Rheo Physics

Soft and Biological Matter
Philippe Coussot
Rheophysics
Matter in All Its States
Soft and Biological Matter
Series editors
Roberto Piazza, Milan, Italy
Peter Schall, Amsterdam, The Netherlands
Roland Netz, Berlin, Germany
Wenbing Hu, Nanjing, People’s Republic of China
Gerard Wong, Los Angeles, USA
Patrick Spicer, Sydney, Australia
For further volumes:

http://www.springer.com/series/10783
‘‘Soft and Biological Matter’’ is a series of authoritative books covering estab-
lished and emergent areas in the realm of soft matter science, including biological
systems spanning from the molecular to the mesoscale. It aims to serve a broad
interdisciplinary community of students and researchers in physics, chemistry,
biophysics and materials science.
Pure research monographs in the series as well as those of more pedagogical
nature, will emphasize topics in fundamental physics, synthesis and design,
characterization and new prospective applications of soft and biological matter
systems. The series will encompass experimental, theoretical and computational
approaches.
Both authored and edited volumes will be considered.
Philippe Coussot
Rheophysics
Matter in All Its States
123
Philippe Coussot
Laboratoire Navier
Université Paris-Est
Champs-sur-Marne
France
Original French edition ‘‘Rhéophysique: La matière dans tous ses états’’ published by EDP
Sciences, EDP Sciences 2012. A co-publication with EDP Sciences, 17, av. du Hoggar
F-91944 Les Ulis, France
ISSN 2213-1736 ISSN 2213-1744 (electronic)

ISBN 978-3-319-06147-4 ISBN 978-3-319-06148-1 (eBook)
DOI 10.1007/978-3-319-06148-1
Springer Cham Heidelberg New York Dordrecht London
Library of Congress Control Number: 2014941116
Springer International Publishing Switzerland 2014

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. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the
work. Duplication of this publication or parts thereof is permitted only under the provisions of
the Copyright Law of the Publisher’s location, in its current version, and permission for use must
always be obtained from Springer. Permissions for use may be obtained through RightsLink at the
Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law.
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.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Foreword to the French Edition
Many materials cannot be strictly categorised as solid or liquid, and this is

something many of us discovered in our earliest childhood, when we played with
the mashed potato on our plates, made models with plasticine or built sandcastles.
Such intermediate states of matter are found in many natural phenomena and
industrial applications, in a range of examples from the flow of blood whose
viscosity depends on the configuration of red blood cells to self-consolidating
concretes which liquefy under vibration to fill out every corner of the formwork,
not to mention mayonnaise, tomato ketchup, and mousses, or cosmetic emulsions,
creams, and gels. Everywhere we look, in health, civil engineering, and the food
industry! In fact, in every walk of life.
Despite the great importance of these complex materials, science has been slow
to take them on board and much remains to be done. And yet they are part of our
everyday lives and it is through these substances that we come face to face with the
reality of a broad range of physical, physicochemical, and mechanical phenomena.
Indeed, these materials have even given rise to a new form of science teaching in
schools, based on the ‘hands-on’ approach.1 There are many advanced textbooks
on rheology that treat the wide variety of systems and their behaviour, usually
basing the discussion on mechanical measurements. The novelty of the present
book is that it follows rather the opposite route, in the spirit of the research carried
out over the past few decades by physical chemists inspired by Pierre Gilles de
Gennes, starting with physical and chemical descriptions and using them to
explain the mechanisms that underlie the behaviour of these systems, today
grouped under the heading of soft matter, and in a way which emphasises their
common features.
Needless to say, the scientific development of the subjects covered in each of
the chapters have very different and quite independent histories, and within each
chapter, quite distinct sources are clearly identified. Just to give one example, the
one that is least foreign to me, consider the study of granular media and finely
divided matter. A body of knowledge was built up independently in civil
1
First experimented by Leon Lederman in Chicago, the Hands-on programme gave rise to La
main a la pâte in France. See www.lamap.fr.
v
vi Foreword to the French Edition
engineering, soil science, agronomy, chemical engineering, planetary science, and

doubtless other fields of investigation. And over the last couple of decades, a
considerable effort has been made by a broad community of physicists and
physical chemists to unify this field of research. To this end, they have combined
model experiments, often unsophisticated, sometimes deriving from classic
experiments in specific fields of application, numerical approaches, and theoretical
models inspired by the statistical physics of microscopic systems. The book
Granular Media—Between Fluid and Solid, published recently [1] in the same
series as the [French edition of the] present book, describes this new approach and
its achievements so far. And the same kind of remarks could be made about each
of the chapters in this ambitious work.
One of the great merits of the present book is to provide a description of the
different kinds of material—suspensions, colloids, polymers, emulsions, and
foams—in simple physical terms and with a limited formalism common to all the
chapters. All these materials share the essential feature of being dispersed systems
with varying extents of interface depending on how fine the granularity happens to
be. Compactness turns out to be a key parameter, and in each case we find a range
of states from the dilute (without interaction) to the semi-dilute or compact, in
each of which the degree of congestion imposes limits that generally depend on
how the material was prepared. While this simplified approach sometimes neglects
fluctuations in size or organisation, the starting point, which one might qualify as
the physics of a bag of marbles, certainly brings out the main qualitative aspects of
the different phases. And of course, one can always come back later to include
geometric details of the constituent elements.
Specific mechanical behaviour is associated with each of these regimes, such as
liquid, pasty (an adjective with a somewhat fuzzy physical meaning, but much
clearer mechanical definition) or solid. The relevant regime depends also on the
nature of the external loading and its associated timescale—something which lies
at the very heart of rheology. Another important point about this book is that it
brings together apparently disparate topics, allowing constructive comparisons
between the different chapters. For instance, the flow of foams can be likened to
the flow of a suspension of deformable grains, and elongation phenomena in
polymer solutions can be usefully compared with problems involving colloidal
grains.
Philippe Coussot has the good fortune to work in an area of applied research
associated with civil engineering. The confluence, or one might even say, the
confrontation, between fundamental properties and applied properties of materials,
taking into account effects of scale or durability, surely help to better meet the
needs of the construction industry.
This exceptionally clear book will prove an indispensable reference for engi-
neers and technicians working with these complex materials. They will be able to
apply the empirical laws to calculate or predict the behaviour of such materials,
but beyond the purely utilitarian aspect, they will also supplement their
Foreword to the French Edition vii
understanding through a physical approach that emphasises the mechanisms

underpinning the phenomena usually described in more standard rheology text-
books. The present book thus offers a cross-disciplinary and novel account, likely
to interest a broad public, from engineers to those involved in research.
Paris, France, March 2014 Etienne Guyon
Reference
1. Andreotti, B., Forterre, Y., Pouliquen, O.: Granular media - between fluid and solid.
Cambridge University Press, Cambridge (2013)
Preface
Our understanding of the physical or physicochemical processes underlying the

mechanical behaviour of materials, which comes under the heading of rheophys-
ics, is based on a wide range of scientific investigation that has progressed to
different extents for different materials. Apart from gases, liquids, and simple
solids, it is in the field of polymer science that progress has been most significant
over the past fifty years or so, and there is no lack of specialised works on this
topic. Much more recently, there have been several books detailing the rheology of
foams, granular media, and colloids. The present book sets out from the idea that,
before we can understand the rheophysics of any given material, it is particularly
instructive, even essential, to master the basic tools for handling a broad range of
materials. This approach has been built up gradually in the context of my own
lectures on Rheophysics and Soft Matter which form part of an M.Sc. in the
physics of building materials, a novel course set up by the late O. Coussy at the
Ecole des Ponts et Chaussées and the Ecole Polytechnique, now run by Xavier
Chateau. My aim here has thus been to bring together in a single textbook, not the
work in progress at the frontier of each field of research, but the basic elements of
our physical understanding of the main classes of material, and this in the most
unified way possible.
The first version of this book appeared in French with a foreword by Etienne
Guyon, reproduced here in English. The whole work has been translated by Ste-
phen Lyle, who also helped to improve certain passages. The author acknowledges
the financial support of Saint-Gobain for the book translation.
Paris, March 2014 Philippe Coussot
ix
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Suspensions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.1 Phase Separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.2 Effect of Particles on Behaviour of the Mixture . . . . . . 11
1.4.3 Other Effects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Colloids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5.1 Colloidal Interactions . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5.2 Yield Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5.3 Thixotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Polymers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6.1 Properties of Polymer Chains . . . . . . . . . . . . . . . . . . . 21
1.6.2 Polymers in Solution . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.6.3 Viscoelasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6.4 Other Properties of Polymers . . . . . . . . . . . . . . . . . . . 25
1.7 Emulsions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.8 Foams. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.9 Granular Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.10 Real-Life Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2 Simple Materials . . . . . . . . . . . . . . . . . . . . . . . . . ............ 35

