degraded or multicomponent materials, such as multilayer
packaging or many textiles such as carpets. Landfill is still often the common disposal place for
plastics around the world. However, landfilling space is limited, and it is not a sustainable solution to
9.4 End-of-Life Management: From Waste to Resource 497

Fig. 9.4 Possible end-of-life options for polymeric materials

deal with the waste. One way to increase the recycling volumes is to design materials and products
that are more suitable for recycling. This could include labelling, the use of known material
compositions and favourable chemical structures. Multicomponent material solutions should be
either avoided or designed for easy disassembling.
Currently 29.1 Mt post-consumer plastics are collected in Europe per year. After sorting, only
ca. 30% of this waste is send to recycling, while the rest is used for energy recovery or ends up in
landfill. Finally, ca. 4 Mt of recyclate is produced in Europe. Almost half of the recyclate is used in
building and construction and one quarter in packaging applications. In Europe, 8 Mt of plastic still
ended up in landfill in 2015, which corresponds to 100 million barrels of oil (PlasticsEurope 2020).
Worldwide, almost half of the produced plastics ends up in landfill, corresponding to 150 Mt each
year. Plastic recycling can be divided into three main loops (Ellen MacArthur Foundation 2019); two
of these are ideally fully circular leading to, in best case, materials and products with equal value as
the polymer used in the original products (Fig. 9.4). These two loops are the polymer loop and
monomer loop, more commonly called mechanical recycling and chemical recycling through depo-
lymerization to original monomers. The third loop is the molecule loop or chemical recycling to other
chemicals than the original monomers. In addition, polymers can be recycled to fuels or incinerated to
recover energy. Some plastics can be composted in industrial compost to produce fertile soil, but
processes should also be developed for their mechanical and chemical recycling to keep the material
value.

9.4.1 Polymer Loop: Mechanical Recycling

Mechanical recycling through thermal reprocessing is the usual way to recycle plastics into new
materials. In the recycling hierarchy, it is also the preferred route for high-quality non-contaminated
plastic waste due to the lowest number of steps required when the original polymer backbone is kept
invariant. This generally leads to the lowest environmental footprint. Mechanical recycling can be
divided into primary and secondary recycling. Primary recycling of clean in-plant waste with well-
498 9 Plastics and Sustainability

known composition and properties is relatively easy and generally provides closed-loop recycling to
products of the same quality and value as the original product. The recycling of post-consumer
plastics faces more hurdles. The plastic waste needs to be collected, separated, washed, grinded,
re-stabilized and thermally reprocessed. The quality of the collected waste is generally unknown
including composition and possible presence of harmful additives and other contaminants
(Hakkarainen et al. 2003). The material may undergo degradation during the different stages
(processing and service) leading to reduced molar mass and potentially inferior properties such
as brittleness, presence of contaminants, discoloration and bad small (La Mantia and Vinci
1994; Gr€oning and Hakkarainen 2001; Gr€oning and Hakkarainen 2002; Xiang et al. 2002; Qian
et al. 2011).
A further complication is the large number of different product formulations even for the same
polymer type. During the first production cycle, the material viscosity and properties were tailored
considering both the anticipated processing method and the application. It is therefore difficult to
retain same quality for mechanically recycled products where different material formulations are
mixed and the recyclate may not have the optimum rheological properties for a particular processing
method (Tzoganakis et al. 1988). Irrespective of the measures taken, the material may also be
contaminated with small amounts of other materials leading to mechanical brittleness, because of
the generally low compatibility of different polymers (Ignatyev et al. 2014; Elhrari 2018). These
negative effects can be counteracted by the addition of compatibilizers, which, negatively, increases
the cost and the material complexity (Maris et al. 2018; Gedde and Hedenqvist 2019). Mechanical
recycling can theoretically produce materials and products of equal quality as the original products,
but more often due to practical limitations described above, it leads to downcycling. One way to have
a better control of the product quality is to blend virgin material with a smaller amount of recycled
material. Successful mechanical recycling is used for the production of parts used in in automotive
industry, where shredded automotive parts are used for the production of composites.
Multipolymer products are a group of materials that are difficult to recycle. Mechanical recycling
of mixed plastics usually leads to inferior properties, and it could also lead to degradation of some of
the components as processing temperature needs to be selected based on the materials with highest
processing temperature. One way to recycle multipolymer materials is through dissolution/precipita-
tion, where the different polymers are selectively dissolved and then precipitated by the addition of
non-solvents, which has been successfully applied for blends of the common volume plastics, LDPE,
HDPE, PP, PS, PVC and PET (Achilias et al. 2009; Schut 2001). Harsh solvents are in many cases
required for this process, although the solvents can be recycled. Purification of materials from odours,
pigments, dyes, degradation products, flame retardants, plasticizers and other additives can also be
achieved by dissolution and precipitation (Côté 2018; Layman et al. 2017a, b). Several companies
have already commercialized dissolution/precipitation processes, which has made possible the
recycling of multipolymer materials. Commercial upcycling of PP carpet fibres to food packaging
applications was made possible by an efficient supercritical extraction process (PureCycle-
Technologies 2019).
To promote mechanical recycling, ‘design for recycling’ is an important concept. It was first used
in automotive industry for design of components that were easy to disassemble, but it has now
expanded to packaging and other industries, such as the carpet manufacturing. The aim is to through
product design and material selection enable easier recycling while maintaining the product quality
(Plastic Recyclers Europe 2018a). Design for recycling is also heavily promoted by several directives
and extended producer responsibility schemes.
9.4 End-of-Life Management: From Waste to Resource 499

9.4.2 Monomer and Molecule Loops: Chemical Recycling

According to Plastics Recyclers Europe (2018a, b): ‘Chemical recycling is a process which converts
polymeric waste by changing its chemical structure to produce substances that are used as raw
materials for the manufacture of new products, which excludes production of fuels or means of
energy generation’.
In the recycling hierarchy, chemical recycling comes after mechanical recycling. Industrial
application of chemical recycling is hitherto limited. However, this route is rendering increasing
attention and is expected to grow in importance, parallel with the development of more efficient
processes and depolymerization catalysts (Rahimi and Garcia 2017). Chemical recycling allows, in
its ideal and simplest form, depolymerization of materials back to original monomer (monomer loop),
which can be repolymerized and reprocessed to materials of equal value compared to the original
materials. This route is most easily accomplished when the polymer has hydrolyzable bonds such as
ester or amide bonds. Another target for chemical recycling is the production of other valuable
chemicals (molecule loop).
Processes utilized for chemical recycling include thermochemical and catalytic conversion pro-
cesses (Ragaert et al. 2017). PET is an example of a volume plastic that is suitable for chemical
recycling, and several established chemolysis processes are available, including hydrolysis,
methanolysis, glycolysis and aminolysis (Geyer et al. 2016). Pyrolysis is a suitable process for
materials that are difficult to depolymerize, such as polyolefins or mixed plastics and multicomponent
materials that pose challenges for mechanical recycling (Al-Salem et al. 2010). Ragaert et al. (2017)
presented a review of the different chemical recycling technologies, including processes such as
chemolysis, pyrolysis, fluid catalytic cracking, hydrogen technologies, gasification and catalytic
pressureless depolymerization.
Chemicals and intermediates from the molecule loop can be used for ‘design from recycled’,
where the produced chemicals are a starting point for new product development. Researched
approaches include utilization of produced chemicals or oligomeric products for production of
plasticizers (Langer et al. 2015; Yang et al. 2015; B€ackstr€ om et al. 2019) and compatibilizers
(Wu and Hakkarainen 2015). PET has been converted to methacrylated or acrylated oligoesters for
fabrication of UV-curable coatings (Nikles and Farahat 2005). As a final step, plastic waste can be
carbonized to value-added graphitic carbon (Choi et al. 2017; Adolfsson et al. 2018).

9.4.3 Additive Manufacturing as Enabling Tool for Plastic Recycling

Additive manufacturing (AM) or 3D printing has emerged as an industrially viable processing

method. AM offers interesting possibilities when it comes to mechanical recycling and could be an
enabling and complementary tool for polymer recycling, not least because it can easily work with
small material volumes and production of personalized products (Chong et al. 2015). The possibility
to 3D print plastic waste has been demonstrated, but the information concerning the influence on
material properties and long-term performance of the produced products is still limited. A few studies
compared the properties of 3D printed parts made of virgin and recycled polylactide (PLA)
(Anderson 2015) and carbon fibre-reinforced PLA before and after recycling (Tian et al. 2017).
These studies showed that the recycled product had only slightly ‘reduced’ properties, e.g. the tensile
500 9 Plastics and Sustainability

strength of PLA decreased by 11% and the hardness by only 2.4%, which is promising for further
utilization of 3D printing as a recycling tool. The mechanical properties of pre-consumer recycled PP
with different fillers were also shown to be adequate (Stof and Pickering 2018). The cost of producing
recycled HDPE filaments for 3D printing was 2–3 USD as compared to the commercial virgin HDPE
filament that was sold for 38 USD (Kreiger et al. 2014).

9.4.4 Dynamic Covalent Chemistry as Enabling Tool for Plastic Recycling

Reversible or dynamic covalent chemistry and polymers with covalent adaptable networks have
appeared as a potential approach to produce repeatedly recyclable plastics with high performance
(Kloxin and Bowman 2013; Durand-Silva and Smaldone 2020; van Zee and Nicolaÿ 2020). Acetal
linkages, disulphide bonds, boronic ester bonds, Schiff based (imine chemistry), urethane and ester
bonds are examples of dynamic covalent bonds that can be utilized. The introduction of dynamic
covalent bonds has the potential to improve the properties of thermoplastics collected for recycling,
and it could make traditionally non-recyclable cross-linked thermosets re-processable and recyclable.
Thermally reversible covalent adaptable networks can be subjected to repeated thermal processing as
thermoplastic polymers and their behaviour mimic traditional thermosets once cooled due to forma-
tion of cross-links. The possible benefits of a cross-linked material over a thermoplastic material are
better durability and adjustable properties through control of the degree of cross-linking. This route
can thus produce materials with properties ranging from elastomeric to those of stiff materials with
high modulus and tensile strength by adjusting the concentration of the dynamic covalent bonds.
A study on epoxy thermosets introduced the concept of vitrimers, a subcategory of covalent
adaptable networks relying on associative bond exchange (Montarnal et al. 2011). Poly(butylene
terephthalate) polyester vitrimers were made by incorporation of glycerol, and these materials
showed excellent processability, and the mechanical properties remained intact up to four cycles
according to DMTA data (Zhou et al. 2019). Conversion of conventionally produced polyurethane
thermosets to covalent adaptable networks has also been demonstrated by Sheppard et al. (2020).
Dioxaborolane metathesis is a type of dynamic covalent chemistry that is applicable to polymers with
carbon–carbon backbone, such as polyolefins and polystyrene (R€ ottger et al. 2017). As an example,
modified polystyrene had excellent environmental stress crack resistance and did not crack even at
flexural stress of 30 MPa, whereas thermoplastic polystyrene broke only after 2 min at a flexural stress
of 28 MPa. Dynamic covalent chemistry can thus be utilized to improve the properties of recycled
plastic products enabling recycling of thermosets and also to enhance the properties of commodity
polymers such as polystyrene.

9.5 Degradation of Plastics in Natural Environments

Plastic pollution is a large environmental problem, and microplastics can nowadays be found in most
natural environments on land, in freshwater and oceans (Thompson et al. 2004; Rochman 2018;
UNEP 2017). Most plastics are persistent in the environment; still abiotic and biotic environmental
factors, such as sunlight, humidity, heat, mechanical wear and microorganisms, cause changes in the
plastic materials, and with time, degradation and fragmentation take place (Roy et al. 2011;
Albertsson and Hakkarainen 2017). The degradation rate of plastics thrown in nature could vary
from months to more commonly years or hundreds of years depending on the interplay between
the specific environment and the physical and chemical structure of the materials (Zumstein et al.
2019; Min et al.2020; Napper and Thompson 2019). The prediction of environmental degradation
9.6 Summary 501

rate is complicated by the significantly different degradation environments, depending whether,

e.g. the material ends up in ocean or is buried in soil or subjected to sunlight and the geological
location. The degradation rate is further influenced by the material characteristics: polymer and its
molar mass, molecular architecture and morphology and its additives. The size and shape of the
plastic material are also important. Even seemingly small material or formulation modifications could
significantly influence the degradation potential in specific environments. Incomplete degradation in
natural environments could lead to release of microplastics or harmful degradation products (Roy
et al. 2011; UNEP 2015; The Guardian 2016). The impact of this on the environment and the potential
ecological risks are still uncertain (Haider et al. 2019; Degli Innocenti and Breton 2020).
The use of biodegradable polymers have been suggested as a solution to the waste problem, but the
degradation of biodegradable or compostable polymers can only be guaranteed under specific
conditions such as industrial compost (Albertsson and Hakkarainen 2017). The property ‘biodegrad-
ability’ is always connected to a specific environment, and it does not guarantee degradation in other
natural environments (Zumstein et al. 2019); natural environments offer huge variety of conditions
and environmental factors, and irrespective of polymers potential to degrade, fast and complete
degradation cannot be guaranteed under all environmental conditions.
Hence, even though biodegradable or compostable plastics should not be considered as universal
solutions to the plastic waste problem, they offer large added value in specific applications. For these
applications, materials can be designed for specific disposal route or degradation environment with
relatively well-defined and established conditions. For example, bioresorbable materials for biomed-
ical applications are designed to degrade under specific conditions in the human body (Albertsson and
Hakkarainen 2017; Shen et al. 2020). Compostable plastic materials have high value in applications,
where plastic material is contaminated by bio-waste, including plastic bags for garden or food waste
(Zumstein et al. 2019). For these products, organic recycling through industrial composting is an ideal
end-of-life management option. Other products like fishing nets have high risk of ending up in nature
and could with benefit be designed with an enhanced environmental degradation rate (Iwata 2015).

9.6 Summary

Polymeric materials are present in essentially all areas of modern life. During the service life,
polymeric materials largely contribute to a sustainable society and lower greenhouse gas emissions
by serving in different applications, such as packaging, medical applications, insulation and energy
production. Recent research illustrates that emissions can in many cases be further reduced by
changing from petroleum-based to bio-based resources for the production of polymers. However,
more work is needed to evaluate the whole life cycle of the newly developed products. Unfortunately,
the end-of-life management of polymeric materials still requires further development. Improvements
in material and product design and recycling technologies are required to further improve the
sustainability of polymeric materials and to keep them in the cycle as long as possible. Changing
the perception from waste to resource will help to tackle the current problem with plastic waste and
especially the release of large amounts of plastics in the natural environments and landfills.
Concerning biodegradable plastics, it is very difficult to guarantee complete degradation of plastic
products in natural environments. However, biodegradable plastics still play an important role in
specific applications.
502 9 Plastics and Sustainability

9.7 Exercises

9.1. (a) Is replacement of plastics with other materials a realistic and sustainable solution?
(b) What are the main benefits of plastics during service life?
(c) What are the two large challenges to further improve the sustainability of plastic materials?
9.2. (a) Are bio-based materials also biodegradable?
(b) Are bio-based materials always sustainable?
9.3. (a) Why is recycling of plastics challenging? Why are several solutions needed to solve the
plastic waste problem?
(b) Explain shortly what is meant with the following recycling processes, when are they suitable
and what is the outcome: mechanical recycling (polymer loop), chemical recycling (mono-
mer loop) and chemical recycling (molecule loop).
9.4. (a) Are biodegradable materials solution to plastic waste problem?
(b) It is difficult to guarantee degradation of plastics in natural environments. Why?