2.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . ............ 35
2.2 Interactions Between Elementary Components
and States of Simple Matter . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2.1 Elementary Components . . . . . . . . . . . . . . . . . . . . . . . 37
2.2.2 Thermal Agitation . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2.3 Interaction Potential . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.2.4 Van der Waals Forces . . . . . . . . . . . . . . . . . . . . . . . . 39
2.2.5 Chemical Bonds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.6 Born Repulsion Force. . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.7 Balance of Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
xi
xii Contents
2.2.8 Hydrogen Bond and Hydrophobic Forces . . . . . . . . . . . 41

2.2.9 States of Simple Matter . . . . . . . . . . . . . . . . . . . . . . . 42
2.3 Gaseous State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.3.1 Velocity Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.3.2 Mean Free Path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.3.3 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.3.4 Ideal Gas Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.3.5 Kinetic Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.4 Liquid State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.4.1 Transition from Gaseous State to Liquid State . . . . . . . 55
2.4.2 Structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.4.3 Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.4.4 Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.4.5 Rheophysical Model. . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.4.6 Interfacial Tension . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.5 Solid State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.5.1 Structures and Interactions . . . . . . . . . . . . . . . . . . . . . 66
2.5.2 Microrheology in the Solid Regime . . . . . . . . . . . . . . . 67
2.5.3 Elongation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.5.4 Behaviour Under Simple Shear . . . . . . . . . . . . . . . . . . 69
2.5.5 Compressibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.5.6 Maximal Mechanical Strength . . . . . . . . . . . . . . . . . . . 72
2.5.7 Solid–Liquid Transition . . . . . . . . . . . . . . . . . . . . . . . 75
2.5.8 Solid–Gas Transition . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.6 Glassy State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.6.1 Glasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.6.2 Glass Transition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.6.3 Mechanical Behaviour Associated
with the Glass Transition . . . . . . . . . . . . . . . ....... 77
2.6.4 Viscosity of Glasses . . . . . . . . . . . . . . . . . . . ....... 79
3 Suspensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.2 Preparing a Suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.2.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.2.2 Volume Fraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.2.3 Energy Involved in Creating the Suspension . . . . . . . . . 85
3.2.4 Dispersing the Particles . . . . . . . . . . . . . . . . . . . . . . . 87
3.2.5 How Many Particles can be Put in Suspension? . . . . . . 87
3.2.6 Resistance of the Liquid to Particle Displacement . . . . . 89
3.2.7 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.3 Effect of Particles on the Behaviour of the Mixture . . . . . . . . . 95
3.4 Effect of Concentration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.4.1 General Considerations. . . . . . . . . . . . . . . . . . . . . . . . 97
Contents xiii
3.4.2 Concentration Regimes. . . . . . . . . . . . . . . . . . . . . . . . 99

3.4.3 Dilute Suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.4.4 Non-dilute Suspension . . . . . . . . . . . . . . . . . . . . . . . . 101
3.5 Effect
of Particle Anisotropy . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.5.1 Ideal Anisotropic Particles: Spheroids . . . . . . . . . . . . . 104
3.5.2 Effect on Viscosity of Anisotropic Particles
with Constant Uniform Orientation . . . . . . . . . . . . . . . 105
3.5.3 Particle Rotation in a Fluid Under Simple Shear . . . . . . 105
3.5.4 Effect of Concentration . . . . . . . . . . . . . . . . . . . . . . . 108
3.6 Effect of Non-uniform Particle Concentration . . . . . . . . . . . . . 109
3.7 Shear Thickening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.8 Suspensions in a Yield Stress Fluid . . . . . . . . . . . . . . . . . . . . 115
3.8.1 Displacement of an Object Through
a Yield Stress Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.8.2 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.8.3 Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4 Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.2 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2.1 Apparent Length of a Chain . . . . . . . . . . . . . . . . . . . . 123
4.2.2 Distribution of Apparent Chain Lengths . . . . . . . . . . . . 124
4.2.3 Radius of Gyration. . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.2.4 Extension of a Chain Under Traction . . . . . . . . . . . . . . 127
4.2.5 Persistence Length . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.3 Polymers in Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
4.3.1 Configurational Free Energy . . . . . . . . . . . . . . . . . . . . 132
4.3.2 Free Energy Associated with Interactions
Between Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.3.3 Total Free Energy and Chain Size . . . . . . . . . . . . . . . . 135
4.4 Several Chains in Solution . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.4.1 Dilute Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.4.2 Semi-dilute Regime . . . . . . . . . . . . . . . . . . . . . . . . . . 138
4.4.3 Concentrated Regime . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.4.4 Entanglement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
4.5 Cross-Linked Polymers and Polymer Gels. . . . . . . . . . . . . . . . 142
4.6 Mechanical Behaviour of Liquid Polymers . . . . . . . . . . . . . . . 144
4.6.2 Dilute Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
4.7 Effect of Temperature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
xiv Contents
5 Colloids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.2 Brownian Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.2.1 Basic Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.2.2 Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.2.3 Rotational Diffusion. . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.2.4 Osmotic Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.2.5 Sedimentation and Brownian Diffusion . . . . . . . . . . . . 165
5.3 Van der Waals Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5.4 Electrostatic Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.5 Effects Due to Adsorbed Polymers. . . . . . . . . . . . . . . . . . . . . 173
5.6 Depletion Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
5.7 Balance of Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.8 Behaviour of Repulsive Systems . . . . . . . . . . . . . . . . . . . . . . 180
5.8.1 Hard Repulsive Suspensions . . . . . . . . . . . . . . . . . . . . 180
5.8.2 Soft Repulsive Suspensions. . . . . . . . . . . . . . . . . . . . . 184
5.9 Attractive Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.9.1 Structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.9.2 Behaviour of Attractive Suspensions . . . . . . . . . . . . . . 194
5.10 Pasty–Hydrodynamic Transition. . . . . . . . . . . . . . . . . . . . . . . 198
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6 Emulsions and Foams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

6.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
6.2 Physical Properties on the Scale of the Inclusions . . . . . . . . . . 203
6.2.1 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6.2.2 Pressure Difference Across an Interface . . . . . . . . . . . . 203
6.2.3 Deformation of a Fluid Inclusion at Zero Speed
and Constant Volume . . . . . . . . . . . . . . . . . . . . . . . . . 204
6.2.4 Displacement of an Inclusion in a Liquid at Rest . . . . . 206
6.2.5 Sedimentation or Creaming. . . . . . . . . . . . . . . . . . . . . 207
6.3 Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
6.3.2 Forming Inclusions by Deformation . . . . . . . . . . . . . . . 209
6.4 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
6.4.1 Coalescence and Stabilisation . . . . . . . . . . . . . . . . . . . 212
6.4.2 Ostwald Ripening . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.5 Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
6.5.2 Concentration Regimes. . . . . . . . . . . . . . . . . . . . . . . . 220
6.5.3 Dilute Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Contents xv
7 Granular Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

7.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
7.2 Main Types of Direct Interaction . . . . . . . . . . . . . . . . . . . . . . 233
7.2.1 Lubricated Contact. . . . . . . . . . . . . . . . . . . . . . . . . . . 233
7.2.2 Frictional Contact . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
7.2.3 Collision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
7.3 Role of Configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
7.3.1 Basic Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
7.3.2 Dilatancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
7.3.3 Settling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
7.3.4 State of the System . . . . . . . . . . . . . . . . . . . . . . . . . . 240
7.4 Regimes of Behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
7.5 Frictional Regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
7.5.1 Simple Shear. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
7.5.2 Constitutive Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
7.5.3 Applications to Quasi-Static Flows . . . . . . . . . . . . . . . 248
7.6 Collisional Regime. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
7.7 Intermediate Regimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
7.7.1 Transition from Frictional to Collisional Regime . . . . . . 254
7.7.2 Transition from Frictional to Lubricated Regime . . . . . . 254
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
8 Rheometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
8.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
8.2 Basic Geometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
8.2.1 Parallel Disk Rheometer . . . . . . . . . . . . . . . . . . . . . . . 263
8.2.2 Cone–Plate Rheometer . . . . . . . . . . . . . . . . . . . . . . . . 265
8.2.3 Concentric Cylinder Rheometer . . . . . . . . . . . . . . . . . . 265
8.2.4 Flow in a Duct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
8.3 Perturbing Factors in Rheometry . . . . . . . . . . . . . . . . . . . . . . 269
8.3.1 Perturbations of the Sample Volume . . . . . . . . . . . . . . 270
8.3.2 Slipping on the Walls. . . . . . . . . . . . . . . . . . . . . . . . . 271
8.3.3 Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
8.3.4 Shear Bands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
8.3.5 Instability Associated with a Decreasing Flow Curve . . . 276
8.3.6 Other Perturbing Factors. . . . . . . . . . . . . . . . . . . . . . . 277
8.4 Experimental Procedures. . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
8.4.1 Choosing the Geometry . . . . . . . . . . . . . . . . . . . . . . . 278
8.4.2 Preparing the Sample . . . . . . . . . . . . . . . . . . . . . . . . . 279
8.4.3 Flow Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
8.4.4 Solid–Liquid Transition . . . . . . . . . . . . . . . . . . . . . . . 281
8.5 Practical Measurement Techniques. . . . . . . . . . . . . . . . . . . . . 282
8.5.1 Squeeze Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
8.5.2 Inclined Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
xvi Contents
Appendix A: Fluid Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Appendix B: Elements of Thermodynamics . . . . . . . . . . . . . . . . . . . . 311
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
Symbols
b Distance between the centres of two molecules or particles