References

A European Strategy for Plastics in a Circular Economy. (2018). https://ec.europa.eu/environment/circular-economy/

pdf/plastics-strategy-brochure.pdf
Achilias, D. S., Giannoulis, A., & Papageorgiou, G. Z. (2009). Polymer Bulletin, 63, 449.
Adolfsson, K. H., Lin, C.-F., & Hakkarainen, M. (2018). ACS Sustainable Chemistry & Engineering, 6, 11105.
Albertsson, A.-C., & Hakkarainen, M. (2017). Science, 358, 872.
Al-Salem, S., Lettieri, P., & Baeyens, J. (2010). Progress in Energy Combustion Science, 36, 103.
Amass, W., Amass, A., & Tighe, B. (1998). Polymer International, 17, 89.
Anderson, I. (2015). 3D printing and Additive Manufacturing, 4, 110.
Andrady, A. L. (2015) Plastics and Environmental Sustainability, John Wiley & Sons, Inc. Hoboken, New Jersey, USA.
Auras, R., Harte, B., & Selke, S. (2004). Macromolecular Bioscience, 4, 835.
Averous, L. (2004). Journal of Macromolecular Science-Polymer Reviews, C44, 231.
B€ackstr€om, E., Odelius, K., & Hakkarainen, M. (2019). ACS Sustainable Chemistry & Engineering, 7, 11004.
Bastioli, C. (1998). Polymer Degradation and Stability, 59, 263.
Benavides, P. T., Dunn, J. B., Han, J., Biddy, M., & Markham, J. (2018). ACS Sustainable Chemistry & Engineering, 6,
9725.
Billiet, S., & Trenor, S. R. (2020). ACS Macro Letters, 9, 1376.
BPF, British Plastics Federation. (2019). https://www.bpf.co.uk/Press/Oil_Consumption.aspx
Buchanan, C. M., Gardner, R. M., & Komarek, R. J. (1993). Journal of Applied Polymer Science, 47, 1709.
Burgess, S. K., Leisen, J. E., Kraftschik, B. E., Mubarak, C. R., & Koros, W. J. (2014). Macromolecules, 47, 1383.
Campanella, A., La Scala, J. J., & Wool, R. P. (2011). Journal of Applied Polymer Science, 119, 1000.
Chapman, A. M., Keyworth, C., Kember, M. R., Lennoxc, A. J. J., & Williams, C. K. (2015). ACS Catalysis, 5, 1581.
Choi, D., Jang, D., Joh, H.-I., Reichmanis, E., & Lee, S. (2017). Chemistry of Materials, 29, 9518.
Chong, S., Chiu, H.-L., Liao, Y.-C., Hung, S.-T., & Pan, G.-T. (2015). Chemical Engineering Transactions, 45, 1669.
Cordier, P., Tournilhac, F., Soulie-Ziakovic, C., & Leibler, L. (2008). Nature, 451, 977.
Côté, R. (2018). WO patent, WO2019095078A1.
Dai, J., Ma, S., Wu, Y., Han, L., Zhang, L., Zhu, J., & Liu, X. (2015). Green Chemistry, 17, 2383.
Davis, P., Law, P., Calver, P., & Quarshie, R. (2011). Sustainable manufacturing guide. https://www.bpf.co.uk/
sustainable_manufacturing/ Guide.aspx.
Degli Innocenti, F., & Breton, T. (2020). ACS Sustainable Chemistry & Engineering, 8, 9239.
Durand-Silva, A., & Smaldone, R. A. (2020). ACS Central Science, 6, 836.
Earhart, A. J. J. E., Faaji, A. P. C., & Patel, M. K. (2012). Energy & Environmental Science, 5, 6407.
Elhrari, W. (2018). Research & Development in. Materials Science, 7, 791.
Ellen MacArthur Foundation. (2016). The New plastics economy-rethinking the future of plastics. https://www.
newplasticseconomy.org/about/publications/report-2016
Ellen MacArthur Foundation. (2017). The New plastics economy- catalysing action. https://www.newplasticseconomy.
org/about/publications/report-2017
References 503

Ellen MacArthur Foundation. (2019). Enabling a circular economy for chemicals with the mass balance approach: A
white paper from the co. Project mass balance. https://www.ellenmacarthurfoundation.org/assets/downloads/Mass-
Balance-White-Paper-2020.pdf
Franklin Associates, A Division of Eastern Research Group (ERG). (2018). Life cycle impacts of plastic packaging
compared to substitutes in the United States and Canada: Theoretical Substitution Analysis. https://plastics.
americanchemistry.com/Reports-and-Publications/LCA-of-Plastic-Packaging-Compared-to-Substitutes.pdf
Fujimaki, T. (1998). Polymer Degradation and Stability, 59, 209.
G€allstedt, M., Mattozzi, A., Johansson, E., & Hedenqvist, M. S. (2004). Biomacromolecules, 5, 2020.
Garcia, J. M., & Robertson, M. L. (2017). Science, 358, 870.
Gedde, U. W. & Hedenqvist, M. S. (2019) Polymer Solutions (Chap. 4). In Fundamental Polymer Science, Cham:
Springer Nature Switzerland AG.
Geyer, B., Lorenz, G., & Kandelbauer, A. (2016). eXPRESS Polymer Letters, 10, 559.
Geyer, R., Jambeck, J. R., & Law, K. L. (2017). Science Advances, 3, 1700782.
Gioia, C., Lo Re, G., Lawoko, M., & Berglund, L. (2018). Journal of the American Chemical Society, 140, 4054.
Grignard, B., Gennen, S., Jérôme, C., Kleij, A. W., & Detrembleur, C. (2019). Chemical Society Reviews, 48, 4466.
Gr€oning, M., & Hakkarainen, M. (2001). Journal of Chromatography, A, 932, 1.
Gr€oning, M., & Hakkarainen, M. (2002). Journal of Applied Polymer Science, 86, 3396.
Haider, T. P., V€olker, C., Kramm, J., Landfester, K., & Wurm, F. R. (2019). Angewandte Chemie International Edition,
58, 50.
Hakkarainen, M., Gr€ oning, M., & Albertsson, A.-C. (2003). Journal of Applied Polymer Science, 89, 867.
Hauenstein, O., Reiter, M., Agarwal, S., Rieger, B., & Greiner, A. (2016). Green Chemistry, 18, 760.
Ignatyev, I. A., Thielemans, W., & Beke, B. V. (2014). ChemSusChem, 7, 1579.
Inkinen, S., Hakkarainen, M., Albertsson, A.-C., & S€odergård, A. (2011). Biomacromolecules, 12, 523.
Isikgor, F. H., & Becer, C. R. (2015). Polymer Chemistry, 6, 4497.
Iwata, T. (2015). Angewandte Chemie International Edition, 54, 3210.
Jambeck, J. R., Geyer, R., Wilcox, C., Siegler, T. R., Perryman, M., Andrary, A., Narayan, R., & Law, K. L. (2015).
Science, 347, 768.
Jawerth, M., Johansson, M., Lundmark, S., Gioia, C., & Lawoko, M. (2017). ACS Sustainable Chemistry & Engineer-
ing, 5, 10918.
Kloxin, C. J., & Bowman, C. N. (2013). Chemical Society Reviews, 42, 7161.
Kreiger, M. A., Mulder, M. L., Glover, A. G., & Pearce, J. M. (2014). Journal of Cleaner Production, 70, 90.
La Mantia, F.-P., & Vinci, M. (1994). Polymer Degradation and Stability, 45, 121.
Langer, E., Waskiewicz, S., Lenartowicz-Klik, M., & Bortel, K. (2015). Polymer Degradation and Stability, 119, 105.
Layman, J. M., Gunnerson, M., Schonemann, H., & Williams, K. (2017a). U.S. Patent US9803035.
Layman, J. M., Gunnerson, M., Schonemann, H., & Williams, K. (2017b). U.S. Patent US9834621.
Lee, S. H., Cyriac, A., Leon, J. Y., & Lee, B. Y. (2012). Polymer Chemistry, 3, 1215.
Lim, L.-T., Auras, R., & Rubino, M. (2008). Progress in Polymer Science, 33, 820.
Maris, J., Bourdon, S., Brossard, J., Cauret, L., Fontaine, L., & Montembault, V. (2018). Polymer Degradation and
Stability, 147, 245.
McConnell, V. P. (2008). Reinforced Plastics, 52, 34.
Michalak, M., Kurcok, P., & Hakkarainen, M. (2017). Polymer International, 66, 617.
Min, K., Cuiffi, J. D., & Mathers, R. T. (2020). Nature Communications, 11, 727.
Ministry of Environment and Food of Denmark, Environmental Protection Agency. (2018). Environmental project no.
1985. https://www2.mst.dk/Udgiv/publications/2018/02/978-87-93614-73-4.pdf
Mohanty, A. K., Misra, M., & Hinrichsen, G. (2000). Macromolecular Materials and Engineering, 276(277), 1.
Montarnal, D., Capelot, M., Tournilhac, F., & Leibler, L. (2011). Science, 334, 965.
Moreno, A., & Sipponen, M. (2020). Materials Horizons, 7, 2237.
Nameer, S., Deltin, T., Sundell, P. E., & Johansson, M. (2019). Progress in organic coatings, 136, 105277.
Napper, I. E., & Thompson, R. C. (2019). Environmental Science & Technology, 53, 4775.
Nikles, D. E., & Farahat, M. S. (2005). Macromolecular Materials and Engineering, 290, 13.
Pang, J., Zheng, M., Sun, R., Wang, A., Wang, X., & Zhang, T. (2016). Green Chemistry, 18, 342.
Panic, V. V., Seslija, S. I., Popovic, I. G., Spasojevic, V. D., Popovic, A. R., Nikolic, V. B., & Spasojevic, P. M. (2017).
Biomacromolecules, 18, 3881.
Patel, M., Angerer, G., Crank, M., Schleich, J., Marscheider-Weidemann, F., Wolf, O. & Hüsing, B. (2005) Technical
Report EUR 22103 EN.
Pilz, H., Schweighofer, J. & Kletzer, E. (2005). The contribution of plastic products to resource efficiency,
PlasticsEurope – Association of Plastic Manufacturers, Brussels, Belgium.
Pilz, H., Brandt, B., & Fehringer, R. (2010). The impact of plastics on life cycle energy consumption and greenhouse gas
emissions in Europe. PlasticsEurope – Association of Plastic Manufacturers, Brussels, Belgium.
504 9 Plastics and Sustainability

Plastic Recyclers Europe. (2018a). https://www.plasticsrecyclers.eu/design-recycling

Plastic Recyclers Europe. (2018b). https://www.plasticsrecyclers.eu/chemical-recycling
PlasticsEurope. (2020). https://www.plasticseurope.org
PureCycle-Technologies. (2019). https://purecycletech.com/2019/09/images-feedstock-evaluation-unit/
Qian, S., Igarashi, T., & Nitta, K.-H. (2011). Polymer Bulletin, 67, 1661.
Ragaert, K., Delva, L., & van Geem, K. (2017). Waste Management, 69, 24.
Rahimi, A., & Garcia, J. M. (2017). Nature Reviews Chemistry, 1, 0046.
Rochman, C. M. (2018). Science, 360, 28.
R€ottger, M., Domenech, T., van der Weegen, R., Breuillac, A., Nicolaÿ, R., & Leibler, L. (2017). Science, 356, 62.
Roy, P. K., Hakkarainen, M., Varma, I. K., & Albertsson, A.-C. (2011). Environmental Science & Technology, 45,
4217.
Schut, J. H. (2001). Plastics Technology, 47, 58.
Shen, L., Worrell, E., & Patel, M. K. (2012). Biofuels, Bioproducts & Biorefining, 6, 625.
Shen, M., Song, B., Zeng, G., Zhang, Y., Huang, W., Wen, X., & Tang, W. (2020). Environmental Pollution, 263,
114469.
Sheppard, D. T., Jin, K., Hamachi, L., Dean, W., Fortman, D. J., Ellison, C. J., & Dichtel, W. R. (2020). ACS Central
Science, 6, 921.
Siracusa, V., & Blanco, I. (2020). Polymers, 12, 1641.
SOU 2018:84. (2018). https://www.regeringen.se/rattsliga-dokument/statens-offentliga-utredningar/2018/12/sou-
201884/
Stoof, D., & Pickering, K. (2018). Composites Part B Engineering, 135, 110.
The Guardian. (2016). Biodegradable plastic: false solution for ocean waste. https://www.theguardian.com/environ-
ment/2016/may/23/biodegradable-plastic-false-solution-for-ocean-waste-problem
Thompson, R. C., Olsen, Y., Mitchell, R. P., Davis, A., Rowland, S. J., John, A. W. G., McGonigle, D., & Russell, A. E.
(2004). Science, 304, 838.
Tian, X., Liu, T., Wang, Q., Dilmurat, A., & Li, D. (2017). Journal of Cleaner Production, 142, 1609.
Trotta, J., Watts, A., Wong, A. R., LaPointe, A. M., Hillmyer, M. A., & Fors, B. P. (2019). ACS Sustainable Chemistry
& Engineering, 7, 2691.
Tzoganakis, C., Vlachopoulos, J., & Hamielec, A. E. (1988). Polymer Engineering & Science, 28, 170.
Ulery, B. D., Nair, L., & Laurencin, C. T. (2011). Journal of Polymer Science, Part B: Polymer Physics, 49, 832.
UNEP. (2015). Biodegradable plastics and marine litter: Misconceptions, concerns and impacts on marine environ-
ment. United Nations Environment Programme (UNEP). https://www.un-ilibrary.org/environment-and-climate-
change/biodegradable-plastics-marine-litter_1dc5344e-en
UNEP. (2017). UN Declares war on ocean plastics. https://www.unenvironment.org/news-and-stories/press-release/
un-declares-war-ocean-plastic-0
Van Zee, N. J., & Nicolaÿ, R. (2020). Progress in Polymer Science, 104, 101233.
Von der Assen, N., & Bardow, A. (2014). Green Chemistry, 16, 3272–3280.
Wang, X., Huang, L. X., Zhang, C. H., Deng, Y. J., Xie, P. J., Liu, L. J., & Cheng, J. (2020). Carbohydrate Polymers,
240, 116292.
Wu, D., & Hakkarainen, M. (2015). European Polymer Journal, 64, 126.
Wu, D., Samanta, A., Srivastava, R. K., & Hakkarainen, M. (2017). Biomacromolecules, 18, 1582.
Wu, B., Sufi, A., Ghosh Biswas, R., Hisatsune, A., Moxley-Paquette, V., Ning, P., Soong, R., Dicks, A. P., & Simpson,
A. J. (2020). ACS Sustainable Chemistry & Engineering, 8, 1171.
Xiang, Q., Xanthos, M., Mitra, S., Patel, S., & Guo, J. (2002). Polymer Degradation and Stability, 77, 93.
Xu, H., Wu, D., Xie, L., & Hakkarainen, M. (2016). ACS Sustainable Chemistry & Engineering, 4, 2211.
Xu, Y., Hua, G., Hakkarainen, M., & Odelius, K. (2018). Biomacromolecules, 19, 3077.
Xu, Y., Odelius, K., & Hakkarainen, M. (2019). ACS Sustainable Chemistry & Engineering, 7, 13456.
Yang, X., Finne-Wistrand, A., & Hakkarainen, M. (2013). Composites Science and Technology, 86, 149.
Yang, X., Clenét, J., Xu, H., Odelius, K., & Hakkarainen, M. (2015). Macromolecules, 48, 2509.
Zhang, M., & Yu, Y. (2013). Industrial & Engineering Chemistry Research, 9505.
Zhang, X., Fevre, M., Jones, G. O., & Waymouth, R. M. (2018). Chemical Reviews, 118, 839.
Zhou, Y., Groote, R., Goossens, J. G. P., Sijbesma, R. P., & Heuts, J. P. A. (2019). Polymer Chemistry, 10, 136.
Zhu, Y., Romain, C., & Williams, C. K. (2016). Nature, 540, 354.
Zumstein, M. T., Narayan, R., Kohler, H.-P. E., McNeill, K., & Sander, M. (2019). Environmental Science &
Technology, 53, 9967.
Chapter 10
Solutions to Problems Given in Exercises

10.1 Chapter 1

1.1. Mass crystallinity (wc) is calculated according to the total enthalpy method (cf. Gedde and
Hedenqvist 2019a):

Δhm
wc ðT 1 Þ ¼ ð10:1Þ
Δh0m ðT 1 Þ

Tð0m


Δh0m ðT 1 Þ ¼ Δhm T m 
0 0
ðCp,a  Cp,c ÞdT ð10:2Þ
T1

The integral in Eq. (10.2) is solved first. The heat capacity data are given according to the
following polynomial expressions:

Cp,c ¼ 8:19697  107 T 3  0:00075737  T 2 þ 0:256479  T  25:827302 ð10:3Þ

Cp,a ¼ 3:80358  10∗ T 3  3:683  106  T 2 þ 0:0066435  T þ 5:380487 ð10:4Þ

Cp,a  Cp,c ¼ 8:235  107 T 3 þ 7:6105  104 T 2  0:25011  T  31:2078 ð10:5Þ

and the integral is given by

ð
418


Cp,a  Cp,c dT  6 J g1 ð10:6Þ
370

The melting enthalpy point at 370 K is according to Eq. (10.2):

# Springer Nature Switzerland AG 2021 505

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3_10
506 10 Solutions to Problems Given in Exercises

ð
418


Δh0m ð370Þ ¼ Δh0m ð418Þ  Cp,a  Cp,c dT ¼ 293  6 ¼ 287 J g1 ð10:7Þ
370

The mass crystallinity is according to Eq. (10.1):

140
wc ð370Þ ¼ ¼ 0:487 ð10:8Þ
287

The last digit is not safe, that is the mass crystallinity is 49  2%.
1.2. The entropy of fusion is given by

Δh0m 293:14
Δ S0m ¼ ¼  9:86 J mol1 K1 ð10:9Þ
T 0m 418

1.3. The enthalpy of fusion per main chain atom is greater for polyamide 6 than for polyethylene. The
average intermolecular energy, expressed, for example, in cohesive energy density (CED;
cf. Gedde and Hedenqvist 2019b) is greater for polyamide 6 than for polyethylene due to the
presence of the strong hydrogen bonds in polyamide 6.
1.4. Medium-density polyethylene in this particular case is a blend of high- and low-density
polyethylenes. The two components crystallize in separate crystal lamellae with different
melting points.
1.5. Polyamide 6 shows a significant uptake of water at room temperature with a pronounced change
in its mechanical properties. Thermogravimetry (TG) is used for assessing the equilibrium water
uptake by heating the considered sample from 25  C to 120  C while recording the mass loss.
Furthermore, please continue to record the mass loss during the extended isothermal treatment
period at 120  C in order to capture all the sorbed water.
1.6. Thermogravimetry (TG) is used. The polymer depolymerizes in nitrogen, and the total mass loss
at 600–700  C is equal to the original polymer content. Further heating in nitrogen releases CO2
from the calcium carbonate, and the latter loses 44% of its mass between 700 and 800  C. By
changing the atmosphere from N2 to 0 2 at 800  C, oxidation of carbon black takes place,
yielding gaseous CO2. The remaining ash consists of pristine calcium oxide.
1.7. The liquid-crystalline component (Vectra) is oriented in the injection-moulded blends and
possesses an approximately invariant thermal expansion coefficient anisotropy in the different
blends.
1.8. The relationship between the linear expansion coefficients (i.e. longitudinal (αL) and transverse
(αT)) and the volume expansion coefficient (αV) can be derived as follows:


V ¼L30  ð1 þ αL Þ ð1 þ αT Þ2 ¼ L30  1 þ 2αT þ α2T þ αL þ 2αT αL þ α2T αL
ffiL30  ð1 þ 2αT þ αL Þ
V 0 ¼L30 ð10:10Þ
V
¼1 þ αV ¼ 1 þ 2αT þ αL ) αV ¼ 2αT þ αL
V0

The following thermal expansivity values are presented in Table 10.1:

10.1 Chapter 1 507

Table 10.1 Thermal Vectra content (%) αL αT αV

expansivity
0 6.5 105 6.5 105 19.5 105
40 2.0105 6.6 105 15.2 105
80 0.2 105 7.5 105 14.8 105
100 1.0 105 8.0 105 15.0 105

Fig. 10.1 Left: Enthalpy as a function of temperature for a sample cooled at 1 K min1 and then heated at two different
rates (0.1 (blue) and 100 (red) K min1). The right-hand graph shows the corresponding heating thermograms

1.9. The sample cooled at the higher rate (100 K min1) shows a break in the liquid enthalpy line at
a higher temperature than the sample which is cooled at the lower rate (1 K min1). The
temperature difference between the two glass transitions is approximately 6–8 K. Figure 10.1
shows, respectively, the enthalpy vs temperature and the differential heat flow vs temperature
(i.e. a DSC thermogram) for the sample heated at the two different rates. Note that the fictive
temperature (Tf) is the same independent on the heating rate (cf. Fig. 10.1 left).
1.10. Cool the sample at a high rate (i.e. by a down-jump) to the temperature of choice, and keep the
different sample for a range of different times at this particular temperature. The change in
fictive temperature is recorded after each experiment according to the method illustrated in
Fig. 10.2. The area marked A should be equal to area B.
The relationship between actual volume (V ) obtained by dilatometry and fictive temperature is
as follows:


V  V 1 ¼ ðT f  T Þ  αl  αg ð10:11Þ

where V1 is the equilibrium volume, T is the actual temperature, αl is the thermal expansivity of
the liquid and αg the thermal expansivity of the glass. A similar equation is valid for DSC
enthalpy (H ) data:


H  H 1 ¼ ðT f  T Þ  Cp,l  Cp,g ð10:12Þ

where H1 is the equilibrium enthalpy at temperature T, and Cp,l is the heat capacity of the liquid
phase and Cp,g is the heat capacity of the glass.

1.11. The sample is probably a liquid crystalline polymer. The high temperature transition is a first-
order transition involving a nematic phase converting into an isotropic phase, which is visible in
an optical microscope equipped with crossed polarizers (a marked decrease in the total light
508 10 Solutions to Problems Given in Exercises

Fig. 10.2 Illustration of a method for the assessment of the fictive temperature (Tf) from DSC thermograms. From
Gedde and Hedenqvist (2019c), with permission from Springer Nature

intensity). The low temperature process is due to a phase transition involving a smectic phase
and a nematic phase. The latter renders just a minor change in light intensity but a more
pronounced change in enthalpy.
1.12. Yes. OIT is much more sensitive to changes in antioxidant concentration at high stabilization
levels.
1.13. OIT varies between 50 and 100 min at 190  C (Fig. 1.65). The change in OIT with temperature
can be calculated from the Arrhenius equation using the activation energy 176 kJ mol1:

 
 
OIT 190 C =OIT 220 C ¼ exp ð176000=ð8:31  463ÞÞ= exp ð176000=ð8:31  493ÞÞ
¼ 16:177 ð10:13Þ

OIT varies between 3 and 6 min at 220  C.

 
 
OIT 190 C =OIT 230 C ¼ exp ð176000=ð8:31  463ÞÞ= exp ð176000=ð8:31  503ÞÞ
¼ 38:003 ð10:14Þ

OIT varies between 1.3 and 2.6 min at 230  C.

1.14. The Arrhenius diagram of the two relaxation processes is shown in Fig. 1.15. The activation
energy (ΔEa) can be calculated from the slope according to:


ΔEa ¼ slope  ln 10  R  slope  2:303  8:31 in J mol1 ð10:15Þ

The low temperature process (β) exhibits a very high activation energy, 202 kJ mol1,
whereas the high temperature process (α) obeys a WLF-temperature dependence, not consistent
with a particular value for ΔEa.
1.15. The derivation of the Kirchhoff law begins from the fact that H is a state function and that two
different paths should lead to the same difference in H. Path A involves heating the reactants to
temperature T2, followed by a transition at T2, whereas path B starts with a transition at T1, after
which the products are heated to T2 (Fig. 10.3). The following expressions are obtained:

Tð2 Tð2

ΔH ðT 1 Þ þ Cp,p dT ¼ΔH ðT 2 Þ þ Cp,r dT

T1 T1
ð10:16Þ
Tð2


ΔH ðT 2 Þ ¼ΔH ðT 1 Þ þ Cp,p  Cp,r dT
T1
10.2 Chapter 2 509

Fig. 10.3 Schematic representation of the enthalpy of reactants and products. Note the two different paths (denoted A
and B) to reach point “stop” from point “start”. From Gedde (2020), with permission from Springer Nature

where the indices p and r represent products and reactants, respectively. Eq. (10.16) is known as
the Kirchhoff law and it is a very useful expression. The enthalpy change associated with a
given process is normally reported for only one (standard) temperature. By the application of
Kirchhoff law, it can be calculated for any other temperature, provided the heat capacity data of
products and reactants are known.
1.16. This is because at constant pressure conditions, the change in enthalpy (ΔH ) is equal to the heat
flow (q) or in differential quantities: dH ¼ dq. This is proven according to:

dH ¼dU þ pdV þ Vdp∧dU ¼ dq  pdV

dH ¼dq  pdV þ pdV þ Vdp ð10:17Þ
dH ¼dq ðp ¼ constantÞ

1.17. The first step is to obtain a completely pristine polymer. This may include removing all the
volatile species in vacuum. The equilibrium concentration of the volatile compound at different
temperatures is obtained by isothermal sorption until constant mass is obtained. This may
require a precision balance depending on the size of the percentages of uptake. After obtaining a
constant sample mass, the sample is transferred to the thermogravimetric balance and the
gradual change in mass recorded at the same temperature is used during the sorption stage.
This will be an input in a numerical solution of the Fickian equations with specified boundary
and starting conditions (cf. Chap. 7).

10.2 Chapter 2

2.1. Prepare thin films (< 5 μm) by solvent casting. Try to make wide samples, preferably 5–10 mm in
diameter. The sample with large spherulites can be studied by polarized microscopy, and with the
λ-plate inserted in the beam, the optical sign of the spherulites can be assessed. The small size of
the spherulites of the other sample makes direct assessment using polarized microscopy
510 10 Solutions to Problems Given in Exercises

Fig. 10.4 Negative and positive spherulites as they appear in an optical microscope equipped with crossed polarizers
and a λ-plate

impossible. Small-angle light scattering is the preferred method. It can be conducted in the
microscope. Use crossed polarizers; close the field and aperture irises, and focus the back focal
plane of the objective lens by using the Bertrand lens. A clover-leaf pattern will appear from
which the spherulite diameter can be determined. Note that this method requires calibration.
More information about this method, which mimics the small-angle light scattering method, is
given by Gedde and Hedenqvist (2019a).
2.2. Insert the sample in the polarized microscopy (crossed polarizers). Slot in the λ-plate oriented as
shown in Fig. 10.4. Colours appear in the spherulites according to either of the following schemes
depending on the sign: negative spherulite (left); positive spherulite (right). If you feel uncertain
about the orientation of the λ-plate, ‘calibrate’ your system by examining a polyethylene sample,
which always shows negative spherulites.
2.3. PVC/NR: Staining by using osmium tetroxide; it will react with NR and increase the density of
the NR phase. Contrast for transmission electron microscopy is thus achieved.
ABS: The blend sample is to be stained with osmium tetroxide, which reacts only with the
polybutadiene phase and increases its density. Contrast for transmission electron microscopy
is thus achieved.
PE/PS: Solvent etching; use chloroform, which will dissolve PS. The two-phase structure is
revealed by SEM.
PET/LCP: Etching with n-propylamine and PET is removed from the sample. The two-phase
structure is revealed by scanning electron microscopy can be used to reveal the morphology.
PMMA/PVC: PMMA is sensitive to the electron beam and degrades into volatile products, which
provides good contrast (marked difference in density) between the two polymers in the
microscope. Both transmission electron microscopy and scanning electron microscopy can be
used to reveal the morphology.
2.4. The sample should be exposed to a mechanical load to fracture at a sufficiently low temperature
to avoid ductile failure. After standard sample preparation, metal coating (sputtering), the
fracture surface is examined by scanning electron microscopy. The image will reveal the
distribution of the two phases. Note that the image represents a two-dimensional cut through a
three-dimensional structure.
2.5. The left micrograph: the sample had been microtomed to obtain a perfectly flat surface and then
treated with a solvent for PS (solvent extraction). The holes in the surface is thus from the
dissolved PS-phase. This phase appears to consist of discrete particles of diameter of a few μm.
This image enables the assessment of the volume fraction of the PS-phase. The right micrograph
10.2 Chapter 2 511

is from a fracture surface (obtained at room temperature). This image reveals in a nice manner
the spherical character of the PS particles. Their diameter is readily determined. The polyeth-
ylene phase is pulled out to fibres providing a three-dimensional view of the structure. From this
image it is more difficult to assess the volume fraction of the PS phase.
2.6. A phase which appears continuous by SEM or TEM is continuous. The phase which appears
discrete may be discontinuous. However, it is also possible that the apparent separate phases are
connected through a ‘bridge’ along the third (invisible) dimension. An analogue: a cut through
a tree (a three-dimensional continuous structure) reveals a great number of elliptical cuts
(seemingly non-continuous structures).
2.7. The typical view is along the spherulite radius, which is along the b-axis, that is [010].
2.8. Permanganic acid etching of a polyethylene sample provides the means to reveal the three-
dimensional shape of the melt-grown crystals. Some areas on the image reveal the edge-on
crystals (view along the b-axis), whereas other areas show the shape of the crystals along the
c-axis (i.e. the lateral habit).
2.9. The dark field for optical microscopy means that the specimen is illuminated from the side, that
is at a large angle to the optical axis. The bright parts of the image are due to the scattered
(deflected) light. A dark field for transmission electron microscopy is accomplished by
adjusting the objective aperture so that it transmits only diffracted, scattered electrons. The
bright parts of the image in this case are also due to the scattered electrons.
2.10. Scanning electron microscopy due to its large depth of field. It has also the capacity to reveal
foreign particles often present in the initiation region of fracture.
2.11. The linear growth rate is determined by hot-stage polarized microscopy taking a number of
photographs over time. Temperature should be kept strictly constant during the experiment. It is
common that several spherulites are growing in the field of view, and thus an average value of
the growth rate can be obtained. Linear growth rate of growing crystal lamellae can also be
assessed by hot-stage AFM of thin-film samples.
2.12. The intercept method can be used. Draw a number of arbitrary lines and record the fraction of
each line, which is “occupied” by the glass fibres. An average value of the volume fraction of
fibres can be obtained by including data from the different lines.
2.13. The following disclination of strength  ½ are found; note that these are characterized by two
meeting black lines. This disclination is indicated with a red-bordered square (cf. Gedde and
Hedenqvist 2019d) (Fig. 10.5).

Fig. 10.5 Polarized micrograph of nematic sample with indicated disclinations of strength  ½
512 10 Solutions to Problems Given in Exercises

10.3 Chapter 3

3.1. The first method is infrared spectroscopy: the 1379 cm1 absorption band is assigned to methyl
groups, whereas the 1465 cm1 band is due to methylene groups. The methyl group concentra-
tion (i.e. branch concentration) is determined by comparison of the two absorption peaks
(internal standard) after suitable calibration with polymers of known chain branching. There is
some interference between the 1379 cm1 band and the adjacent methylene bands. This ‘back-
ground’ can be removed by taking the difference spectrum between the sample spectrum and the
spectrum of a branch-free polymer. The second method is 13C NMR spectroscopy, by which both
the type (branches up to six carbon atoms can be distinguished) and their concentrations can be
determined.
3.2. (a) Small-angle X-ray scattering (SAXS) assessing the long period (L ) combined with wide-angle
X-ray scattering determining the volume crystallinity (ϕc). The average crystal thickness (Lc)
is calculated according to: Lc ¼ Lϕc.
(b) Raman spectroscopy by measuring the frequency (ν) of the longitudinal acoustic mode
(LAM):
rffiffiffi
n E
Lc ¼  ð10:18Þ
2ν ρ

where n is the order of the vibration, E is the Young’s modulus in the chain-axis direction and
ρ is the density of the crystalline component.

(c) A third method, possibly less attractive than the former two, is to measure the width of the
crystalline wide-angle reflections (WAXS) and to calculate the crystal thickness perpendicu-
lar to the diffracting planes by using the Scherrer equation.
3.3. It may be possible when strong specific interaction between the two polymers occurs. Hydrogen
bonding may cause a frequency shift of vibrations of the participating groups, for example the
carbonyl stretch band in polyesters. Both IR and NMR spectroscopy are short-sighted, and the
interaction between nearby groups have an effect on the chemical shift, which may provide some
information about interaction between two different polymers. However, this doesn’t provide
sufficient information to construct a phase diagram.
3.4. X-ray scattering measures the crystallinity considering only crystals thicker than 2–3 nm. IR
measures the total content of helix structures. The very thin crystals (< 2–3 nm) are included in
the IR crystallinity assessment. It is also possible that the helices are present in the amorphous
phase.
3.5. X-ray diffraction records changes in the unit cell parameter. IR spectroscopy records frequency
shift of certain absorption bands which, in addition to a shift in the peak frequency, will also turn
asymmetric due to the variation in stress among the different segments. Small-angle neutron
scattering of labelled (deuterated) guest chains provide information about deformation of whole
molecules.
3.6. Oxidation of polyolefins yields a number of different carbonyl-containing compounds, such as
ketones, aldehydes and esters, all with carbonyl-stretching vibrations at slightly different
frequencies. The assessment of their concentrations can only be made by a resolution of the
broad carbonyl band into its basic components (ketone-, aldehyde- and ester-carbonyl bands).
3.7. Infrared dichroism measurement provides information about the orientation of groups. If the
angle between the transition moment vector and the chain axis is known, it is possible to
determine a Hermans orientation parameter for the chain axis. The c-axis orientation (crystal-
10.4 Chapter 4 513

phase orientation) can be obtained from wide-angle X-ray scattering. The orientation of an
amorphous polymer can be assessed by IR spectroscopy.
3.8 Because the concentration of 13C is very low and the average spin surrounding such atom is
essentially 100% protons.
3.9. The triplet due to the CH2 group is downfield (less shielded carbon) compared to the methyl
groups (more shielded carbon) occurring as a quartet. The quaternary carbon occurs as a singlet
between the CH2 and CH3 groups on the chemical shift axis. As determined from the Pascal
triangle, the middle peak in the triplet is twice the size of the outer two peaks, and the middle two
peaks in the quartet are three times the size of the outer two peaks. The peaks are broadened
because of unresolved multiple two and three bond proton couplings to the carbon (Ibbett 1993).
3.10. Use Eq. (3.21): δ ¼ 106(500/(50106) ¼ 10 ppm. At 10 ppm, it may be a proton in, for example,
a carboxylic group.
3.11. Apply the Larmor equation: B0 ¼ υr/γ. The magnetogyric ratio for the proton is 42.6 MHz/T
(267.5 (rad/s)/T), yielding a magnetic field of B0 ¼ 500/42.6 ¼ 11.7 T. The magnetogyric ratio
for 13C is 10.7 MHz/T, and the resonance frequency becomes: υr ¼ 10.7 11.7 ¼ 125.2 MHz,
which is four times lower than that of a proton.