Length of a section of a polymer chain
c Speed (magnitude of velocity vector)
d Size of a molecule or particle (apparent diameter)
e Charge on the electron
Internal energy density
g Acceleration due to gravity
h Distance between the surfaces of two particles
i Slope of an inclined plane
kB Boltzmann constant 1:38 1023 JK1
m Mass of a molecule or particle
n Number of molecules or particles per unit volume
n0 Number of neighbours of a molecule or particle
Ion concentration in a solution
p Pressure
r Distance between two molecules
Apparent root mean squared length of a polymer
x, y, z Coordinates along three orthogonal axes
u, v, w Components of the velocity along three orthogonal axes
w Maximum interaction potential energy between two molecules
A Area of a surface
D Diffusion coefficient
E Young’s modulus Energy
F, f Force
FD Drag force
G Shear modulus
G0 Elastic modulus
G00 Viscous modulus
H Height or thickness of material
K Uniform compression modulus
Hamaker constant
xvii
xviii Symbols
M Torque
Molar mass
N Total number of molecules or particles in a given sample
Number of sections of a polymer chain
Normal force on a grain or granular medium
N Number of repeat units (monomer units)
PðxÞ Probability density of variable x
Q Heat
R Radius of a spherical particle
RG Radius of gyration of a polymer
S Entropy
T Temperature
Tangential force on a grain or granular medium
U Internal energy
US Slip speed
V Speed
W Work
Z Number of microstates of a system
c Deformation under simple shear
d Dirac distribution
e Deformation under elongation
Average roughness
e0 Electric permittivity
g Apparent viscosity
h Viscoelastic relaxation time
j1 Debye length
k Mean free path of a molecule
Relaxation time of a polymer
l Viscosity of a Newtonian fluid, possibly containing inclusions
l0 Viscosity of a Newtonian interstitial fluid
Viscosity of the interstitial liquid of a suspension or emulsion
l
Viscosity of the fluid phase of an inclusion
m Poisson coefficient
Volume of a molecule or particle
q Density
qS Density of solid phase
qL Density of liquid phase
rAB Interfacial tension between materials A and B
r0 Charge density per unit area
rxx ; rxy ; . . . Components of the stress tensor
s Shear stress
sc Yield stress
u Internal friction angle of a granular medium
/ Volume fraction
Symbols xix
/m Maximum packing fraction

/c Critical concentration of steric hindrance
x Oscillation frequency
C Angular speed
X Sample volume
Xm Volume available per molecule in a sample of material
h Viscoelastic characteristic time
U Interaction potential energy
Pe Péclet number
Ca Capillary number
St Stokes number
Ba Bagnold number
Le Leighton number
i; j; k Unit vectors in a 3D orthonormal frame
n Unit normal vector to a surface
u Velocity vector
R Stress tensor
T Extra-stress tensor
D Strain rate tensor
Operators
r Gradient
r Divergence
D Laplacian
o Partial derivative
tr Trace
d Derivative
O(x) Number smaller than x
hxi Average of variable x
Energies
P Dissipated power
F Helmholtz free energy
Chapter 1
Introduction
Abstract This chapter reviews the main classes of material and sets out the qual-
itative basis for rheophysical investigation. Most complex fluids are produced by
mixing together mesoscopic elements with a liquid. The simplest materials from this
point of view are suspensions, which consist of solid particles immersed in a liquid.
When liquid volumes are replaced by solid particles, this makes it more difficult
for the mixture to flow. When in addition the particles interact with one another at
a distance through the liquid, as happens with colloidal particles, the mixture may
have more complex mechanical properties, lying somewhere between those of a solid
and those of a simple liquid. Similar effects are found with emulsions and foams.
For their part, polymers constitute a class on their own, owing to the fact that their
chains can stretch like springs and get entangled, leading to novel mechanical behav-
iour: elastic in certain situations, liquid in others. Finally, granular materials made
up of concentrated assemblages of solid grains in direct contact with one another can
remain jammed together to constitute a solid or flow like liquids depending on the
circumstances.
1.1 Introduction
Between the solid which deforms only very slightly and the fluid that flows easily
there is a broad range of materials with intermediate mechanical properties, including
polymers, emulsions, foams, and granular materials, among others. The aim of this
book is to provide the physical background required to understand the mechanical
behaviour of these intermediate materials. Simple liquids and solids have a homoge-
neous structure on the atomic or molecular scale. For such materials, one can deduce
the macroscopic behaviour from properties on the atomic or molecular scale. But
in many other cases, the observed deformations cannot be explained by the atoms,
nor even the molecules making up the material. It is rather the relative motions
of elements comprising large numbers of atoms or molecules that set the scene.
P. Coussot, Rheophysics, Soft and Biological Matter, 1

DOI: 10.1007/978-3-319-06148-1_1,
© Springer International Publishing Switzerland 2014
2 1 Introduction
These ‘mesoscopic’ elements, that is, neither macroscopic nor microscopic, can be
bubbles, droplets, polymer chains, solid grains, globules, cells, and so on. The
mechanical behaviour of the corresponding materials arises from the interactions
between these mesoscopic elements. Depending on the material, the macroscopic
behaviour results from the averaged local behaviour, cooperative phenomena involv-
ing large numbers of elements, or again some collective structure. The specific
features of the constitutive elements, such as their deformability, their elasticity,
the long-range interactions they engage in, and their impact on collective phenom-
ena can lead to extremely varied and novel macroscopic properties that fall between
those of the standard solid or liquid.
This chapter gives an overview of the main classes of materials and puts
forward qualitative guidelines for rheophysical analysis. We begin with solid mate-
rials in Sect. 1.2. When crystalline, they provide a unique standard thanks to the
simplicity of their structure, made up as it is by juxtaposing identical copies of some
specific local arrangement. When we move on to simple liquids in Sect. 1.3, we
already encounter some of the problems that arise in the rheophysics of fluids, since
their mechanical behaviour results from relatively complex physical phenomena,
namely, thermal agitation and disorder of atoms or molecules. Most complex fluids
comprise a mixture of mesoscopic elements with a liquid. Concerning components
and associated interactions, the simplest materials are suspensions (Sect. 1.4). These
consist of solid particles immersed in a liquid. When liquid volumes are replaced by
solid particles, this increases the resistance of the mixture to flow. When in addition
these particles interact at a distance through the liquid, as happens with colloidal
particles (Sect. 1.5), the resulting mixture can have more complex mechanical prop-
erties somewhere between those of a solid and those of a simple liquid. Analogous
phenomena are found with emulsions (Sect. 1.7) and foams (Sect. 1.8). For their
part, polymers constitute a distinct category owing to the fact that their chains may
(i) occupy very large apparent volumes in comparison to their effective volume,
(ii) stretch out like springs, and (iii) become entangled. These various phenomena
lead once again to novel types of mechanical behaviour: effective elastic behaviour
under certain conditions and liquid behaviour under others (Sect. 1.6). Finally, gran-
ular materials (Sect. 1.9), concentrated assemblages of solid grains able to come into
direct contact with one another, also form their own class. They may remain blocked
like solids or flow like liquids depending on the circumstances, but in the latter case,
in contrast to liquids, the pressure plays a critical role.
1.2 Solids
We are surrounded by solid materials which ensure the success, stability, and
durability of everything we do: the ground we stand on is solid, we build solid
houses and buildings (see Fig. 1.1a), and we surround ourselves with solid pieces
of furniture on which we set a whole range of objects that we would like to
remain in place. We also use all kinds of vehicles which carry us around without
1.2 Solids 3
Fig. 1.1 Concrete on different length scales. a The bridge across to the Ile de Ré, off the French
Atlantic coast. Built with concrete and comprising 28 piers, this is the longest bridge in France,
measuring some 2926 m. This curved structure joins the island to the continent. b Centimetre scale
concrete structure. Credit G. Grampeix and H. Delahousse (IFSTTAR). c Structure of interstitial
cement paste (cryo-SEM image) on a scale of one tenth of a millimetre. Credit J. Hot and C. Castella
(IFSTTAR)
deforming. The human body itself is made up of solid elements, some of which
deform only slightly (the bones), while others are highly deformable (the skin). To
grasp the importance of the solidity of all these objects on our everyday lives, we need
only try to imagine the same world but made up of liquid objects. We would gradually
sink into the ground, our houses would deform, our beds would flatten out underneath
us, and all objects would begin to flow away shortly after their manufacture.
Naturally, all these materials remain solid only over a certain range of conditions
that varies from one object to another: under a large enough deformation, they will
break. Looking at them more closely, all solid materials are in fact elastic when
subjected to small deformations. Some objects, referred to as ductile, can be signif-
icantly deformed before reaching breaking point, while others, described as brittle,
snap cleanly beyond some very small deformation (see Fig. 1.2). Finally, the com-
mon feature of all these solid objects is that they cannot be deformed indefinitely
without eventually breaking.
While all these objects may be solid, their composition and structure can vary enor-
mously. Indeed, what common features are shared by a rock made up of agglomerated
crystals, a steel bar, a rubber made from vulcanised entangled polymers, a wooden
4 1 Introduction
Fig. 1.2 Consequences (right) of applying a bending force to an initially solid material at rest (left)
beyond the elastic regime. a A ductile nail. b A fragile biscuit
plank made from plant matter, or a concrete wall comprising a stack of grains of many
different sizes bonded together in a water–cement matrix (see Fig. 1.1b)? In fact, in
all these materials there is a length scale on which the constitutive elements form a
strongly interacting structure. It is often on the atomic or molecular scale that one
observes these strong bonds guaranteeing the solid behaviour of the object as a whole
(see Sect. 2.2). On this scale, one has in particular crystal structures, in which the
atoms or molecules are arranged in an ordered way. Such structures generally char-
acterise pure substances or simple alloys, comprising molecules of comparable size.
In glasses, the material may have an amorphous structure without specific ordering.
Many solids encountered in everyday life are nevertheless much more complex,
composed of elements of different sizes such as cells, grains, clay particles, fibres, and
so on, immersed in crystalline or amorphous phases of varying degrees of complexity
(see Fig. 1.1c) or simply compacted together. This is the case when a rock cools
from a liquid magma. Here there is a process of fractional crystallisation in which
crystals first form from one kind of species and then later on from others, so that
the final structure is an ensemble of different crystals dispersed or concentrated in
a glassy matrix made from a species that has solidified without crystallising. Most
materials used for building, like plasters, concretes, mortars, terracotta or mud (baked
or unbaked earth), etc., are also made up from grains of various sizes coated in a thin
matrix (hydrated cement, clay paste). The effect of these grains is often to increase
the mechanical strength of the material in its solid state.
In all these structures, the various elements are trapped between several near
neighbours with which they interact through mutual interaction forces. At rest, each
element is in a position of equilibrium with respect to the forces exerted on it, just
as though it were at the bottom of a potential energy well. What happens when
a force is applied to this type of structure? A macroscopic force applied to the
solid sample as a whole will induce local forces between the elements in addition
1.2 Solids 5
to the existing interaction forces. These elements will then shift slightly into new
equilibrium positions associated with the newly established set of forces. They thus
appear to climb to new positions relative to their initial potential wells. For small
applied forces, the induced displacements will be correspondingly small and the
system remains in a ‘linear’ regime in the sense that the displacements are roughly
proportional to the force. Furthermore, if the applied force is removed, the elements
will naturally return to their initial equilibrium positions. Put another way, they will
fall back to the bottom of their individual potential wells. The macroscopic behaviour
of the material will follow suit: when subjected to small enough external forces, it
will be linearly elastic. For simple solid structures, the mechanical resistance to
deformation can be calculated directly in this regime if one knows the interaction
forces between its constituent atoms at the local level (see Sect. 2.5).
Beyond some critical level of deformation, certain elements may leave their
potential well, whereupon the initial arrangement will be permanently destroyed.
This irreversible breakage happens because, given the complexity of the distrib-
ution of interactions, it would be impossible for the system to recover its origi-
nal configuration through deformation in the opposite direction. In certain cases,
the structure may separate into several parts, and one speaks then of brittle frac-
ture (see Fig. 1.2b). In other cases, the structure may break only at several points
(dislocations) while the material maintains its apparent integrity. One then speaks of
ductile fracture (see Fig. 1.2c) and the material is said to undergo plastic defor-
mations, these being associated with irreversible changes in the initial structure
(see Sect. 2.5.6).
So the fundamental property of solids is that they cannot survive unlimited
deformation without breaking or at least permanently losing their initial charac-
teristics. This sets them apart from fluids, which can be infinitely deformed without
losing any of their original mechanical characteristics.
1.3 Liquids
Our lives are also dependent on fluids. We have to breathe air and drink water, and
various liquids flow in our bodies (blood, synovial fluid). Unlike solids, these fluids
(gases, simple liquids) are infinitely deformable under any force, no matter how small.
This is what we observe when a glass of water is spilt on a perfectly plane surface.
The water spreads widely until it forms a very thin film over a large area. In this
case, the force applied to the fluid is the force of gravity, which tends to reduce
the liquid level as long as no obstacles are encountered. When the thickness of the
liquid film is small enough, surface interaction phenomena (which determine wetting
properties) finally prevent any further spreading. Another key difference between liq-
uids and solids lies in the fact that, whatever deformations are imposed on the liquid,
its mechanical properties are never altered, in contrast to what happens with solids,
which are likely to fracture beyond a certain critical deformation. Finally, liquid
flow can occur in very different ways, from mass flow in a canal, river, or duct,
6 1 Introduction
through sprays of droplets in pesticide applications, to vapour condensation produc-