10.4 Chapter 4

4.1. The following input data are used:

m1 ¼ 3 g, m2 ¼ 5 g and m3 ¼ 2 g,
M1 ¼ 20,000 g mol1, M2 ¼ 50,000 g mol1 and M3 ¼ 90,000 g mol1.

which makes possible the calculation of the number of moles of the three species: n1 ¼ m1/
M1 ¼ 3 g/20000 g mol1 ¼ 0.00015 mol, n2 ¼ m2/M2 ¼ 5 g/50000 g mol1 ¼ 0.0001 mol and
n3 ¼ m3/M3 ¼ 2 g/90000 g mol1 ¼ 0.000022 mol
P
mi
3þ5þ2
¼ 36 765 g mol1
i
Mn ¼ P ¼ ð10:19Þ
ni 0:00015 þ 0:0001 þ 0:000022
i

P
mi Mi
3  20000 þ 5  50000 þ 2  90000
¼ 49 000 g mol1
i
Mw ¼ P ¼ ð10:20Þ
mi 3þ5þ2
i

Mw 49000
D¼ ¼ ¼ 1:33 ð10:21Þ
Mn 36765

4.2. (a) Assume 1 g of sample ¼> m1 ¼ 0.99 g and m2 ¼ 0.01 g

P
mi
1
¼ 40 080 g mol1
i
Mn ¼ P ¼ ð10:22Þ
ni 0:99=200000 þ 0:01=500
i
514 10 Solutions to Problems Given in Exercises

P
mi Mi
0:99  200000 þ 0:01  500
¼ 198 005 g mol1
i
Mw ¼ P ¼ ð10:23Þ
mi 1
i

(b) For Mn, the impact of a small mass fraction of oligomer is significant, because this tiny mass
fraction contains great many molecules, whereas Mw shows only a minor change. The
numbers of moles of the two species are: n1 ¼ 0.99 g/ 200,000 g mol1 ¼ 0.00000495 mol
and n2 ¼ 0.01 g/ 500 g mol1 ¼ 0.00002 mol. Note that n2/n1  4, whereas m2/m1  0.01.
4.3. Sample A has the highest molar mass, and it also shows a variation in molar mass between
different molecules, which imply that the retention time will be shorter than for the other samples
and that Mw/Mn > 1. Samples B and C have approximately the same molar mass, but sample C is
a branched star polymer. A branched polymer molecule has a smaller hydrodynamic volume
than a linear polymer molecule with same molar mass. This means sample C is retained in the
column for a longer period of time than sample B. Both samples B and C show monodispersity
(Mw/Mn  1). Sample A ¼ SEC 3, sample B ¼ SEC 2, sample C ¼ SEC 1.
4.4. Polymer A has a higher solvent uptake, and its molecules expand more, that is the molecules
possess a higher hydrodynamic volume than those of polymer B. A molecule with larger
hydrodynamic volume travels through the column faster and thus elute earlier.
4.5. SEC chromatogram 1 shows sample B. The width of the peak indicates a molar-mass polydis-
persity. The retention time of the longer chains in sample B should be shorter than those of the
chains in Sample A. SEC chromatogram 2 thus shows sample A, a monodisperse polymer with a
narrow SEC peak. SEC chromatogram 3 shows sample C. The cross-linked sample is not soluble
and no SEC peak is detected.
4.6. (a) A chromatogram with a single peak. (b) Since the molar mass differences are clear and the
molar-mass dispersity is expected to be low, a chromatogram with three peaks is expected. If
the resolution is less good, a single peak with two shoulders is detected, suggesting the
presence of three different molecular species. (c) A chromatogram with two peaks, one
associated with homopolymer A and the other with homopolymer B. In the case of a poor
resolution, a single peak with shoulder would be revealed.
4.7. The calibration gave the relationship between molar mass and retention time shown in Fig. 10.6.

Fig. 10.6 Calibration with monodisperse PS standards providing a correlation between molar mass and retention time
10.4 Chapter 4 515

Table 10.2 SEC data for PMMA in THF at 25  C

Retention time (min) Log Mi Mi (g/mol1) Height (Hi) (mm) Hi/Mi (mm mol g1) Hi Mi (mm g mol1)
30 4.938 86,696 1.0 1.15  105 86,696
31 4.746 55,719 6.2 1.11  104 345,458
32 4.554 35,810 30 8.38  104 1,074,300
33 4.362 23,014 40 0.00174 920,560
34 4.170 14,791 27 0.00183 399,357
35 3.978 9506 11 0.00116 104,566
36 3.786 6109 2 3.27  104 12,218
Note that Hi is proportional to (δwi/δ(log M)); point sampling needs to be uniformly distributed in log(M)-space

The calculated molar mass averages and the molar mass dispersity are based on the following
results obtained from the data of Table 10.2:
X X X
H i ¼ 117:2 ðmmÞ, ð Hi =Mi Þ ¼ 0:00602 ðmm mol g1 Þ and ð H i Mi Þ
¼ 2 943 155 ðmm g mol1 Þ

P
Hi
117:2
¼ 19 468 g mol1
i
Mn ¼ P Hi ¼ ð10:24Þ
0:00602
Mi
i

P
H i Mi
2943155
¼ 25 112 g mol1
i
Mw ¼ P ¼ ð10:25Þ
Hi 117:2
i

Mw 25112
D¼ ¼ ¼ 1:29 ð10:26Þ
Mn 19468

4.8. The molar mass of PMMA is obtained by applying the universal calibration method:
     
1 K PS 1 þ aPS
log MPMMA ¼  log þ  log MPS ð10:27Þ
1 þ aPMMA K PMMA 1 þ aPMMA

and by inserting the Mark-Houwink-Sakurada constants, the following result is obtained:

     
1 1:03  102 1 þ 0:74
log MPMMA ¼  log þ  log 97000 ¼ 5:076 ð10:28Þ
1 þ 0:76 5:7  10 3 1 þ 0:76

which results in MPMMA ¼ 119,124 g mol1.

4.9. With the help of the calibration curve, Mi for each fraction was calculated; Hi is related to amount
of sample with molar mass Mi (Table 10.3).
516 10 Solutions to Problems Given in Exercises

Table 10.3 SEC data for polystyrene in THF at 25  C

TR (min) Log Mi Mi (g mol1) Hi (mm) Hi/Mi (mm mol g1) Hi Mi (mm g mol1)
35 4.8 63,096 2 3.17  105 126,192
36 4.6 39,811 19 4.47  104 756,409
37 4.4 25,119 136 5.41103 3,416,184
38 4.2 15,849 184 0.0116 2,916,216
39 4.0 10,000 59 0.0059 590,000
40 3.8 6310 24 0.0038 151,440
41 3.6 3981 8 0.0020 31,848

Fig. 10.7 Calibration curve for PS

Summation yields the following values: ∑Hi ¼ 432 (mm), ∑ (Hi/Mi) ¼ 0.02919 mm mol g1
and ∑ ( HiMi) ¼ 7 988 289 (mm g mol1), which are used to obtained the following average molar
mass values:
P
Hi
432
¼ 14 800 g mol1
i
Mn ¼ P Hi ¼ ð10:29Þ
0:02919
Mi
i

P
H i Mi
7988289
¼ 18 491 g mol1
i
Mw ¼ P ¼ ð10:30Þ
Hi 432
i

Mw 18491
D¼ ¼ ¼ 1:25 ð10:31Þ
Mn 14800

Hence, the requirement for monodispersity is not fulfilled, because D > 1.1.
4.10. (a) Prepare a calibration curve and use it to calculate the relative M (in PS equivalents) for
monodisperse polymer X with 38 min retention time (cf. Fig. 10.7):

The molar mass of polymer X sample is thus:

log M ¼ 0:1912  38 þ 11:743 ¼ 4:4774 ) M ¼ 30 019 g mol1 ð10:32Þ

10.5 Chapter 5 517

(b) The universal calibration equation to calculate the molar mass of polymer X is
     
1 K PS 1 þ aPS
log MX ¼  log þ  log MPS
1 þ aX KX 1 þ aX
      ð10:33Þ
1 0:0049 1 þ 0:79
¼  log þ  log 30000
1 þ 0:77 0:0054 1 þ 0:77
¼4:473 ) MX ¼ 29 717 g mol1

The K and a values for polymer X and PS are relatively similar. The deviation in molar mass
due to difference in hydrodynamic volumes is, thus, small: 29,700 g mol1 (after universal
calibration) and 30,019 g mol1 (in PS equivalents).
4.11 (a) SEC is the method of choice for the analysis of high molar mass compounds. Polymers
cannot be vaporized, which is required for GC, and high molar mass compounds would also
block a normal HPLC column. (b) HPLC is the most suitable method. The expected products
are both polar, and they have molar mass as high as 1200 g mol1, which is too high for
GC. The low molar mass species are outside the convenience range for SEC. (c) The molar
mass species can be analyzed by using GC. In addition, the cyclic monomers are less polar,
which decreases their boiling point and facilitates the analysis with GC.
4.12 (a) The boiling point is the main factor influencing the retention time, in combination with
polarity, which also has an impact on the boiling point. The compound with lowest boiling
point will elute first. The expected order with increasing retention is hexane, hexanol and
hexanoic acid. (b) In normal mode, the mobile phase is non-polar and stationary phase is
polar. The polar compounds are expected to interact more strongly with the stationary phase
and analyzed compound to be retained longer. The order with increasing retention time is
thus same as in case (a), that is hexane, hexanol and hexanoic acid. (c) In reverse mode, the
mobile phase is polar and stationary phase is non-polar. The non-polar compounds are
expected to interact more strongly with the stationary phase and hence to be retained longer.
The order with increasing retention time is thus hexanoic acid, hexanol and hexane.

10.5 Chapter 5

5.1. Note that the solutions are only examples of correct answers. Many other correct solutions are
also possible and special techniques can sometimes be used to overcome the stated weaknesses.
QC. Strength: QC can be used to compute complex (sub-) atomistic phenomena involving electron
transport, like chemical reactions and electron distributions. Weakness: Slow computations,
which limits the number of atoms and the length of the simulated time-period.
MD. Strength: MD is a relatively fast atomistic method where time-dependent properties of
polymeric systems with thousand or millions of atoms can be simulated. Weakness: Electron
transport is not included, and only “physical chain movements” are possible, making it difficult
to completely equilibrate polymer systems with long chains.
MC. Strength: MC methods are time-independent, stochastic atomistic methods which enable
“unphysical chain movements” and can be optimized for equilibrating polymer systems or for
constructing polymeric starting structures close to equilibrium. Weakness: Electron transport is
not included and time-dependent properties cannot be computed.
Meso. Strength: Faster than atomistic models (QC, MD and molecular MC) and more detailed
than continuum methods like FEM. Weakness: Sometimes they are limited in application, and
518 10 Solutions to Problems Given in Exercises

in multi-scale modelling, they are hard to construct such that they efficiently bridge and couple
the atomistic and continuum ranges.
FEM: Strength: Can be used to model a very wide range of macroscopic phenomena, ranging
from fluid dynamics and diffusion simulations to mechanical and electrical simulations, in
complex geometries. Weakness: The atomistic detail is completely lost.
Statistical methods: Strength: These are very fast methods which can efficiently utilize informa-
tion from large databases to (for instance) predict properties of new materials or to suggest how
to construct new materials with desired properties. Weakness: They are limited by the quality of
the database and by the analytical expressions suggested as the basis for the calculations.
5.2. QC: Modelling chemical reactions between polymers and reactive molecules. MD: Modelling
diffusion of small penetrant molecules through a polymer matrix. MC: Efficient construction of
atomistic polymer structures close to equilibrium. Meso: Modelling viscous flow of molten
polymer blends. FEM: Computing the effective diffusion coefficient of a composite with known
polymer diffusivity. Statistical methods: Predicting physical properties of a new polymer
molecule with known chemical structure.
5.3. Unless the simulation box is twice as large as the radius of gyration of the polymer chain, the
polymer chain might interact with itself in an appropriate manner. Additionally, if the polymer
chains are long, it will take very long time to equilibrate the molecular system with MD until it
reaches complete equilibrium according the ergodicity criteria.
5.4. Coarse-grained force-fields, as well as united-atom force-fields, increase the computational speed
of MD simulations considerably as compared to all-atom force-fields. One drawback is that they
accelerate certain processes, like diffusion and mechanical phenomena.
5.5. One repeat unit of polystyrene consist of one aromatic C6H5 group, one CH2-group and one
CH ¼ -group. According to Park’s and Paul’s table (Table 5.1), the molar masses (Mi) of these
groups adds to (77.10 + 14.03 + 13.02) ¼ 104.15 g/mol. The weighted sum of the molar volumes
(Vw) with weight factors (bi) is (45.85·1.51 + 10.25·1.33+ 6.8*1.30) ¼ 91.7 cm3/mol, giving the
density 104.15/91.7 ¼ 1.14 g/cm3. This is a slight overestimate compared to the experimental
1.05 g/cm3.
5.6. The Hückel-determinant of one ethane molecule, which has 2 sp2-hybridized carbons, is:

αE β
¼0 ð10:34Þ
β αE

which gives:

E1 ¼ α þ β
ðα  E Þ2  β 2 ¼ 0 ) ð10:35Þ
E2 ¼ α  β

Three ethane molecules have a total energy of E ¼ 3(E1 + E2) ¼ 6α. The Hückel determinant of
benzene (C6H6) is:
αE β 0 0 0 β
β αE β 0 0 0
0 β αE β 0 0
¼0 ð10:36Þ
0 0 β αE β 0
0 0 0 β αE β
β 0 0 0 β αE
10.6 Chapter 6 519

The Hückel energy for Benzene with 6 Sp2 electrons is E ¼ 2E1 + 2E2 + 2E3 ¼ 2(α + 2β) +
2(α + β) + 2(α + β) ¼ 6α + 8β. So the energy difference compared to three ethane molecules is 8β,
where β < 0. The difference is called the delocalization energy and is the gain in energy when the
normal double bounds are delocalized in an aromatic group.

10.6 Chapter 6

6.1. By the use of Eq. (6.23), we have:

pffiffiffiffiffi pffiffiffiffiffi
σ nx ¼ð101 þ 53 þ 55Þ= 35 ¼ 50= 35
pffiffiffiffiffi pffiffiffiffiffi
σ ny ¼ð51 þ 103 þ 155Þ= 35 ¼ 40= 35 ð10:37Þ
pffiffiffiffiffi pffiffiffiffiffi
σ nx ¼ð51  153 þ 205Þ= 35 ¼ 60= 35

Hence: σn ¼ (50, 40, 60)/√35. The absolute value of σn is:

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi
j σn j ¼ 502 þ ð40Þ2 þ 602 =35 ¼ 14:8 MPa ð10:38Þ

The size of the normal stress is obtained with Eq. (6.27):

0 1
50
1 B C
σ ¼ ð1, 3, 5Þ @ 40 A ¼ 6:6 MPa ð10:39Þ
35
60

The shear stress is obtained with Eq. (6.28):

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
τ ¼ jσn j2  σ 2 ¼ 14:82  6:62 ¼ 13:2 MPa ð10:40Þ

6.2. Take the moment around B clockwise:

Fg cos ð45Þ  1  FA  1:5 ¼ 0 ð10:41Þ

where Fg is the force due to gravity (9.81 m/s times 30 kg) ¼ 294.3 N. FA ¼ 138.7 N. Force balance in
the vertical direction gives:
FB ¼ Fg  FA  cos ð45Þ ¼ 196:2 N ð10:42Þ

6.3. The two principal stresses are: σ 1 ¼ 15.1 MPa and σ 2 ¼ 3.9 MPa, obtained by Eqs. (6.35) and
(6.36). The angle of reorientation in the anti-clockwise direction is obtained by Eq. 6.37:
φ ¼ 63.4 .
6.4. The permissible force is P A τcr, where A ¼ πDL is the area over where the force is acting.
A ¼ 1257 cm2, yielding P 12.7 MN.
520 10 Solutions to Problems Given in Exercises

6.5.

ΔV ¼ xyz  x0 y0 z0 ¼ 0 ð10:43Þ

ΔV ¼ λ1 x0 λ2 y0 λ3 z0  x0 y0 z0 ¼ 0 ð10:44Þ

λ1 λ2 λ3 ¼ 1 ð10:45Þ

6.6.
 
L
ε ¼ ln ; L ¼ L0 exp ðεÞ ð10:46Þ
L0


ΔV ¼ x0 exp ðεx Þ y0 exp εy z0 exp ðεz Þ  x0 y0 z0 ð10:47Þ


ΔV x0 exp ðεx Þ y0 exp εy z0 exp ðεz Þ  x0 y0 z0

¼ ¼ exp ðεx Þ exp εy exp ðεz Þ  1 ð10:48Þ
V0 x 0 y 0 z0

At constant volume:


exp ðεx Þ exp εy exp ðεz Þ ¼ 1 ð10:49Þ

6.7.

ΔV=V 0 ¼ εx ð1  2υÞ ¼ 0:04 ð1  2  0:45Þ ¼ 0:004 ð10:50Þ

New density:

ρ ¼ ρ0 =ð1 þ ΔV=V 0 Þ ¼ 1096 kg m3 ð10:51Þ

6.8 (Fig. 10.7).

6.9. Upon unloading, the stress is zero, hence:

Eε þ η_ε ¼ 0 ð10:52Þ

that is integrated:
ðε ðt2
∂ε
¼ ∂t ð10:53Þ
ε
ε1 t1

which yields the final expression:

10.6 Chapter 6 521

E ðt  t 1 Þ
ε ¼ ε1 exp  ðt  t1 Þ ¼ ε1 exp  ð10:54Þ
η τ

6.10. The shear strain after 2 years (730 days) is 0.6%. The strain is higher than the critical value:
0.5%, and hence it is not recommended for the application.
6.11. The modulus can be estimated as the secant modulus by reading off the stress at a strain of 1.3%.
The modulus is 8 MPa / 0.013 ¼ 615 MPa. The volume increase can be estimated from
Eq. (6.42). The resulting volume increase is 0.4%.
6.12. a2, b5, c1, d3, e4.
6.13. The strain after n tensile stress steps is:

X
n
εð t Þ ¼ Δσ i Dðt  ti Þ, t > tn ð10:55Þ
i¼1

where for t ¼ 750 s, n ¼ 2, t1 ¼ 0 s, Δσ 1 ¼ 2 MPa, t2 ¼ 500 s, Δσ 2 ¼ 0.5 MPa, which yields a

strain of 0.63%. For t ¼ 1200 s, n ¼ 3, t1 ¼ 0 s, Δσ 1 ¼ 2 MPa, t2 ¼ 500 s, Δσ 2 ¼ 0.5 MPa and
t3 ¼ 1000 s, Δσ 3 ¼ 2.5 MPa, which yields a strain of 0.47%.
6.14. Choose 30  C as the reference temperature T1 and 40  C as the temperature T2: D40 (a40 t) ¼ D30 (t).
This can be rearranged as: D40 (t) ¼ D30 (t/a40), where the shift factor is calculated as:
   
ΔH 1 1 150 000 1 1
aT ¼ exp  ¼ exp  ¼ 0:149 ð10:56Þ
2 R T2 T1 8:31441 313 303

The compliance behaviour at 40  C is:

   0:5
t t
D40 ðtÞ ¼ D30 ¼ 0:1 ¼ 0:259 t0:5 GPa1 ð10:57Þ
0:149 0:149

6.15. The shifting has to be performed in two steps via the glass transition temperature:
 

 t15 C 17:44 ð15 þ 70Þ
log aT ¼ log ¼ ¼ 10:85 ð10:58Þ
t70 C 51:6 þ 15 þ 70

t15 C 10 h
¼ 1:4  1011 ) t70 C ¼ ¼ 1:1  1012 h ð10:59Þ
t70 C 5:01  1012

and
 
t30 C 17:44 ð30 þ 70Þ
log ¼ ¼ 7:6 ð10:60Þ
t70 C 51:6  30 þ 70

t30 C
¼ 2:4  108 ) t30 C ¼ 2:4  108  1:1  1012 ¼ 25850 h ð10:61Þ
t70 C

Thus, stress relaxation takes much longer time at the lower temperature.
522 10 Solutions to Problems Given in Exercises

6.16. The change in volume due to 1% strain is 0.2% (Eq. (6.42)). The volume change due
to the temperature variation is: 3αL ΔT ¼ 0.0003 ¼ 0.03%. An estimated error is
0.03/0.2 ¼ 0.15 ¼ 15%.
6.17.
   
ΔH 1 1 150 000 1 1
ðaÞ a50 ¼ exp  ¼ exp  ¼ 0:0033 ð10:62Þ
R T2 T1 8:31441 323 293

f50 ¼ 150/0.0033 ¼ 45.5 kHz, (b) rubbery, (c) glassy, (d) 150 Hz, (e) 150 Hz.
6.18. The lifetime (tf) at 5 MPa and 40, 60 and 80  C, are, respectively, 55, 505 and 5573 h. Arrhenius
extrapolation using these times and the corresponding temperatures yields: ln (tf) ¼ 12,764/
T + 32.1, where T is absolute temperature. The activation energy is hence 125.2 kJ. Using this
equation for 20  C yields a lifetime of ca. 92,000 h (ca 10 years).
6.19. Use

ε3 ¼ S13 σ 1 þ S23 σ 2 þ S33 σ 3 ð10:63Þ

Knowing Poisson’s ratio, it is possible to rewrite the equation as:

ε3 ¼ υS33 σ 1  υS33 σ 2 þ S33 σ 3 ð10:64Þ

Then in the first case

ε3 ¼ S33 σ 3 ¼ ð1=10Þ  10 ¼ 1 ð10:65Þ

in the second case we have

ε3 ¼ υS13 σ 3 þ S33 σ 3 ¼ 0:5  ð1=10Þ  10 þ ð1=10Þ  10 ¼ 0:5 ð10:66Þ

In the third case when the same stress is applied on the three orthogonal directions, the net strain is
zero.
6.20. a) E (20, 3000) ¼ 0.80 MPa, E (40, 3000) ¼ 0.85 MPa b) E (20, 7000) ¼ 0.34 MPa,
E (40, 7000) ¼ 0.36 MPa.
6.21. If one uses the specific volumes at 0 and 50 MPa, the calculated bulk modulus is ca. 3 GPa.
6.22. a-I, b-II, c-III. Modulus: 10 MPa, yield stress: 10 MPa and extensibility: almost 10%.
6.23. Fracture stress: 31 MPa, tensile index: 156 kNm/kg.
6.24. Lowest degree of crosslinking/crystallinity (1), highest (3).
6.25. The equation can be formulated:
 
ΔA ε
¼ ε þ εT ¼ ε 1 þ T ¼ εð1  υÞ ð10:67Þ
A ε

where ε and εT are the strains in the main and transverse direction and υ is the two-dimensional
Poisson’s ratio. Then it follows that for the area to remain constant, υ is 1.
6.26. The crystal thickness of 12.5 nm yields a yield stress of ca. 45 MPa and a 20 nm thickness yields
a yield stress of ca. 60 nm.
6.27. The stress intensity factor is calculated from Eqs. (6.195) and (6.192):
10.7 Chapter 7 523

K c ¼ ðE Gc Þ0:5 ¼ 1:41 MPa m0:5 ð10:68Þ

E Gc
am ¼ ¼ 1:8 cm ð10:69Þ
π σ2

Hence, 2am ¼ 3.6 cm, which is larger than the actual crack, and the crack will not propagate.
6.28. The stress intensity factor is calculated from Eqs. (6.206) and (6.205):

 0:5
E GIC
K IC ¼ ¼ 0:95 MPa m0:5 ð10:70Þ
ð1  v2 Þ

K IC 2
am ¼ ¼ 0:8 cm ð10:71Þ
π σ2

2am ¼ 1.6 cm, which is less than the actual crack, and the crack will propagate.
6.29. With the stress varying from 1 to 3 MPa and a crack length of 2 cm, the stress intensity varies as:

ΔK ¼ Δσ ðπaÞ0:5 ¼ 2 ð0:01 π Þ0:5 ¼ 3:54 MPa m0:5 ð10:72Þ

With Paris law and the values for Af and m, the crack growth rate is 20 μm/cycle.

10.7 Chapter 7

7.1. Consider only the first two terms in the series of Eq. 7.87:
   
Mt 8 π2 1 9π 2 1
 1  2 exp  2 D t0:5  exp  2 D t0:5 ¼ ð10:73Þ
M1 π 4l 9 4l 2

By substituting 4lπ 2 D t0:5 with x and rearranging the above equation, the following is obtained:
2

π2 1 1
¼ exp ðxÞ  exp ð9xÞ ¼ exp ðxÞ  ð exp ðxÞÞ9 ð10:74Þ
16 9 9

If only one term is considered in Eq. (10.74), one obtains

π2
exp ðxÞ  ð10:75Þ
16

which allows the second term in Eq. (10.73) to be replaced by π2/16:

 9
π2 1 π2
¼ exp ðxÞ  ð10:76Þ
16 9 16

Rearrangement and removing the substitution (x) yields:

524 10 Solutions to Problems Given in Exercises

"  9 #
1 π2 1 π2
D ¼  π2 t ln  ð10:77Þ
0:5
2
16 9 16
4l

which if the plate thickness is l instead of 2 l becomes the final expression:

0:049
D¼ t0:5 ð10:78Þ
l2

7.2. Using the initial slope between origo (0,0) and half the uptake yields a D ¼ 1.227 106 cm2/s, and
the half-time method yields D ¼ 1.225 106 cm2/s. It should be noted that the curve in the figure
was generated with Eq. (7.87) using a D ¼ 1.250 106 cm2/s and 15 terms in the series.
7.3. The time (θ) where the straight slope crosses the time-axis is 20 s, which gives a diffusivity of
8.3·107 cm2/s. The equation to start with in the derivation is:

     2 2 
2C X 1
1
x nπ nπ
Cðx, tÞ ¼ C1 1   1 sin x exp  2 Dt ð10:79Þ
l π n¼1 n2 l l

The loss rate on the right-side of the plate is:

   2 2 
DC1 2DC1 X
1
∂C nπ
Q ¼ D ¼ þ ð1Þn exp  2 Dt ð10:80Þ
∂x x¼l l l n¼1 l

The accumulated loss is:

ðt ðt X
1  2 2 
DC1 2DC1 nπ
Qt ¼ dt þ ð1Þn exp  2 Dt dt ð10:81Þ
l 0 l 0 n¼1 l

"  2 2 #t
DC1 t 2C1 l X ð1Þ
1 n
nπ
Qt ¼  2 exp  2 Dt ð10:82Þ
l π n¼1
n 2
l
0

which with the insertion of integration limits is:

 2 2 
DC1 t 2C1 l X ð1Þ 2C1 l X ð1Þ
1 n 1 n
nπ
Qt ¼  2 exp  D t  ð10:83Þ
l π n¼1 n2 l2 π 2 n¼1 n2

Since
X1
ð1Þn π2
2
¼ ð10:84Þ
n¼1
n 12

the accumulated loss is at long times then:

10.7 Chapter 7 525

 
DC1 t C1 l DC1 l2
Qt ¼  ¼ t ð10:85Þ
l 6 l 6D

and, when Qt is zero

l2
D¼ ð10:86Þ
6θ

7.4. (1) n-hexane, (2) cyclohexane, (3) 2,2-dimethylbutane, (4) n-decane and (5) tetradecane. The two
non-linear molecules behave more bulky in the diffusion process.
7.5 (Figs. 10.8 and 10.9).

Fig. 10.8 Modulus as a function of time

Fig. 10.9 The benzene-concentration profiles within a sheet of poly(ethyl acrylate)

526 10 Solutions to Problems Given in Exercises

7.6. Using Eq. (7.134) with a w/l ratio of 50 yields a tortuosity factor of 5.2. With Eq. (7.132), the
following diffusion coefficients are obtained:
O2: D ¼ 2.6·107 cm2/s and Ar: D ¼ 1.7·107 cm2/s.
7.7 Rewrite

2
∂C ∂ CD
ðCD þ CH Þ ¼ DD ð10:87Þ
∂t ∂x2

as
 
∂ C0H bp 2
∂ CD
CD þ ¼ DD ð10:88Þ
∂t 1 þ bp ∂x2

and replace p by considering that the dissolved and adsorbed species are in equilibrium with the same
vapour pressure: p ¼ pD ¼ pH. Thus, p ¼ cD/kD and
0 1
C0H b 2
∂@ C ∂ CD
C D þ k D C0 b A ¼ D D
D
ð10:89Þ
∂t 1þ H ∂x2
kD

Differentiation with respect to time of the left side leads to

2 3
C0H b 2
∂CD 6 kD 7 ∂ CD
41 þ  2 5 ¼ DD ð10:90Þ
∂t ∂x2
1 þ kD CD
b

Thus, Fick’s second law becomes

2  2 3
∂CD 1 þ b
C 2
6 kD D 7 ∂ CD
¼ DD 4 0  2 5 ð10:91Þ
∂t CH b ∂x2
k D þ 1 þ k D CD
b

7.8 (Fig. 10.10).

Fig. 10.10 The diffusion

coefficient as a function of
the volume fraction of
benzene
10.7 Chapter 7 527

The faster the mass uptake, the more mobile the adsorbed species are. Consequently, the solid,
dotted and dashed curves correspond, respectively, to DH equal to zero, DD and 10 DD. As observed
below, CD/C increases in a non-linear fashion with the increase in C (solid line). The effective
diffusivity is higher in the case where both type of species are mobile (DH ¼ 10 DH, dashed curve) as
compared to when only the dissolved species are mobile (DH ¼ 0, dotted curve). In the former case,
the diffusivity decreases with concentration since the relative amount of adsorbed species decreases
with increasing concentration (according to the CD/C curve). This is also the reason why the
diffusivity increases when the concentration increases when the adsorbed species are immobile
(dotted curve) (Fig. 10.11).
7.9 (Fig. 10.12)

Fig. 10.11 Normalized

uptake as a function of
square root of time

Fig. 10.12 Normalized concentration of dissolved species and diffusivities, as defined above, as a function of total
concentration
528 10 Solutions to Problems Given in Exercises

The curves in Fig. 10.13 correspond to a glass transition concentration Co1 equal to 0.5 and 0.9 at
L ¼ 20. The higher the glass transition concentration, the steeper the concentration profiles will
be. This, in turn, when being integrated, leads to more linear uptake vs time curve (Fig. 10.14).

Fig. 10.13 The change in D as a function of C at L ¼ 20 (dotted curve), 10 (dashed curve) and 1 (solid curve)

Fig. 10.14 Mass uptake with Co1 equal to 0.5 (curve 1) and 0.9 (curve 2). The unsmooth thin curve is due to numerical
instabilities in the simulated concentration profiles
10.7 Chapter 7 529

7.10.
Theory of Barrer/Brandt. In the theory of Barrer, a molecule can diffuse in a polymer if the amount
of energy absorbed in the region around the molecule is equal or larger than a critical energy value.
This energy is divided between the solute molecule and the polymer chains and gives rise to an
increase in the degree of freedom of the polymer chains. These can rearrange synchronically and
force the solute into a new position. The diffusion coefficient is dependent on the displacement of the
polymer chains, the vibration frequency of the solute molecule and the activation energy.
Brandt, on the other hand, proposed that to overcome the critical energy for the solute to move, it is
necessary that the polymer chains can bend and that the intermolecular energy has to be overcome.
The thermal energy, however, is not necessary for the activation. In Barrer’s theory, both the size of
the activated zone around the solute and the degrees of freedom are temperature dependent.
Theory of DiBenedetto/Paul and Pace/Datyner. In the theory of DiBenedetto/Paul the polymer
chain consists of n-centre polymer segments which correspond to a repeating unit. A cylinder of the
size sufficient to host the solute molecule has to be created for the solute to move. This cylinder is
created parallel to the chain axis. The interactions between the centres of the polymer segments are
described by the Lennard-Jones potential and therewith the energy that is needed to form the cylinder.
The interactions between the solute and the polymer chains and compressive stresses are neglected.
In contrast, the theory of Pace and Datyner extends the theory of DiBenedetto/Paul. Solute
molecules are able to diffuse both parallel and perpendicular to the polymer chains. The solute
diffuses along the chain axis until it reaches, for example, an entanglement, where the chains open up
space so that the solute can move also perpendicular to the chains. To diffuse along the chains no
activation energy is needed. The energy that is needed for the chains to open up for perpendicular
diffusion, corresponds to the experimentally determined activation energy. This energy is also
described by the Lennard Jones potential and include factors like, for example, chain stiffness and
cohesive energy.
7.11. The calculated relative oxidation rate and oxidation profiles are given in Figs. 10.15 and 10.16,
respectively. Observe that the oxidation rates are highest at the surface and consequently also
the degree of oxidation.