ing droplets on a window. These examples show that a liquid can be indefinitely
deformed or separated into pieces which reconstitute a liquid mass with the same
properties when brought back together again.
The properties of liquids just described are extremely useful to us. They allow us
to take in various substances, including water, which first adopts exactly the shape
of the glass, then the shape of our mouth, before entering the digestive system and
spreading throughout the body. Likewise for the blood, which must circulate in our
veins and arteries, irrigating the tiniest of blood vessels. We also encounter liquids
in our everyday lives, in the form of detergents, petrol, oils, and others. Thanks to
their specific features described above, all these liquids can be spread over very wide
areas or vigorously stirred up without altering their natural qualities.
These mechanical properties arise from two main characteristics of the liquid
structure (see Sect. 2.2):
• As in a solid, the atoms or molecules of a liquid are held in contact with one another
by van der Waals forces. However, there is a permanent spontaneous agitation of
thermal origin which means that two neighbouring elements may suddenly be
caused to move apart.
• The elements are arranged in a disordered way, but the average properties of
this arrangement remain constant whatever deformations the structure may have
undergone in the past.
Even when it is at rest, this agitation occurs throughout the liquid, implying that, on
the atomic or molecular scale, the arrangement of the elements changes all the time
in a natural way. Locally, there are as many relative motions in one direction as in any
other, whence the absence of any macroscopic motion. However, when a shear force
is applied to the liquid, the system is caught off balance. It becomes easier for the
elements to leave their potential wells in the direction of the external force. The liquid
thus deforms, and since its properties remain the same during the deformation, the
effect of maintaining this force is that the liquid keeps on deforming at the same
rate. Put another way, it flows. In this motion, the shear deformation is the ratio of
the relative displacement of two fluid layers to the distance between them. The shear
force applied per unit area is called the shear stress (see Fig. 1.3a). This stress reflects
the fluid’s (viscous) resistance to flow. The basic rheological characteristic of a fluid
is its viscosity, which is the ratio of the shear stress and the shear rate (see Chap. 2
and Appendix A). The point about simple liquids like oil, alcohol, water, honey,
mercury at room temperature, etc., is that their resistance to shear, in other words the
shear stress, is simply proportional to the shear rate and does not depend on the flow
history: these are the so-called Newtonian fluids (see Chap. 2 and Appendix A).
For a Newtonian fluid, the viscosity is its principal mechanical characteristic.
Roughly speaking, it increases with the length of the molecules making up the liquid,
since longer molecules will find it harder to slide over one another, just as it is easier to
produce a flow with a mixture of beads than with a packet of entangled needles. Note,
however, that the viscosity of a liquid decreases when its temperature is increased
1.3 Liquids 7
(b)
(a)
Fig. 1.3 In a simple shear deformation, the liquid layers are considered to slide parallel to one
another (a). The tool of choice for measuring the viscosity of a liquid is the rotational or shear
rheometer, which may involve various measurement configurations such as parallel disks (b). After
moving together and slightly squeezing the sample, these rotate relative to one another about the
same axis of rotation
Fig. 1.4 Volcanic lava flow. The upper surface of the flow cools down quickly by radiation. This
increases the viscosity of those regions close to the surface and tends to form a crust (black) which
breaks as deformation proceeds. Those regions located below the crust (yellow) remain very hot
and highly fluid
(see Fig. 1.4) because the molecules are more agitated and can more easily exit their
potential wells to move relative to one another (see Sect. 2.4.5).
The viscosity has a determining impact on flow characteristics. For the same
volume of fluid subject to the same force, the resulting flow velocity will decrease
with increasing viscosity (see Fig. 1.5). As the thickness of a liquid layer or filament
is reduced, capillary or surface tension effects will play an ever greater role. These
are produced by the liquid–air interfaces. For example, a liquid filament will tend to
8 1 Introduction
Fig. 1.5 Flows of liquids with increasing viscosities under gravity. a Water on a plant. b Alcohol
poured into a glass. c Oil. d Honey
separate out very quickly into droplets (see Fig. 1.5a) to minimise the area of these
interfaces, but this effect may disappear if the liquid is viscous enough (see Fig. 1.5c
and d).
The basic tool for measuring the viscosity of a fluid is the rheometer (see Fig. 1.3b).
Such an instrument works in the following way. A sample of the liquid is placed
between two solid surfaces, one of which is held fixed while the other is displaced
in a direction parallel to the first. Different surface configurations are possible, such
as coaxial cylinders or parallel disks, where one of the cylinders or one of the disks
rotates about its axis while the other remains fixed. The flow induced in the fluid
(simple shear) can be conceived of as a relative motion of the fluid layers, one
parallel to the other (see Fig. 1.3a). By using a relative rotational motion, very large
deformations can be induced without the sample leaving the imposed geometry. The
value of the applied shear stress is deduced by measuring the torque imposed on the
device to induce a given flow velocity (see Chap. 8).
1.3 Liquids 9
(a)
(b)
Fig. 1.6 a Laminar and b Turbulent regimes: flow of a Newtonian liquid in a tube. A dye has
been injected into the liquid without modifying its properties. At low flow velocities (a), the liquid
elements remain in layers which slide over one another. The dye thus moves with the layer in which
it has been injected and hence generates a straight line. For faster flows (b), the elements making up
a layer diffuse into the other layers, taking the dye with them, so that it gradually spreads throughout
the tube
Finally, in this context, we can measure the friction between two liquid layers
sliding one relative to the other at a given velocity. The result thus corresponds in
principle to an intrinsic property of the material. If we impose a flow under different
conditions but leading once again to a relative slipping of the liquid layers, the local
stresses will be exactly the same as those measured in a rheometer for the same
shear rate.
Note that the above description of liquid flow is only valid for laminar flows.
In this type of flow, liquid elements are assumed to remain in planes that slip over
one another (see Fig. 1.6a). However, this would not be the case for faster flows,
which are said to be turbulent (see Sect. A.8 in Appendix). In this situation, the
inertial forces of the fluid elements are greater than the viscous friction obtained
under the assumption of laminar flow. These inertial forces induce complex motions
in all directions around the average motion, whereupon the fluid elements tends to
diffuse (see Fig. 1.6b). As a consequence, a significantly greater force is required to
induce turbulent flow than to induce laminar flow. But the main difficulty in modelling
turbulence is that this phenomenon is not intrinsic to the fluid, i.e., the characteristics
of turbulent phenomena are not simply related to the local velocity field, as in the
laminar case, but depend on macroscopic features of the flow. Turbulence cannot be
predicted at the local level without taking into account the boundary conditions of
the flow. In the remainder of this book, we shall only consider laminar flows, where
turbulence can be neglected. Indeed, rheophysics is concerned only with this regime.
Despite the apparent simplicity of the structure of these liquids, we still do not
have any model capable of predicting their velocity in all flow regimes solely from
knowledge of local interactions. This suggests that the analysis would be at least as
difficult in situations involving fluids with more complex interactions and structure.
However, predictive rheophysical methods can be developed for certain structures
10 1 Introduction
with particular types of organisation. It is also possible to predict the behaviour of

certain materials made by adding elements to a liquid of known viscosity.
Many materials are obtained by mixing into a simple liquid various elements
which differ significantly from the atoms and molecules of this liquid, e.g., polymers,
cells, grains, bubbles, droplets, etc. The behaviour of these elements differs from the
behaviour of the liquid molecules for various reasons, such as their size, shape, ability
to deform, or the interactions between them. Quite generally, whenever these objects
are sufficiently far away from one another, the behaviour of the mixture is dominated
by the behaviour of the liquid matrix. The fluid will behave qualitatively rather like
the liquid, with a flow resistance that depends on the type and amount of suspended
objects. When the objects come closer together and significant interactions become
possible between them, the behaviour of the mixture will be very different from the
behaviour of the liquid alone.
1.4 Suspensions
The simplest case of objects placed in a liquid is a simple dispersion of solid elements.
This is called a suspension. This type of material is what we obtain when we make
a soup, which usually consists of a dispersion of small pieces of vegetable, fish, or
meat in water, or when we prepare a hot chocolate by dispersing a mixture of cocoa
powder and sugar in milk or water. In civil engineering, we also employ all kinds
of suspensions (ceramics, concretes, mortars, cements, asphalt surfaces, paints, and
others). Indeed, the addition of solid particles provides a cheap way to strengthen
the final structure of the material, after setting or drying. Nature commonly uses
suspensions as a way to transport sediments from one point to another. Floodwaters
in rivers and streams pick up solid particles which they then carry over great distances
in the case of the smaller particles.
1.4.1 Phase Separation
In practice, this kind of mixture rarely remains homogeneous for very long. Indeed,
since the solid particles usually have different densities to the carrier liquid, the
particles will either fall through the liquid, in which case one speaks of sedimentation
(see Fig. 1.7), or rise up to the surface, in which case one speaks of creaming. The
time required for this phase separation increases in inverse proportion to the square
of the particle sizes and proportionally to the viscosity of the liquid (see Sect. 3.2.6).
Hence, if the particles are very small or the liquid very viscous, the phase separation
may not have time to occur to any significant extent during a given experiment. In
some cases, the particles have densities very close to the density of the liquid and
this will considerably slow down any sedimentation or creaming.
1.4 Suspensions 11
Fig. 1.7 Sedimentation of mud obtained by mixing a little clay with water. a Initial state after
uniform dispersion. b Final state after resting for several hours. The particles gradually fall to the
bottom of the glass to produce a concentrated deposit
1.4.2 Effect of Particles on Behaviour of the Mixture
Here we are specifically concerned with the effects of adding particles to

homogeneous mixtures in the case where phase separation can be neglected.
For example, one might pour a little icing sugar into oil. The sugar will not dis-
solve in the oil and so will remain in the form of solid grains. The result is then
stirred to obtain a homogeneous dispersion of sugar throughout the liquid. The mix-
ture appears to behave like the pure oil. As we continue to add sugar, the behaviour
of the mixture does not change in any fundamental way, and in particular, it still
flows easily. In fact, accurate measurement reveals that the viscosity increases very
gradually with the fraction of added sugar. However, a sudden change occurs for a
particular value of the added fraction. The mixture suddenly becomes very viscous
or even paste-like. It seems that a percolation threshold has been reached, beyond
which the particles begin to get in each other’s way.
The effects observed in this oil–sugar mixture can be found in many other
materials. That is, the effect of incorporating large solid particles on the viscos-
ity of the mixture often depends only on the added volume fraction and in no way
on the physicochemical characteristics of the liquid and solid phases. These effects
have a hydrodynamic origin. The flow of the mixture depends on the flow of the
liquid between the solid particles. The latter tend to complicate the characteristics
of the liquid flow. When the suspension is sheared between two solid planes, the
liquid cannot flow around the particles in the form of parallel planes in relative
sliding motion. However, whatever stress is imposed, it always causes a flow of the
mixture. When the organisation of the mixture, i.e., the spatial distribution of the
particles, does not change, we know that increasing the speed only requires greater
stresses since they are simply proportional to the local velocity. So as long as the par-
ticle distribution is the same in all directions and remains constant, the homogeneous
12 1 Introduction
mixture of a Newtonian liquid and particles will itself constitute a Newtonian mixture
(see Sect. 3.4).
The main effect of including solid particles is that they get in the way of the
flow, simply by the fact that they replace liquid by a solid phase that is itself unable
to flow. It follows that it is also more difficult to produce a flow in a suspension
than in the liquid alone. The viscosity increases with the amount of particles in
suspension. This quantity is described using the ratio between their volume and the
total volume of the sample, which is called the volume concentration. This effect
of the concentration on the viscosity is explained by the fact that, in order to obtain
the same apparent shear in a suspension, i.e., the shear calculated from the relative
motions of the material layer boundaries, the interstitial liquid must be sheared with
greater force because the suspended solid phase does not contribute to the local
relative motions. The increase in viscosity with concentration is nevertheless very
slight at low concentrations. The viscosity of the mixture is only three times that of
the liquid alone when the particles already occupy close to 1/3 of the sample volume.
This explains why there is no obvious change in the viscosity of the mixture when
we begin to add sugar to oil.
1.4.3 Other Effects
1.4.3.1 Size Effects
Chocolate is a suspension of particles of sugar, cocoa, and milk (with a solid fraction
of about 45 %) in cocoa butter, which is a fat. The latter liquefies between 30 and
35 ◦ C, which explains why chocolate melts in the mouth. At a given temperature,
the viscosity of the mixture increases, as expected, with the concentration of solid
particles. An increased viscosity is also observed when the size of the suspended
particles is reduced, e.g., by crushing them, while keeping the solid fraction at the
same level. This effect is often explained as being due to the increased solid–liquid
interactions associated with the greater interfacial area between the two phases, which
in turn results from the smaller size of the particles. This is a fallacy, however. While
the reduced size of the particles does indeed increase the area of contact between solid
and liquid at given concentration, as happens, for example, if a spherical particle is
cut into two halves, that would have no direct impact on the viscosity of the mixture.
In the context of the present discussion, the interactions between the two phases play
no role a priori. Only the flow characteristics of the interstitial liquid influence the
viscosity of the system. Two suspensions containing the same volume concentration
of particles of the same shapes with similar relative size distributions (granulometry)
clustered around two different mean values will have identical apparent viscosities.
This is a really fundamental result: the viscosity of a suspension of identical particle
types depends on the solid volume fraction, and not on the size of the particles. To
understand this, we need only reflect that the two suspensions described above appear
1.4 Suspensions 13
Fig. 1.8 a Deposit from a mudslide originating in a mountain stream bed. b Internal structure of
such a mudslide, containing a broad granulometric spread, ranging from clay particles with size
less than the micron to blocks several metres across
perfectly identical, and hence with identical viscosity, if each is observed at a length
scale proportional to the average size of the particles it contains.
However, if the particles are not identical, the viscosity of the mixture decreases
as the size distribution broadens. This happens because, for the same solid fraction, it
is easier to distribute the particles in space if this size distribution is broad than if the
particles are all of uniform size. A simple demonstration of this effect is obtained by
considering solid grains in a box: higher concentrations of such grains can be reached
as their size distribution is broadened. Particle concentrations of the order of 90 %
can be achieved in this way, whereas a pile of identical beads in a bucket will have
a concentration of only about 60 %. This is observed in mudslides (see Fig. 1.8), but
also freshly mixed concretes, materials which flow easily despite high concentrations.
Coming back to the effect observed with chocolate, it is more likely that the
increased viscosity observed with reduced particle size results from the existence or
emergence (down to a certain size) of colloidal interactions (see Sect. 1.5), that is,
interactions at a distance between particles, which add to hydrodynamic interactions.
The impact of forces associated with colloidal phenomena does indeed increase as
the particle size is reduced.
1.4.3.2 Orientation Effects
A spherical particle suspended in a simply sheared liquid moves along with the sliding
plane it happens to be in, but it will also rotate about its own axis at a frequency equal
to half the shear rate. This is not really surprising since the different levels within
the sphere are associated with fluid planes sliding at different speeds. In particular,
the two poles along an axis perpendicular to the flow direction are dragged along by
the liquid at different speeds, causing the sphere to rotate about an axis normal to
the plane of observation. In the case of an anisotropic particle, the effect will depend
on its orientation relative to the direction of flow. For example, a rod will be made
14 1 Introduction
Fig. 1.9 a Blood is a suspension of red blood cells (about 45 %), shaped like flattened cushions,
together with white blood cells and platelets, the carrier liquid being plasma, which is a mixture of
water (91 %), proteins, and various other components. b Schematic view showing the flow of such
a suspension, although at a much lower concentration than would be found in reality. Given their
high concentration in the mixture, the red blood cells play an all-important role in the viscosity of
blood. In particular, they tend to aggregate face to face in long stacks to form rouleaux. These are
even more anisotropic and can deform significantly, allowing the blood to enter even very small
capillaries. The blood viscosity can be reduced by inhibiting this aggregation and enhanced by
hindering deformation of the cells
to rotate about its centre of gravity, but its angular velocity will be greater as its
principal axis approaches the direction normal to the flow. Indeed, in this situation,
the difference in velocity between the liquid regions surrounding each end of the
rod is then maximum. On the other hand, the angular velocity of the rod is very
small when its principal axis is almost aligned with the flow direction, because then
the relative velocity of the liquid regions surrounding each end of the rod is very
small. As a consequence, the rod will spend more time close to the flow direction
(see Sect. 3.3). This is because, whenever it moves significantly out of line, it will
soon be realigned by the flow. Arguing in terms of time averages, everything happens
as though the anisotropic particles were trying to line up in this direction. This effect
facilitates the relative motion of the liquid layers, and this in turn will reduce the
viscosity of the suspension as compared with its value for a random distribution of
particle orientations.
On the other hand, if the anisotropic particles do not line up, as might happen if
the volume concentration gets too high, the viscosity of the mixture will be higher
than for a suspension of spheres at the same solid fraction. The viscosity of the
mixture then increases with the aspect ratio of the particles, i.e., the ratio of the large
to the small axis of the particles. This effect plays a key role in blood, for example
(see Fig. 1.9), when red blood cells stack up to form anisotropic aggregates known
as rouleaux.
1.4 Suspensions 15
Fig. 1.10 Behaviour of a mixture of water and corn starch. a The mixture flows like a simple liquid
when disturbed gently, e.g., by slowly tipping over the container. b With a sudden disturbance of
the same mixture, a solid lump forms for a brief instant. c Taking this lump in the hand, the material
flows like a simple liquid once again
1.4.3.3 Configuration Effects
The situation becomes more complicated when the concentration is such that the
particles get very close to one another, because they can then impede one another’s
relative movements and even enter into direct contact. The viscosity of the mixture
then grows very quickly with the concentration, but the behaviour of the mixture is
no longer Newtonian above a certain concentration and new effects arise. When the
particles are small, in a range that seems to be from 10 to 50 µm, a kind of blockage
is sometimes observed in these systems above a critical shear rate. This leads to a
sudden steep increase in the apparent viscosity, an effect known as shear thickening
(see Sect. 3.7). It seems that the particles are blocked into a configuration where
their relative separations are very small, and this induces greater viscous resistance
than if they were dispersed randomly. This phenomenon is encountered, for example,
in aqueous suspensions of corn starch. When such a suspension is stirred gently, it
reacts like a low viscosity fluid (see Fig. 1.10a), but if the sample is subjected to a
more sudden application of force, it reacts like a solid (see Fig. 1.10b).
With larger particles, another phenomenon occurs at concentrations close to the
maximal stacking concentration: the mixture behaves like a granular medium for
low shear rates and like a Newtonian suspension for high shear rates. This transition
is probably related to the fact that, at low velocities, the particles are able to form
a network of direct contacts throughout the whole sample, so that the shear stress
results mainly from friction between the grains and the mixture effectively behaves
as a Coulomb solid (see Sect. 1.9). At high shear rates, however, the particles remain
separated by a liquid layer, whereupon the system appears to behave like a Newtonian
fluid (see Sect. 7.7.2).
1.5 Colloids
The simple presence of particles in a liquid may become negligible compared with
other effects if the particles are small enough, for they may interact together at a
distance through what are known as colloidal interactions within the liquid. Such
16 1 Introduction
colloidal effects can start to play a role when the particle size is less than about
100 µm, but they only become clearly predominant for particle sizes below 1 µm.
Colloidal particles can be found in a whole range of industrial products. Examples are
silica particles in toothpastes, latex beads and pigments in paints, clays in cosmetic
creams, nanoparticles in cement pastes, and so on. The most widespread natural
colloidal particles on the surface of the Earth are clays, which result from the chemical
or mechanical decomposition of rocks. Clays constitute one of the fundamental
components of natural muds and drilling muds, and they also occur as components
in many industrial products, e.g., cosmetics, ceramics, paper making, coatings, etc.
1.5.1 Colloidal Interactions
Colloidal interactions are first and foremost van der Waals forces which tend to cause
particles, like two atoms or two molecules, to aggregate (see Sect. 1.3). When these
forces are predominant, the suspension will not be stable, since the particles will
tend to ‘stick’ together to form a compact lump at the bottom of the liquid container
(see Sect. 5.7). In this case, simply stirring will not redisperse the particles because
the attractive forces in these clusters can be considerable.
To ensure homogeneous dispersion of the particles throughout the liquid,
adequate repulsive forces must be introduced between the colloidal particles. Such
an effect occurs when ions are adsorbed onto the particle surfaces. Repulsive elec-
trostatic forces can then arise between the charged surfaces of neighbouring particles
(see Sect. 5.4). Another solution is to coat the particles with polymers which, grafted
to the particle by one end, form a kind of hair. When two neighbouring particles come
together, it is difficult for these ‘hairs’ to interpenetrate, whence they tend to hold
the particles apart, preventing them from actually sticking together (see Sect. 5.5).
The superposition of these different interactions may lead to a weak aggregation
mechanism in which neighbouring particles end up at a distance corresponding to
the balance point between the above repulsive and attractive forces. The particles
can then be considered to be connected to one another, but such bonds can be easily
broken and reconstituted when the material is in flow. Such techniques provide ways
to stabilise the colloidal dispersion. Here we shall consider only this scenario.
1.5.2 Yield Stress
If there are enough particles, they can constitute a continuous network of connections
throughout the sample. This network leads to completely different behaviour as
compared with the liquid on its own. In this case, at rest, we are dealing with a solid
structure that can only be broken by applying a force greater than some critical value.
The fluid is then said to have a yield stress. As long as the applied stress is less than
this threshold, the material will behave like a solid, with finite deformation. When
1.5 Colloids 17
Fig. 1.11 Practical uses for fluids with a yield stress. a Toothpaste spread on a toothbrush does not
flow. b The surface of fresh concrete placed in formwork can be smoothed and will subsequently
remain smooth. c An adhesive mortar for tiling can be applied to the wall and deformed without
flowing
the applied stress exceeds the threshold, the material will flow like a viscous fluid.
However, the transition is reversible: if the stress is gradually reduced below the yield
stress during flow, the fluid will come to a halt and only start flowing again if a stress
above the threshold is once again imposed.
This is a rather surprising behaviour if we think about the usual division of
materials into two main classes, solids and liquids. These yield stress fluids behave
like solids in some circumstances and liquids in others. Despite this novel behaviour,
these materials are fluids in the sense defined above, because they can be deformed
at will without losing any of their mechanical properties. Such threshold behaviour
is not specific to concentrated colloidal suspensions. It is also found in foams, gels,
and concentrated emulsions. It is useful from the practical point of view because an
arbitrary shape can be given to the material and it will conserve it despite the effects of
gravity. This behaviour can be observed, for example, if we squeeze toothpaste onto
a toothbrush (see Fig. 1.11a), apply paint to the wall, spread concrete (see Fig. 1.11b),
apply a mortar (see Fig. 1.11c), decorate a cake with Chantilly cream, or shape objects
with salt dough.
A remarkable property of these colloidal particles is that a small volume fraction
of such particles in a liquid can yield a mixture with a very high yield stress, owing
to the fact that the particles exert mutual forces at distances that may be of the order
of their own dimensions. This effect is used for many kinds of coating and cosmetic
products. Dispersing a small fraction of laponite particles (a synthetic clay) in paints
18 1 Introduction
Fig. 1.12 Paints (a) are mainly composed (b) of solid particles, pigments, and polymers (latex)
in suspension in water or an organic solvent, with various additives such as surfactants and anti-
foaming agents. The pigments give the paint its final colour, while the polymers are there to ensure
cohesion of the film after drying. A paint can have a rather complex microstructure, as we see
in the example (b) which is a sample of total width about 120 microns of the painting entitled
Moissonneurs by V. Van Gogh, observed by scanning electron microscope (J. Salvant, Centre de
Recherche et de Restauration des Musées de France). The image shows whitish grains and clusters
of white lead and dark spots of calcium carbonate, the whole thing bathed in an oil that is not visible.
The manufacturer must formulate the material with a high enough yield stress to ensure that it soon
ceases to flow once applied to the wall. On the other hand, the yield stress must not be too high,
otherwise it will be difficult to soak up the paint on the brush. In some cases, thixotropic paints are
used (see Sect. 1.5.3). These become more liquid when stirred up, and more viscous when they flow
slowly or remain at standstill. Such paints are easily manipulated but behave well on the wall
(see Fig. 1.12), gels, or creams has the effect of increasing their viscosity and even
giving them a significant yield stress. At the same time, these very small particles,
present in very low concentrations in the mixture, have a negligible impact on the
characteristics of the final dry product.
1.5.3 Thixotropy
Some clay or silica suspensions form this kind of structure and behave like yield
stress fluids, but at the same time they seem to liquefy beyond the yield stress. Put
another way, they can be kept in flow under a smaller stress than the one required
to break the initial structure. Somewhat surprisingly, this effect is reversible. If the
material is left alone, the initial structure will gradually re-establish itself and the
effective yield stress of the material will increase in consequence (see Fig. 1.13).
Such materials are said to be thixotropic.
This strange property can be explained by referring to another physical
phenomenon resulting from the very small size of the colloidal particles. This phe-
nomenon is Brownian motion, named after the botanist R. Brown, one of the first
to observe and describe it for pollen grains. The thermally agitated liquid molecules
enter into collisions with the suspended solid particles. The direction and magnitude
1.5 Colloids 19
Fig. 1.13 In this experiment, a bentonite mud (bentonite is a natural clay, widely used in drilling
muds) was vigorously mixed and placed behind a dam on an inclined plane (a). When the dam
is opened after a short resting time, the mud flows easily into the channel like a low viscosity
fluid (b). If the experiment is repeated, but this time allowing the mud to stand for several minutes
before opening the dam (drying remains negligible), the same volume of mud spreads out to form
a tongue which comes to a halt in the middle of the channel (c). The formation of this kind of
deposit is the typical behaviour of a yield stress fluid. If the mud is now left to stand for 1 hour,
the flow characteristics are more complex. Part of the mud separates off and slides quickly along
the channel (d), just like in a mudslide. In this case, thixotropic effects dominate. The material
now has a relatively high yield stress, but it liquefies when a large enough stress is applied. This
is the phenomenon occurring in the thin layer between the two blocks. If the experiment is now
repeated but leaving the mud for several hours before opening the dam, it merely deforms slightly
or fractures, but there is no flow (d). In this case, the mud has significantly restructured itself to
the point where its effective yield stress is too high for it to flow along the channel under its own
weight. If the mud is stirred up again, it returns to its original liquid state and we obtain the same
results as observed previously. Credit H. Chanson
of the forces exerted on a given particle as a result of these collisions are perfectly
unpredictable if taken individually. However, the average force over all possible col-
lisions at a given time is related to the temperature of the system, which is itself
associated with the thermal agitation of the liquid molecules. This instantaneous
force induces the particle to move through the liquid, and the motion will be all
the greater as the particle is lighter (see Sect. 5.2). Finally, because the total force
varies erratically over time, the particles will move in an equally erratic way, thereby
diffusing to differing degrees through the liquid. This agitation increases as the size
of the particles diminishes.
20 1 Introduction
Fig. 1.14 Evolution of a layer of latex beads, of diameter about 1 µm, dispersed on the surface of
an aqueous solution when salt is added. The salt reduces repulsive electrostatic forces. From left
to right and from top to bottom, the images show the structure at different times after adding the
salt: 15, 75, 105, 135 min. Subjected to Brownian agitation and strong attractive interactions, the
particles come together and bind to form ever bigger aggregates. The final structure is an aggregate
extending throughout the whole sample. This must first be broken up before flow can be induced.
Reprinted with permission from [1]. Copyright (1993) by the American Physical Society
The Brownian motion coupled with the interparticle interaction forces gradually
strengthens the network of interactions that determines the yield stress of the material.
The particles explore and eventually reach more and more stable positions, that is,
positions associated with deeper potential wells, forming an ever more extensive
interaction network (see Fig. 1.14). In the end, the structure gets stronger and stronger
(see Sect. 5.9). In flow, this network is partly broken, but it immediately begins to
re-establish when the material is at rest.
Thixotropic behaviour is useful in practice when we require a fluid that is very
viscous at rest but offers little resistance to flow. This is particularly true of paints
(see Fig. 1.12). The user prefers a paint that easily soaks into the brush and is equally
easy to spread on the wall, but which nevertheless immediately becomes viscous
once spread out, i.e., when at rest, so that it does not run under the effects of gravity.
The so-called self-placing concretes are formulated with similar aims in mind. These
materials have low apparent viscosity when they flow, so they are easy to pump and
distribute themselves efficiently throughout the formwork which gives them their
final shape. As soon as flow ceases, they quickly restructure, whence large particles
of sand or gravel do not have time to sediment out. Drilling muds provide a further
example of thixotropic materials. These are used to facilitate boring and extraction
of rock. They must therefore be able to flow easily when injected into the borehole,
1.5 Colloids 21
where they serve to lubricate the drill head, but they must also be able to hold pieces
of rock in suspension when the system is at standstill, and this is enabled by the fast
increase in the yield stress at rest.
1.6 Polymers
The twentieth century brought with it the age of plastic. So-called plastics contain
polymers, that is, very large molecules made by joining up many copies of the same
small building block. In our daily lives, we encounter them all the time. For example,
they are used in packaging, toys, certain textile fibres, many vehicle parts, and so on.
Most of the time we use polymers in their solid form, but they are liquid when they
are synthesised, at very high temperatures, which makes it possible to shape them as
required.
1.6.1 Properties of Polymer Chains
To understand the properties of polymers, we must first examine the properties of