Fig. 10.15 Relative oxidation rate profiles with respect to the surface boundary oxidation rate as a function of time
Arrow indicates the profile development with increasing time
530 10 Solutions to Problems Given in Exercises

Fig. 10.16 The degree of

oxidation as a function of
time and space

Fig. 10.17 A sketch of the

tortuous path along two
clay sheets

7.12. Since the diffusivity and solubility of the clay filler is zero, it is possible to write: D ¼ Dp/τ and
S ¼ Spvp, where index p is the polymer matrix, τ is the tortuosity and ϕ is volume fraction. The
permeability is P ¼ DS and consequently (Fig. 10.17):

ð 1  ϕf Þ
P¼ ð10:92Þ
τ

Ignoring the clay layers yields a path which is, from bottom to top, H + L + H + L. In the presence
of clay, the path is H + L + H + W/2 + L + W/2. Thus, the ratio in path length and consequently the
tortuosity is:
W þ 2L þ 2H W
τ¼ ¼1þ ð10:93Þ
2L þ 2H 2ð L þ H Þ

The volume fraction of filler is approximated as ϕ f = LþH

L
. Inserting this into Eq. (10.93) yields:
ϕf W
τ ¼1þ ð10:94Þ
2L

This is implemented into Eq. (10.92) to give the final expression:


1  ϕf
P¼ ð10:95Þ
1 þ ϕf 2L
W

7.13. The solution to the boundary and initial conditions is (compare with Eq. (7.85)):
10.7 Chapter 7 531

Fig. 10.18 Concentration profiles in the region 0 < x < l generated with Eq. (10.96)

Fig. 10.19 Solute self-diffusivity (solid curve) and mutual binary diffusivity (broken curve) as a function of weight
fraction solute

   
4 X ð1Þ
1 n
ð2n þ 1Þπ ð2n þ 1Þ2 π 2
Cðx, tÞ ¼ C0 cos x exp  Dt ð10:96Þ
π n¼0 ð2n þ 1Þ 2l 4l2

The concentration profiles for this special case is shown in Fig. 10.17. The curves are generated
using five terms in Eq. (10.96). Note that five terms are not enough to yield a sufficiently converging
solution/profile at small times (as time goes to zero).
7.14. The best fit is obtained if the diffusion coefficient is 1.2•107 cm2/s.
7.15. The permeability depends on both solute diffusivity and solubility. Although CO2 has a lower
diffusivity than O2, its solubility is much greater than that of O2, resulting in a larger
permeability.
7.16 (Fig. 10.19).
532 10 Solutions to Problems Given in Exercises

10.8 Chapter 8

8.1. Compounding (mixing) can be divided into two general types: distributive mixing, which
distributes the polymer and additives evenly on a coarse level. This is normally the first stage
of mixing, which is followed by a method which is referred to as dispersive mixing, by which all
the additives are uniformly mixed on a fine scale with the polymer. If the additives are soluble,
they should be in a true solution with the polymer. The assessment of dispersion of filler particles
involves high resolution methods, such as scanning electron microscopy. The particles appear
white in the micrographs, and it is possible by painstaking analysis of a number of micrographs
to express the degree of dispersion in mathematical terms (Fig. 10.20). Heating of a powder
mixture can be helpful if some of the additives are melting by the heat treatment. The melt can

Fig. 10.20 Scanning electron micrographs of a polyethylene nanocomposite containing coated ZnO nanoparticles
(appear white in the micrographs). From Pourrahimi et al. (2016), with permission from Wiley. These micrographs
show that the nano-particles are well dispersed in the polymer matrix
10.8 Chapter 8 533

potentially cover unmolten polymer particles and hence be efficiently distributed in the polymer
phase.
The homogeneity on a coarser scale can be assessed by thermogravimetric analysis of
5–10 mg samples. The remaining ash after heating to 700  C in nitrogen provides information
about the ZnO content in each of the 5–10 mg samples. An extensive variation of the ash
percentage indicates an uneven distribution of the nanoparticles in the sample. The distribution
of other soluble additives can be assessed by IR microscopy or DSC (oxidation induction time,
OIT; cf. Chap. 1) in the case of antioxidants.
8.2. Rotational moulding, which is a low-pressure process, is cost-effective for large products.
Injection moulding is a high-pressure process and is much more efficient for small-to-
medium-sized objects. It is not possible to make hollow objects by injection moulding. However,
two pieces prepared by injection moulding could be assembled by welding to form a hollow
object. Blow moulding (which makes hollow objects) is effective for samples sized in the
medium size range. Injection moulding is superior to both rotational moulding and blow
moulding with regard to dimensional precision. Injection moulding is a high-pressure technique
and the tools (mould and machine) are expensive, but the cycle time is very short, which makes
injection moulding apt for long series, that is the 106 products. Rotational moulding has low tool
costs but very long cycle time, and it is the winner for the 100 products. Blow moulding is best
for the intermediate case, the 1000 objects.
8.3. The following expressions are useful:

πpr 4
Q¼ ð10:97Þ
η8l

and
1
η / M3:4 ∧Q / ) Q / M3:4 ð10:98Þ
η

MFI (i.e. Q) is thus proportional to M–3.4 and the normalised case (M ¼ 10,000 g mol1):

MFI ð10 000Þ ¼ 10 0003:4 ¼ 2:51189  1014 ð10:99Þ

The following relative values were obtained for the other polymers:
1000 g mol1: 2512  MFI (10000)
50,000 g mol1: 0.0042  MFI (10000)
300,000 g mol1: 105  MFI (10000)
106 g mol1: 1.6  107  MFI (10000).
This strong variation in MFI is difficult to deal with; a high molar mass sample has a very high
viscosity and a very low MFI. In order to extend the possible molar mass scale, the mass used is
changed from the most used 2.16 kg to 21.6 kg. This corresponds to a tenfold increase in the
pressure ( p) in Eq. (10.97).
8.4. On the back-side of the hole, a cold knit line is formed. The orientation of the molecules is due to
the fountain flow parallel with the knit line, and thus the strength is reduced. Liquid crystalline
polymers show very low strength values, of the order of 10–20% of the full strength. Brittle
amorphous polymers have also knit lines with significantly reduced strength (by 50%). Short
fibre composites show a strong reduction in strength, almost similar to that of the liquid
crystalline polymers. What can be done to make the sample tougher? (i) Injection mould the
534 10 Solutions to Problems Given in Exercises

Fig. 10.21 Effect of change of gate position on the knit line strength

detail but without the hole and drill a hole afterwards. (ii) Change the position of the gate
according to Fig. 10.21.
8.5. Examples of processing involving extensional flow: film blowing, blow moulding, extrusion
along a narrowing channel (convergent flow), and expansion of polymers forming cellular
structure. The polymers that show an elongational viscosity with a pronounced increase with
increasing strain are apt to work in these processes.
8.6. A snap is fastening the two ends (closing the two ‘legs’), the hinge enables a free motion of the
‘legs’, and finally the ribs provide stiffness to the two ‘legs’. Injection moulding can in one shot
obtain all three functions and is a very efficient method for small objects produced in great
numbers. PP and POM are excellent polymers for snaps and hinges.
8.7. Left: the melamine thermoset is produced by compression moulding; the curing occurs in the
mould. The cover is produced by injection moulding. The cover must fit perfectly the bowl in
order to form a circular snap. Any warpage or dimensional change is banned. The requirement on
the components: scratch resistance of the bowl (melamine resin is a good choice), elegant shiny
and coloured (again melamine resin is a perfect choice), FDA approved (food contact), micro-
wave safe (booth polymers), no swelling of the bowl in water or in hydrophobic media (again
melamine resin is a perfect choice because its highly crosslinked). Any swelling of the cover in
water is not acceptable; polyolefin (PE) is a good choice. Right: Both parts are produced by
injection moulding. This combined object is for storage of food, and the tightness is very
important, which means that the swelling from handling liquid-containing food must be con-
trolled small. The fact that this particular item worked for 20 years is impressive; a careful
materials selection and a very careful processing.
8.8. The uniformity in wall thickness is due to two facts: (a) The temperature of the metallic mould is
(must be) uniform; (b) polymers have a low thermal conductivity. A thin polymer layer is apt to
collect powder, which melts on contact. A thicker polymer layer will hold a lower surface
temperature, and the polymer powder will bounce off without melting. Together will both
these control the thickness of the plastic layer to a certain value.
8.9. The start-up of the extrusion is carried out in steps wherein initially the barrel and the die are
pre-heated to the standard operating temperature of the polymeric material. The screw speed is
gradually increased to the operating level to avoid excessive load on the motor and to discourage
10.9 Chapter 9 535

high pressure build-up within the barrel. Usually, before the polymeric material is introduced in
the extruder, purging is done. This is to confirm that any contaminated and degraded material
that are remaining inside the barrel and within the screws from the previous operation are
expelled. Purging is a cleaning process of the barrel and the screw, and it is interrupted when
clear extruded polymer (without any specs of dirt/contamination) leaves the die.
8.10. Thermal oxidation is the problem. The extrusion of a big, thick-walled pipe is slow, and the
access to oxygen is the problem. If the stability is not sufficient, too little efficient antioxidant
(usually a combination of hindered phenols and phosphites) is available, the inner wall material
is oxidized and it will initiate early fracture. This can be avoided by either of the following
actions: (i) Carry out extrusion with inert gas atmosphere (e.g. nitrogen) inside the pipe, or
(ii) add more stabiliser to the polymer; the latter is not so simple; the solubility of these
compounds is limited in polyethylene. How can oxidation be detected? In IR spectroscopy,
you use an IR microscope to reveal the oxidation profile through the cross-section. Another
method to cut out a specimen from the wall and to tensile test it. The inner wall surface is brittle;
it makes a very clear crack pattern on the surface, which is even visible without microscope.
With SEM, the thickness of the oxidized can be measured.

10.9 Chapter 9

9.1. (a) No, the choice of most sustainable material depends on the application requirements. In many
cases, replacement of plastics with alternative materials leads to increased greenhouse gas
emissions during production and service-life.
(b) There are many benefits, such as low density and a wide range of materials and properties.
Thanks to this, plastics contribute in many ways to, for example, energy effective buildings and
transportation, to food production and preservation, to health care and energy production. Read
more in Chapter 9.2.
(c) Changing from petroleum based to biobased plastic materials and from linear material use to
circular material flows.
9.2. (a) No biobased materials can be biodegradable or non-biodegradable. It is the chemical structure
and not origin that determines the degradability in different environments. Bio-polyethylene is not
more biodegradable than petroleum-based polyethylene.
(b) No, the whole life-cycle from production of biomass to end-of-life management of the product
should be considered. For example, sustainability could be reduced if the production competes
with food production or if the production is energy and fresh water consuming.
9.3. (a) We have great variety of plastics and plastic formulations developed for different applications
and different processing methods. This complicates the recycling process. Further complications
are caused by, for example, low quality (contamination, degradation), unknown additives and
multicomponent materials (read more in 9.4).
(b) Good quality, sorted and clean plastic waste can be reprocessed to new products by mechanical
recycling (read more in 9.4.1). Polymers can also be chemically recycled back to original
monomers (monomer loop), which can be re-polymerized to products that have same quality
than the original polymer or to other chemicals (molecule loop) (see 9.4.2). Product design (single
material/multi-material), type of additives, quality, purity and chemical structure among others
determine which route is most suitable.
9.4 (a) Biodegradable plastics are biodegradable in the defined specified environment, for example
compost in case of compostable plastics. Waste management and industrial composting facilities
would still be needed. It is difficult to see that biodegradable plastics would provide a general
solution to plastic waste problem, but they can with benefit be used in applications where
536 10 Solutions to Problems Given in Exercises

degradability is part of the function, for example plastic bags for food or garden waste to be
composted.
(b) The conditions in different natural environments differ greatly and many environments, such as
marine environment, are not very favourable for plastic degradation. It would be very difficult to
ensure rapid degradation in all kinds of environments, while maintaining good properties during
the service life.

References

Gedde, U. W. (2020). Essential classical thermodynamics. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019a). Chapter 7: Morphology of semicrystalline polymers. In Fundamental
polymer science. Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019b). Chapter 8: Crystallization kinetics. In Fundamental polymer science.
Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019c). Chapter 5: The glassy amorphous state. In Fundamental polymer science.
Cham: Springer Nature Switzerland AG.
Gedde, U. W., & Hedenqvist, M. S. (2019d). Chapter 6: The molten state. In Fundamental polymer science. Cham:
Springer Nature Switzerland AG.
Pourrahimi, A. M., Hoang, T. A., Liu, D., Pallon, L. K. H., Gubanski, S., Olsson, R. T., Gedde, U. W., & Hedenqvist,
M. S. (2016). Advanced Materials, 28, 8651.
Index

A B
Aberrations, 65 Bacterial cellulose, 494
Ab initio method, 216–218 Ballistic regime, 380
ABS plastics, 48, 357, 359, 360, 457 Barostat, 232
Activated zone, 402, 403 Barrer model, 403
Activation energy Basis functions, 212, 214–216, 218–220, 254–259
apparent, 403, 407 Beer-Lambert law, 114
Additive manufacturing (AM), 477–480, 482, 483, Biased Monte Carlo, 236, 239
499, 500 Binder jetting 3D printing process, 480, 481
Affinity constant, 421, 422 Biobased monomers, 493, 495
All-atom AA (force-field), 228, 234, 241 Biobased polyethylene, 493
Amorphous polymers Biobased polymers
fictive temperature challenges, 494
definition and assessment, 4 challenges and opportunities, 494
fictive temperature-cooling rate dependence, 46, 47 Bio-based thermoplastic elastomers, 494
vitrification, 4 Biobased thermosets, 494, 495
Anelastic Bio-based unsaturated polyester, 495
definition, 268 Biodegradable, 492–494, 496, 501, 502
Angular momentum vector, 123 Bio-poly(ethylene terephthalate) (bio-PET), 493
Anomalous diffusion, 401, 421, 424–426, 442 Bio-polyethylene (bio-PE), 493
Antioxidant concentration Blow moulding, 456, 457, 473, 474, 484
assessed by DSC, 53 Boltzmann-distribution, 232
Antiplasticisation, 329 Boltzmann superposition principle, 23, 27, 314–316,
Argand diagram 321, 369
amorphous and semicrystalline poly(ethylene Born–Oppenheimer approximation, 209
terephthalate), 40 Bragg’s law, 109, 145, 148, 153, 161, 163
Arrhenius law, 318 Brandt model, 403, 405
Artificial structures, 63, 82, 85 Brittle fatigue failure, 362
Assesment of volatile species by thermogravimetric Brittle/ductile transition temperature, 340, 350
analysis, 5, 16, 533 Brownian motion, 378, 380
Atomic force microscopy, 61–63, 73–76, 84–87, B-splines, 254, 258
90, 91, 99 Bueche model, 407, 408
assessment of mechanical properties, 290 Bulk flow, 380, 399, 400
cantilever, 73–76 Bulk modulus
contact mode, 73–75 definition, 294
harmonic imaging, 76 effect of crystallinity, 296
intermittent contact mode, 73, 75 Burgers element, 308–311
non-contact mode, 73, 74
sample preparation, 99
solution grown single crystals, 86 C
spherulites in melt-crystallized polymer, 24, 90, 91 Calculation of d-spacings
tapping mode, 73, 75, 76 based on h, k and l values, 153, 155

# Springer Nature Switzerland AG 2021 537

U. W. Gedde et al., Applied Polymer Science, https://doi.org/10.1007/978-3-030-68472-3
538 Index

Calendering, 457, 476, 477, 483 Crack propagation resistance parameter

machine sketch, 481 definition, 348
Calorimetry, 1, 5, 6, 8–14, 33, 42, 54 Crack tip opening displacement (CTOD), 342, 347
Canonical ensemble, 232, 236 Craze, 325, 333, 339, 346, 347, 353, 355, 360–363
Capacity factor Craze incubation time, 361
definition of concept, 173 Crazing
Catalytic conversion, 499 effect of entanglement density, 325
Cellulose, 363, 365 Creep data
chemically modified, 494 described by stretched exponential function, 321
Cellulose acetate, 494 Creep failure, 288, 352–355
Cellulose fibre systems, 363–365 Critical crack size, 349
Chain disentanglement Critical shear stress, 272
in fracture, 340 Critical stress amplitude, 361
Chain entanglement, 298, 323, 324, 327–329, 332, 334, Critical stress intensity factor, 344–346, 349
340, 355, 363 Critical temperature, 386, 431
Chain orientation Crystallization kinetics
assessment by spectroscopy, 105, 122, 136, 165 assessed by DSC, 41
Chemical recycling, 497, 499, 502 assessed by DSC at ultrahigh cooling rates, 43
Chromatography Crystallography basics, 109
definition, 171 Cup method, 436
Circular economy, 490, 496 Cyclic stress
Cis-polyisoprene, 432 definition, 360
Closed shell systems, 214, 215
Cluster
size, 420 D
Coarse-graining (CG), 205, 222, 223, 228–229, 241 Dashpot, 301
Cohesive energy density, 389, 402, 407, 410 Deborah number, 426
Collocation (FEM strategy), 258, 259 Degenerated energies, 212
Column Degradation of plastics in nature, 500, 502
general, 189 Degradation products
open tubular capillary, 172, 191 assessment by gas chromatography, 197, 198
Complex compliance assessment by high performance liquid
definition, 307 chromatography, 187, 197
Complex modulus, 20, 22 Delta Dirac functions, 258
definition, 305 Density functional theory (DFT), 205, 207, 218–221, 259
Compliance Depolymerization, 497, 499
definition, 298 Design for recycling, 498
Compostable, 492, 496, 501 Detector
Compounding, 460, 461, 483, 484 flame ionization, 192
Compression moulding, 459, 474–476, 483 nitrogen-phosphorus, 192
Confocal Raman spectroscopy, 114 thermal conductivity, 192
depth profiling, 118 Dielectric spectroscopy, 25, 30, 31, 38
Constant tensile load test, 354 Dielectric thermal analysis, 1, 5, 25–31, 48, 50, 54
Copolymer structure argand diagram, 28, 29, 40
randomness index hyperbranched polyester, 30, 56
NMR, 132 atomic polarisation, 26, 27
Coulomb operator, 214 Cole–Cole equation, 28
Coupled SEC systems complex dielectric constant, 27, 28
matrix assisted laser desorption ionization-mass Davidson-Cole equation, 29
spectrometry, 187 dielectric constant, 26, 27, 30, 40
Covalent adaptable networks, 500 dielectric loss, 27–31, 51
Crack blunting, 343 dielectric relaxation process, 26, 30
Crack growth resistance dipolar polarisation, 26
definition, 348 electronic polarisation, 26, 27
Crack initiation energy release rate Havriliak-Negami equation, 29, 40
definition, 348 Kohlrausch–Williams–Watts (KWW) equation, 29
Crack initiation time Lorentz-Lorenz equation, 25
definition, 355 relaxed dielectric constant, 27
Crack opening displacement (COD), 341, 347, 355, 356 Schering bridges, 30
Index 539