chains. A polymer chain is a rather special kind of entity, made by juxtaposing
thousands or even millions of identical small molecules which join together through
one of their carbon atoms. One might expect such a chain to occur in the form of
a long straight fibre whose length would be proportional to the number of atoms in
the chain. The reality is much more complex. Three connected molecules are not
generally aligned, and furthermore, their relative orientation is not unique. The first
implication is that a chain comprising a given number of atoms may assume many
different forms in space, from a long strand, through coils of various tortuous shapes,
to globules (see the example in Fig. 1.15a and also Sect. 4.2). The special properties
of polymer-based materials, and in particular their elasticity or plasticity, come from
the fact that a chain in a given configuration can evolve to some other configuration,
depending on the stresses it is subjected to. Put another way, polymer chains are
deformable (see Fig. 1.15).
It is worth thinking for a moment about the deformability of the chains. Consider a
given chain and imagine that we fix the apparent length, i.e., the distance between its
two ends. Several configurations may yield this apparent length. If we now increase
the distance between the two ends, we will reduce the number of possible configura-
tions of the chain. For example, when the chain is stretched as far as possible, there
will be only one available configuration, corresponding to this maximum extension.
It follows that the entropy of the chain, i.e., the number of microscopic states that
would yield this macroscopic state, is reduced when the chain is stretched, since
the number of possible arrangements or configurations of the various atoms is also
reduced. The laws of thermodynamics teach us that energy must be supplied in order
to lower the entropy of a system. This means that a certain force must be applied to
22 1 Introduction
Fig. 1.15 a Representation of a polypropylene molecule in a barely stretched conformation.