time-domain spectrometers, 30, 31 Dynamic mechanical thermal analysis

unrelaxed dielectric constant, 27, 40 forced oscillation instruments, 19
Vogel–Fulcher–Tammann–Hesse (VFTH) loss modulus, 20, 22
temperature-dependence, 30 storage modulus, 20, 23
Wheatstone bridges, 30 torsion pendulum, 19, 20
Die swell, 454 Dynamic mechanical therrmal analysis
Differential interference-contrast microscopy, 65, 66, glass transition
69, 99 Vogel–Fulcher–Tammann–Hesse equation, 22, 23
Differential scanning calorimetry (DSC), 1, 5–14, 25, 33,
35–37, 42–51, 53–56
Differential thermal analysis (DTA), 1, 6–14, 23 E
Diffusion Effective media models, 206, 242
barrier, 377 Elastic behaviour
cage-like, 383, 384 definition, 268
case II, 401, 421, 424–427, 442, 446 Elastic strain energy, 343
concentration-dependent, 401, 418, 419 Elastic stress
confined, 383, 384 internal energy, 326
constrain by crystal-amorphous interphase, 417 Elastic-plastic fracture mechanics (EPFM), 341, 342, 346,
constraining effect, 413 347, 366
depence on spherulitic structure, 415–419 Elasto-viscous state
DiBenedetto and Paul model, 403, 405, 446 definition, 328
dual mobility, 424 Electron anti-symmetry, 210, 213
Einstein relation, 378 Electron densities, 207, 218–220, 261
Fick’s first law, 391, 423, 439 Electron diffraction, 144, 148, 162, 163, 165
Fick’s second law, 439, 445, 446 Electron microcsopy
fluctuation theory, 407 bright-and dark-field, 64, 73
liquid-like, 383, 384, 408 Electron microscope
Pace and Datyner model, 405, 406 resolution, 71, 72, 99
partial immobilisation model, 423 Electron spectroscopy for chemical analysis (ESCA), 139
selective, 430 Electron spin, 210, 213, 214, 219
shell-like free volume, 411 Electron spin resonance (ESR) spectroscopy, 105,
Vrentas-Duda free volume model, 409, 422, 447 142–143, 165
Diffusivity detection of radicals, 143
activation energy by Meares model, 402, 403 Electrospray ionization mass spectrometry (ESI-MS), 187
assessment by pulsed field gradient NMR technique, Electrostatic Coulomb interactions, 227
437, 439 End-of-life management, 493, 496, 497, 501
assessment by time lag method, 419 Endurance strength
correlation with solute molar volume, 385 definition, 361
dependence on diameter of solute molecule, 385, Energy consumption, 490
404, 406 Ensemble average, 233
Digital model Ensembles (thermodynamic), 232
development, 478 Enthalpic fracture, 341
Dilatometry, 16–18, 41, 44, 54, 287, 288 Entropic fracture, 341
Dilatometry and thermomechanical analysis Entropy production
linear variable differential transformer (LVDT), 17 general aspects, 2, 3
Dispersive mixing, 460, 461 melting of semicrystalline polymers, 33, 34
Dissipative particle dynamics (DPD), 206, 242–243, 260 Entropy-elastic force
Distributive mixing, 460, 461, 467 explanation of origin, 298
Drop-in plastics, 493, 495 Environment
Drum tumbler, 460 abiotic, 500
Dual sorption, 386, 387 biotic, 500
hole saturation constant, 422–424 natural, 496, 500–502
Ductile failure, 349–352 Environmental stress cracking (ESC), 355
viscoplastic models, 355 EPDM rubber, 137
Dynamic covalent chemistry, 500 Epoxy, 472, 481
Dynamic fatigue failure, 360–363 Ergodic theorem, 233
Dynamic mechanical analysis, 18, 54, 287–288 Essential work of fracture (EWF), 348–350
torsion pendulum measurements, 288 Ethylene-propylene rubber, 405, 406
Dynamic mechanical data–Arrhenius diagram, 21–23, 30 Euler–Lagrange formulation, 257
540 Index

Evaporation constant, 395 Fourier transform infrared spectroscopy, 112

Exchange operator, 214 Fourier’s law, 391, 439
Exchange–correlation energy, 220 Fracture
Exchange–correlation functional, 219, 220 chain scission, 340
Explicit Euler forward method, 398 impact of tie molecules, 351
Extended Hückel method, 213 Fracture energy
Extended-chain crystals critical value, 359
heating rate dependence of melting, 33 definition, 344
Extraction molar mass depence, 340
headspace, 194–196 Fracture mechanics, 288, 289, 341–350, 366, 371
liquid–liquid, 172, 194, 195 full notch creep test, 354
liquid–solid, 195 inner fracture process zone, 348
microwave-assisted, 195 notch, 288, 345
multiple headspace, 194 notch tip radius, 357
solid-phase, 195, 196 outer plastic dissipation zone, 348
solid-phase microextraction, 195–197 rapid crack growth, 353
solvent, 194 single edge notch test, 354
solvent-free, 196 slow crack growth, 353
sonication, 195 Fracture mechanics testing
soxhlet, 195 surface energy reducing agent, 354
Extraction methods Fracture testing, 288–289
dissolution–precipitation, 195 impact, 289
Extrusion Fracture toughness
general principle, 470 definition, 176
screw design, 468 impact of ß-transition, 323
zones, 469 trapped entanglements, 351
Extrusion blow moulding, 474 Free volume
Extrusion mixing, 460 fractional, 409, 410, 417, 418, 445
holes, 378, 408, 410, 421, 445
models, 407–412
F Free-induction decay
Far-infrared (FIR) spectrscopy, 110 free-induction decay (FID), 128, 129, 133
assessment of hydrogen bonding, 126 Free volume model
Fast multipole techniques, 218 Cohen, Turnbull and Fujita, 408
Fast scanning calorimetry, 5, 12, 33, 42, 54
Fatigue failure
thermal fatigue, 362 G
Fick’s second law Galerkin (FEM strategy), 250, 255–257
solved by method of separation of variables, 395 Gas chromatography (GC), 171, 172, 187, 189–199
Film blowing, 470, 483 calibration curve, 193, 194
Finite difference methods (FDM), 206, 252, 253, 256, 260 carrier gas, 190, 192
Finite element methods (FEMs), 206, 242, 247–260 split/splitless injector, 190
Finite volume methods (FVM), 206, 252, 253, 260 volatile compounds, 190, 191, 196, 197
First-order phase transition, 1, 2, 6, 9, 44 Gases
Fitting of X-ray diffraction peaks condensable, 386, 433
pearson VII function, 155 Gas separation, 428, 430–433
Flash-DSC, 14 Gaussian white noise signal, 382
Flory–Huggins interaction parameter, 418, 448 Gel permeation chromatography (GPC), 171
Flux, 391, 392, 394, 395, 399, 423, 438, 439, 441 Generalized gradient approximations (GGA), 220
Fock matrix, 215, 216 Ghost chain, 236, 240, 241
Fock operator, 214 Gibbs ensemble, 232
Folded-chain polymer crystals Gibbs–Rivlin theory, 321
thermodynamic nature, 4, 5, 31–34 Glass transition temperature
Force field (FF), 205, 221–225, 228, 229, 234, 235, 237, assessment, 3, 4, 21, 22
241–243, 259, 260 Glass transition
Fourier transform (FT) NMR assessed by DSC, 1, 8, 44, 46–48, 54
radio frequency (RF) pulse, 128 assessed by modulated DSC, 5, 9, 12, 33, 47, 48
Fourier transform infrared spectrometers definition, 325
Michelson interferometer, 112 Gradient elution, 188
Index 541

Grand canonical ensemble, 232, 236 Infrared spectroscopy (IR)

Greenhouse gas, 489, 490, 493, 495, 496, 501 asssessment of chain brancing in polyethylene
Group contribution methods, 244, 410, 448 105, 117
Group Interaction Modelling (GIM), 335 asssessment of oxidation in polyolefins, 117, 165
Guggenheim-Anderson-de Boer (GAB) model, 420 asssessment of unsaturated groups in polyethylene,
117
attenuated total reflection (ATR), 112, 113,
H 118–121
Hagen–Poiseuille equation, 455 basic principles, dipole moment, 121
Hamilton operator, 207–210, 214, 218 curve resolution, 118
Hansen’s solubility parameters, 244 group frequencies, 110
Hartree’s method, 214 infrared dichroism, 110, 122
Hartree–Fock equation, 214, 215, 218 mid-IR range, 105, 112
Hartree–Fock method, 214 transmittance, definition, 114
Hat functions, 254, 256, 258 wave number, definition, 114
Hazardous, 197 Initial modulus, 295, 312, 316, 370
Heat capacity Injection blow moulding, 474
assessment by DSC, 11 Injection moulding, 453, 456–458, 460–466, 471,
non-reversing, 11 483, 484
reversing, 11 corner design, 465
Heat convection transfer coefficient, 442 design philosophy, 464
Heat flux rib design, 466
Stefan-Boltzmann’s law, 441 screw characteristics, 463, 464, 466, 467
Heat-flux differential scanning calorimeter, 6 stages in cycle, 463, 472
Henry’s law, expression, 388 tap design, 466
Henry’s law constant, 388, 424 Instrumented falling weight impact tests, 358
Herman’s orientation functions, 300 Integral sorption measurements, 435
High performance liquid chromatography (HPLC), 172, Interlamellar slip, 336
187–199 Intermolecular forces, 224, 241
assessing flame retardant, 197 Internal stress, 465, 476
coupled with mass spectrometer, 193, 199 Intramolecular forces, 224, 241
ion-exchange, 189 IR microscopy, 76
isocratic elution, 188 Isothermal–isobaric ensemble, 232
normal-phase, 189, 190 Iterative Boltzmann inversion, 228
reverse-phase, 189, 190
UV detector, 187
High-density polyethylene, 121, 416 J
Hohenberg–Kohn existence theorem, 218 J-integral, 347, 350
Hooke’s generalised law, 292, 294
Hooke’s law, 269, 292–294, 296, 342
Hookean response, 267 K
Hoop stress Kinetic Monte Carlo (kMC), 222, 235, 242
definition, 345 Kirchhoff’s law, 3, 56
Hopfenberg’s map, 400 Knit line
Hückel model, 212 cold, 459, 466
Hybrid functionals, 220 warm, 466
Hybrid QM/MM method, 217, 259 Kohlrausch–Williams–Watts (KWW) function, 303
Kohn–Sham equation, 219

I
Impact failure L
Izod test, 356 Lagrange multiplier, 211, 224
tensile impact method, 357 Langevin equation, 379
Impact strength Langmuir isotherm, 421
definition, 356 Lattice plane index system, 152
notched specimen, 288, 357, 363 Leathery state, 328
Impact testing Lennard–Jones (LJ) potentials, 226, 227, 229, 231, 242,
Charpy V, 355 243, 404
Importance sampling, 236–239, 242 Life cycle assessment, 491, 493, 495
542 Index

Light scattering Modelling, 205–261

assessment of molar mass and radius of gyration Modulated DSC, 5, 9, 12, 33, 47, 48
108, 165 Mohr strain circle, 278, 284, 285
basic principles, 139 Mohr stress circle, 276–278, 294
Lignin, 492, 495 Molar mass assessment
Linear combination of atomic orbitals (LCAO), 211, by colligative properties, 178
212, 214 by end-group analysis, 178
Linear dilatometer, 16, 17 by intrinsic viscosity, 181, 184
Linear elastic fracture mechanics (LEFM), 341, 342, 346, by ultracentrifugation, 178
348, 350, 354, 358, 366, 371 Molar mass averages
Linear elastic materials, 288 definitions, 185
Linear viscoelasticity, 267, 301–311, 368 Molar mass dispersity
Linear-scaling quantum techniques, 217 assessment by size exclusion chromatography, 177, 178
Liquid chromatography (LC), 171, 172, 187, 188, 195 definition, 177
two-dimensional, 187 Molecular dynamics (MD), 205, 207, 221–236, 240–243,
Liquid crystalline polymers 259–261
dielectric spectroscopy, 25, 38 Molecular flexibility
phase transitions, 49 definition, 323
thermal behaviour, 49, 50, 55, 56 Molecular modelling (MM), 205, 217, 240, 259
thermal optical analysis (TOA), 5 Molecular orbital method (MO), 205, 207, 210, 216,
Liquid separation, 428–430 217, 219
Local density approximation (LDA), 220 Molecule loop, 497, 499, 502
Logarithmic strain Monomer loop, 497, 499, 502
definition, 284 Monte Carlo (MC), 205, 222, 228, 233–242, 248, 259, 260
Loss compliance Monte Carlo Metropolis-Hastings, 236–239
definition, 314 Mooney equation, 326, 327
Loss modulus Multi-angle laser light-scattering (MALLS), 181
definition, 305 Multicomponent product
Loss tangent recycling issue, 497
definition, 305 Multiscale modelling, 205, 234, 242, 259
Mutual diffusion coefficient, 399

M
Magnetic force microscope (MFM), 62, 76 N
Magnetic resonance imaging (MRI), 106 NBR-rubber, 96
Magnetic suspension balances, 434 Near-infrared (NIR) spectroscopy, 110, 112, 121
Material extrusion (FDM) 3D printing process, 481 Necking, 300, 319, 333–335, 337, 350
Material informatics, 243 formation of needle crystals, 338
Material jetting 3D printing process, 481 Neutron scattering, 108, 109, 144, 163–165
Material use Newton’s equations, 205, 221, 229, 231, 235
circular model, 489 Nitrile rubber, 408, 428
13
Maxwell element, 301–306, 309, 311, 367 C NMR spectroscopy
Mechanical fatigue, 362 group frequencies, 110, 111
Mechanical property assessment 1H NMR spectroscopy
uniaxial testing, 286 group frequencies, 110, 111
Mechanical recycling, 497–499, 502 NMR instrumentation
Melting point-true value, 33, 34 broadband decoupling, 130
Membrane separation, 377, 428–433 Nominal strain
Mercury dilatometer, 16 definition, 278
Mesoscale models, 206, 242, 260 Non-ideal mixing, 400
Michel-Levy birefringence chart, 68 Non-linear viscoelasticity, 318–322
Microcanonical ensemble, 232 Non-recoverable strain, 309, 332
Microscopy Normal stress
comparison of different methods, 79, 88 definition, 271
discipline, 61 Nuclear magnetic resonance (NMR) spectroscopy
history, 61–63 2D NMR, 133, 135
Microtomy, 79 2D COSY, 133
Mobile amorphous phase-assessed by dielectric Nuclear magnetic resonance spectroscopy (NMR)
relaxation data, 40, 41 assesssment of tacticity, 130
Mobile phase Carr-Purcell-Meiboom-Gill (CPMG) pulse, 137
definition, 171 coupling constant, 126, 128, 130, 131
Index 543

cross-polarization (CP), 130, 137 tensile index, 363–365, 370

diffusion coefficient tensile stiffness index, 364
Stokes-Einstein relation, 133 Paris relationship, 362
distortionless enhancement by polarization transfer Parison variator, 474
(DEPT), 133 Partial differential equations (PDEs), 206, 208, 247
dynamic nuclear polarization (DNP), 130 249–253, 255, 257, 260
fourier transform spectrometer, 128, 129, 133 Partial vapour pressure, 387
gradient diffusion probe, 133 Partition coefficient, 173
magic angle, 126, 130 Pauli principle, 213, 214
magnetic moment, 123 Penetrant
magnetogyric ratio, 123 random walk, 380, 381, 383
nuclear magnetic spin quantum number, 123 Periodic boundary conditions, 222, 223, 226
nuclear Overhauser effect (NOE), 130, 132 Permselectivity, 430, 431
Pascal’s triangle, 126–128 Pervaporation, 428, 430
proton quantification, 132 Phase-contrast microscopy, 65, 66, 68, 99
relaxation time in the rotating frame, 129 Phase-space, 233
resonance frequency, 128, 130 Physical ageing, 44, 46, 47, 56
shielding, 124, 128, 130–132 Pipe testing
spin decoupling, 130 hydrostatic pressure testing, 285
spin-lattice relaxation time, 124 Plane strain
spin-spin relaxation time, 124 definition, 284
use of deuterated solvents, 129 Plane stress
definition, 277
Plastic deformation
O definition, 267
Optical microscopy (OM) Plastic packaging, 491, 496
aperture iris, 66, 67 Plastic pipe lifetime
bertrand lens, 63 antioxidant concentration, 198
bright-and dark-field, 64 Plastic pollution, 496, 500
compensator, 68 Plastic production
conjugate planes, 66, 67 annual, 489
eyepiece, 61, 64, 66 linear, 489
field iris, 63, 66, 67 Plastic waste, 489, 496–499, 501, 502
general, 61, 63 Plasticisation, 387, 412, 418, 420–422, 425, 426
K€ohler illumination, 67 Plasticizer concentrtion
lambda (λ) plate, 68–70 assessed by HPLC, 197
numerical aperture (NA), 65, 66 Poisson’s ratio
objective lens, 65, 66, 71 definition, 279
principle of magnification, 64 impact on volume strain, 279, 295
resolution, 61, 65, 66 values for different materials, 279, 280
sample preparation, 99 Polarized microscopy, 65–69, 91, 99, 100
spherulites liquid crystalline polymers, 91, 99
negative, 68–70 spherulites, 68–70, 87, 89, 90, 100
positive, 68, 69 Poly(1-trimethylsilyl-1-propyne) (PTSMP), 432, 433
Optimum material-processability and solid state Poly(3-hydroxybutyrate) (PHB), 494
properties, 456 Poly(4-methyl-1-pentene), 86, 87
Ordinary differential equations (ODEs), 247, 249, 252, Poly(butylene bifuranoate), 131
253, 259 Poly(butylene furanoate), 131
Osmosis, 428–430 Poly(butylene succinate) (PBS), 493
Overlap integrals, 211, 212, 217 Poly(butylene succinate-co-butylene adipate)
Oxidation induction time (OIT), 51–53, 56 (PBSA), 493
Poly(butylene terephthalate), 455
Poly(dimethyl siloxane), 431
P Poly(ether-ether-ketone), 411
Paper Poly(ethylene naphthalate) (PEN), 427, 428
fibre pull-out, 364 Poly(ethylene oxide), 91
fibre-to-fibre strength, 364 Poly(ethylene terephthalate) (PET), 11, 12, 33, 36, 39–42,
surface weight, 364 48, 110, 141, 245, 246, 427, 428, 447, 493
tensile energy absorption index, 364 Poly(lactic acid), 493
544 Index

Poly(methyl methacrylate), 339, 340, 409 Polystyrene, 47, 48, 383–385, 421, 422, 425, 471, 482,
Poly(oxy methylene), 300, 331 500
Poly(phenylene oxide), 431 atactic, 47
Poly( p-hydroxybenzoic acid-co-ethylene terephthalate), high-impact, 330, 331, 355, 360
92 Polyvinyl chloride (PVC), 477
Poly(trimethylsilyl propyne), 432 Portable NMR equipment, 129, 137
Poly(vinyl acetate), 404, 418 Post-consumer plastics
Poly (vinyl chloride), 328, 329, 355, 471 annual collection, 497
Poly(vinyl chloride-co-vinyl acetate), 408 Powder bed fusion 3D printing, 482
Poly(ε-caprolactone), 23, 24 Powder diffractometer
linear, 24 Johann type, 150
star-branched, 23–25 Power-compensation differential scanning calorimeter, 6
Polyamide, 482 Principal strain
Polyamide 6, 33, 53, 55 definition, 284
Polyamide 6,6, 117–119, 355 Principal stress
Polybutadiene definition, 275
trans, 35, 36 Profiling by confocal Raman microscopy in PE/PMMA
Polycarbonate, 47, 311, 312, 323, 334–336, 340, 349, two-layer structure, 117, 118
350, 355, 362, 366, 471 Progressive damage model, 341
semicrystalline, 311, 312 Pseudo-spectral methods, 255, 259
Polyester Ptychographic X-ray nanotomography, 76
hyperbranched, 30, 56 PVC/poly(ethylene-co-vinyl acetate), 94
Polyester resin
unsaturated, 54
Polyether ether ketone (PEEK), 482 Q
Polyethylene (PE), 4, 5, 17, 25, 31–40, 52, 53, 55, 56, 62, QSAR method, 206, 243, 260
76, 77, 80–90, 93, 95, 97–100, 110, 111, 115, Quantum chemistry (QC) methods, 207, 218, 259
117, 121, 123, 136, 137, 143, 154–156, 161,
164, 165, 220, 228, 235, 236, 240–242, 280,
296, 300, 323, 325, 326, 334, 337, 338, R
351–356, 369, 371, 384, 386, 403, 406, 409, Radial distribution function
412, 413, 415–420, 426, 428, 437, 441, 445, concept, 157
447, 468, 470, 471, 485, 493, 494, 496 Raman microscopy, 63, 76
Polyethylene furanoate (PEF), 493 Raman scattering, 110, 113, 114, 121, 122
Polyethylene nanocomposites, 95, 97, 98 Raman spectroscopy
Polyhydroxyalkanoates (PHA), 494 asssessment of unsaturated groups in polyethylene,
Polylactide (PLA), 499 117
Polymer crystallization monochromator, 121
revealed by atomic force microscopy, 43 fourier transform spectrometer, 112, 128
Polymer degradation longitudinal acoustic mode (LAM)
studied by DSC, 51, 53, 54 crystal thickness, 123
Polymer loop, 497, 498, 502 Rayleigh scattering, 108, 110, 145, 164, 165
Polymer melt Rayleigh–Ritz (FEM strategy), 250, 255, 258
creep compliance, 454 Recoil growth algorithms, 240
deformation mechanisms, 299 Recycling
relaxation modulus, 454 biological, 497
Polymer processing chemical, 497, 499, 502
relation to fundamental science disciplines, 458–460 material, 492, 495, 496
relation to polymer physics, 458, 459 Relaxation, 222, 241
relation to rheology, 453, 458 Relaxation strength
Polymers dielectric relaxation, 27, 40
bio-based, 495 mechanical relaxation, 39, 40
biodegradable, 494 Relaxation time
compostable, 501 definition, 302
from waste and greenhouse gases, 495 Resolution
petroleum-based, 493, 495 definition, 173
Polyoxymethylene, 84 Resonance integrals, 211–213
Polypentapeptide, 135 Resource, 489, 491–494, 496–501
Polypropylene Retardation time
isotactic, 33–36, 43, 83, 84, 89, 90, 100, 406, 417, 484 definition, 307
Index 545

Retention factor melt-crystallized PE, 86

definition, 173 Thomson-Gibbs equation, 31, 34
Retention time total enthalpy method, 37
definition, 173 upper and lower bound of modulus vs. crystallin,
Retention volume 38, 39
definition, 173 Semi-crystalline polymers–thermal behaviour, 31–43
Reverse osmosis, 428–430 Semi-empirical methods, 216, 217, 259
Reversible process Shear bands, 360, 363
characteristics, 2 Shear strain
Ribbon blender, 460, 477 definition, 282
Rigid amorphous fraction (RAF), 413, 417, 424 Shear stress
Rigid amorphous phase-assessed by dielectric relaxation definition, 272
data, 40–42 Shear-bands, 360, 363
Rigid amorphous phase-assessed by DSC, 40–42 Sheet lamination 3D printing, 482, 483
Roothaan–Hall equation, 215, 216, 220 Silicone rubber, 141
Rotational moulding, 456, 472–474, 483–485 Simulation, 205–260
Rubber elasticity, 292, 317, 326–327 Size exclusion chromatography (SEC), 171, 172, 175, 198
Langevin spring, 334 calibration
phantom network, 326, 327 Mark-Houwink-Sakurada equation, 184
statistical mechanical models, 326 using narrow standard, 182, 183, 186
Rubber elasticity theory calibration curve, 182, 184, 186
Gaussian chain, 326 calibration standards, 183, 185
Rubber plateau, 328–330 calibration using light scattering, 178, 181
calibration using narrow standard, 182, 186
detector
S refractometer, 181
Sample preparation, solvent etching, p-xylene, 81, 82, 88 equipped with nuclear magnetic resonance
Sanchez-Lacombe equation of state model (SL-EOS), spectroscopy, 187
388, 390 equipped with thermal field-flow fractionation, 187
Scanning electron microscopy (SEM) equipped with triple detector, 182
etched polymer blends, 93, 94 high temperature analysis, 179
general, 63 long-chain branching, 175, 177, 181, 182, 185, 187
polymer nanocomposites, 95, 97 molar mass distribution, 171, 175, 177, 178, 182, 183,
rubber filled with carbon black, 96, 98 187, 198
sample preparation, 79 molecular architecture, 171, 175, 181, 182, 185–187
Scanning thermal microscopy, 76 relative calibration, 182, 183
Scattering methods, history, 108 universal calibration method, 184
Schr€odinger equation, 205, 207–210, 212, 215, 218, 252 Size exclusion chromatography system
Secant modulus components, 191
definition, 321 Slater determinants, 214, 219
Second-order phase transition characteristics, 1, 2 Slater-type orbitals (STO), 218, 219
Secular equation, 212 Small-angle neutron scattering (SANS), 108, 109, 164
Selectivity factor chain conformation in condensed states, 164
definition, 173 chain conformation in semicrystalline polymers, 164
Self-avoiding random walks, 240 Small-angle X-ray scatterering (SAXS)
Self-diffusion, 410, 433, 447 desmearing, 161
Semicrystalline polymers collimated X-ray beam, 159
Avrami equation, 43 Kiessig camera, 159
Bimodal melting, 35, 36 Kratky camera, 160
crystal thickening, 32 Lorentz correction, 161
mechanical relaxation strength, 39–42 pin-hole camera, 159, 160
Tsai-Halpin equation, 38, 40 slit-collimation camera, 159, 160
melting peak temperture Statton camera, 159
heating rate dependence, 33 Small-angle X-ray scattering (SXAS)
mobile amorphous phase revealed by dielectric crystal thickness, 109, 144, 160, 161
spectrscopy data, 12, 38 long period, 109, 144, 160, 161
mobile amorphous phase revealed by DSC, 12, 37 Solbility
multimodal melting, 35–37 glassy polymers by non-equilibrium lattice fluid
spiral growth about screw dislocation, 90 model, 389
546 Index

Solid state NMR, 129, 130, 136 Surface mecahnics

Solubility hardness testing, 290
dependence on heat of solution, 387 Surface mechanics methods, 289–290
glassy polymers by non-equilibrium thermodynamics, Sustainability, 489–502
391 Swelling, 377, 390, 393, 399, 418–420, 426, 435, 442
permanent gases, 387 induced mechanical stress, 377, 418, 426
perturbed hard sphere chain theory, 391
pressure decay method, 435
statistical thermodynamics models, 387, 391 T
glassy polymers by non-equilibrium perturbed hard Tensile impact test, 356
sphere chain model, 391 Tensile testing, 286–287
Solubility coefficient, 386, 387, 390, 392, 432, 435 Theoretical plates
Solubility parameter, 386, 389, 400, 448 definition, 175
Solubility selectivity, 428, 430, 431, 433 Thermal analysis
Sorption isotherm, 420–423, 435 definition, 1, 37
Sorption kinetics focussing on structure, 1
assessment by quartz crystal microbalance, 433 Thermal conductivity, 439–441
assessment by quartz spring balance technique, 434 dependence on sound velocity, 440, 441
Spectral element method, 259 Thermal diffusivity, 439, 440
Spectral methods, 206, 259, 260 Thermal mechanical analysis (TMA), 287–288
Spectroscopic methods Thermal optical analysis (TOA), 1, 5, 6, 23–25, 41, 54, 56
history, 108 linear spherulite growth, 42
Spherulites, 68–70, 79, 87, 88, 90, 93 Thermal oxidation
Spin orbital, 214 assessment by infrared spectroscopy, 117
Splitting of ESR spectrum, 142 Thermobalance
Starch, 493–495 buoyancy force correction, 15
amorphous, 379 Thermodynamic second-order glass transition, 4
Stationary phase, 171–173, 175, 188–190, 192, 195 Thermoforming, 471–473, 483
Statistical associated fluid (SAFT) models, 391 Thermogravimetry (TG)
Statistical methods, 206, 243–247, 260 assessment of transport properties, 5, 53
Stereolithography (SLA), 479, 480 basic principles, 14, 15
Storage compliance thermobalance, 14
definition, 368 Thermomechanical analyser, 17
Storage modulus Thermomechanical analysis, 1, 16–18
definition, 305 Thermomechanical processes, 1, 16–18, 357, 472
Strain hardening Thermometry, 6
definition and origin, 334 Thermoplastic starch, 494
Strain hardening modulus, 334, 354 Thermostat, 232, 234
Strain softening in glassy polymers, 333 Thin-layer chromaotgraphy (TLC), 172
Strain tensor Three D printing, 457, 477–483, 499, 500
definition, 283 Three-point bending testing, 285, 286, 354
Stress balance Time average, 233
internal force, 270 Time-temperature superposition
Stress concentration factor glass transition temperature, 317, 318
definition, 357 Time-temperature superposition principle, 316
Stress intensity factor interval, 357 shift factor, 317
Stress relaxation, 267, 287, 301–303, 306, 309, 311–314, Tortuosity, 412–415, 417
316, 321, 330, 340, 361, 366, 367 Transmission electron microscopy
Stress relaxation spectrum general, 63, 99
definition, 303 aluminium oxide nanoparticles, 95, 96
Stress tensor sample preparation, 99
definition, 274 polyethylene-vapour decoration, 84, 86
Strong form, 253, 257 staining, chlorosulphonic acid, 80–82, 85, 87, 99
Styrene-acrylontrile copolymer (SAN), 466 staining, osmium tetroxide, 80, 82, 99, 510
Styrene-butadiene rubber, 21, 22 staining, ruthenium tetroxide, 80, 82
Sub-glass process, 318, 328 solution grown single crystals, 85, 86
Subglass process–Arrhenius equation, 21–23, 29 stained polymer blend, 81
Superheating Transmission electron microscopy of etched sample
polymer crystal melting, 32–34 melt-crystallized PE, 86
Index 547

Transmission electron microscopy of stained sample Wet skid resistance, 310

melt-crystallized PE, 86 Wide-angle X-ray scattering (WAXS), 108, 137, 144,
Transport properties 148–163, 165
effect of crystallinity, 377–447 crystal thickness, 109, 144, 160
True strain Debye–Scherrer camera, 149
definition, 284 flat-film camera, 149
True stress focussing camera (Guinier-Hägg camera), 150
definition, 271 focussing camere according to Guinier, 109
2D NMR spectrroscopy interfacial component, 154, 155, 160
heteronuclear single-quantum correlation Widom insertion–deletion algorithm, 236
(HSQC), 134 Williams–Landel–Ferry (WLF), 30
Williams–Landel–Ferry (WLF) equation, 22, 318, 369
W€
ohler diagram, 361
U Wood
Ultrafiltration, 428, 430 mechano-sorptive creep, 365
Ultraviolet-visible (UV-VIS) spectroscopy, 105, 143
UNIFAC model, 244, 387
United-atom UA (force-field), 228, 241 X
X-ray generation
high-vacuum sealed X-ray tubes, 148
V monochromator, 148–150, 159
Van der Waals potentials, 226 synchrotrone source, 148, 159
Van Krevelen’s group-contribution, 244 X-ray photoelectron spectroscopy (XPS), 105, 106,
Variational principle, 211, 214, 218 139–141, 165
Vat-photopolymerisation printing process, 479 high resolution spectrum, 141
Vectra A: a copolyester based on p-hydroxy benzoic acid relative sensitivity factor, 140
and hydroxy naphthoic acid, 93 survey spectrum, 139, 140
Velocity autocorrelation function, 381–383 X-ray ptychographic tomography
Verlet algorithm, 231 strained nanocomposite, 95, 98
Vibrational spectroscopy X-ray scattering
conformation bands, 117, 118 monochromator, 148–150, 159
monitoring of chemical reaction, 117, 120 X-ray scattering and unit cell structure, 151–153, 155,
Vibrational spectroscopy (IR and Raman) 157, 159
vibrational modes, 110 X-ray scattering detection
Viscoelastic modelling cameras equipped with photographic film, 149
standard linear solid model, 309 diffractometer, 150
Viscosity
apparent vs. shear rate, 455, 458
constitutive equation, 304 Y
at zero shear rate, 454, 455 Yielding
at zero shear rate vs. molar mass, 454 definition, 267
Viscous flow, 402, 409 dislocation theory, 334
heat evolved, 8 Eyring model, 334
Viscous state free volume theories, 334
definition, 327 theories, 334
Vitrimers, 500 Young’s modulus
Voigt–Kelvin element, 306–308, 310, 367 definition, 292
Volume dilatometry, 16 effect of crystallinity, 512
Volume expansion coefficient, 407 rubbers and glassy polymers, 292

W Z
Wave functions, 207, 210–212, 214–218 Zener model, 248, 249
WAXS study of semicrystalline polyethylene, 109, 122, Zero-entropy production path melting, 4
148, 156 Zimm plot, 147
Weak form, 253, 257