b Deformation of a ‘plastic’ (polyethylene) bottle obtained by immersing it in boiling water. The
polymer chains are deformed and keep the new form after cooling. c Reversible deformation of an
elastomer (a ‘rubber’ balloon), which resumes its initial position once the force is removed
the chain to stretch it out. It turns out that, for small enough deformations, this force
is simply proportional to the extension, and when the force is removed, the chain
resumes its initial shape. What we are saying is that the behaviour of the chain is
essentially elastic (see Sect. 4.2.4).
1.6.2 Polymers in Solution
In various applications or when manufacturing certain materials, polymers are

suspended in a simple liquid. The effect the polymers can have on the behaviour
of the mixture depends in particular on the shape assumed by the polymer chains
when they are immersed in the liquid. When they have a certain ‘affinity’ with the
liquid, they tend to occupy a rather large apparent volume. This affinity relates to the
interaction energy between the chain components and the molecules of the liquid.
On the other hand, when this affinity is low, they tend to coil up (see Sect. 4.3).
To a first approximation, the chains in solution can be considered to occupy a
volume contained within the spherical envelopes that encompass each of them. The
polymer density within such a volume is usually very low, but everything happens
as though the fraction of liquid contained within this region is simply trapped by the
chain. Under these conditions and regarding the viscosity, a polymer solution can be
treated as a suspension containing solid particles with volume equal to the apparent
volume of the polymer chains. We then observe that the viscosity of the mixture
increases much more quickly with the polymer concentration than if the polymer
chains were really small compact solid objects (see Sect. 4.6). This behaviour is
widely exploited in practice. Liquids can be ‘thickened’ by adding a very small
polymer fraction. For this reason, it has become commonplace to add polymers to all
kinds of products, such as shampoos or washing-up liquids, in order to significantly
increase their viscosity without greatly altering their overall composition. Note also
that polymer solutions are shear thinning, i.e., their apparent viscosity decreases with
1.6 Polymers 23
the applied shear rate (see Sect. 4.6). This is due to the straightening and alignment
of the chains, phenomena which have ever greater impact as the stress is increased.
When the polymer concentration reaches the point where the chains begin to
get in each other’s way or even become entangled, the viscosity evolves in a much
more complex way. Eventually, at very high concentrations, we come close to a pure
polymer. In this case, the structure starts to look somewhat like a plate of spaghetti
(see Sect. 4.6.3). The flow of such a system can no longer occur by simple displace-
ment of parallel layers, but involves subtle displacements of the chains relative to
one another, a phenomenon we shall return to shortly.
1.6.3 Viscoelasticity
The most spectacular property of polymers is their viscoelasticity, that is, a partly
elastic and partly viscous mechanical behaviour. This property can be observed in
dilute polymer solutions, but viscoelasticity is particularly evident in concentrated
solutions and polymer melts. In practice, when a stress is imposed on such a material,
it will behave initially like an elastic solid, over a rather short period of time, but if
the stress is maintained, it will end up flowing indefinitely like a liquid.
The practical consequences of this behaviour can be observed in a mixture of
silicone oil and boric acid, a substance commercialised as a toy under the trade name
of Silly Putty. When a ball of this is dropped on the floor, it will bounce up as high as
a football or a rubber ball (see Fig. 1.16a, b). However, if we apply a longer lasting
but lesser force, for example by setting the ball on the floor and leaving it to readjust
itself under the effects of gravity, we find that it will slowly deform and spread out
into a kind of puddle. We can also allow it to be drawn out under its own weight, in
which case it can extend significantly without breaking (see Fig. 1.16c, d). Finally,
if we strike it with a hammer or if we subject it to a sudden traction force, it can
literally shatter (see Fig. 1.16e).
What happens inside this material in the different steps of the experiment? When
a stress is applied to the material, the whole chain network first deforms. As for the
chains themselves, the behaviour of the network as a whole is essentially elastic.
Indeed, over short time lapses, the chains do not have time to slide relative to one
another as they must in a flow. They stretch or squash up under the action of the
applied mechanical stress, while maintaining their initial organisation, i.e., keeping
the same neighbours. If the stress is applied too quickly, it does not even leave time
for the chains to deform and the structure may simply fracture. On the other hand, if
the stress is applied for long enough, there will be time for the network to rearrange
itself through various relative motions of the chains. Under the action of such a stress,
the structure will thus deform slowly but surely like a liquid.
This description hints at a mechanism that does not have any obvious explanation.
For how would we expect a mass of entangled molecules to flow? In fact, the relative
motions of the chains are made possible by thermal agitation in the system. Each
chain evolves in an environment made up of its neighbouring chains and which is
24 1 Introduction
Fig. 1.16 Response of Silly Putty to different types of stress: a and b bouncing on the ground,
c and d gradual stretching, e sudden stretching leading to breakage
itself evolving. In a sense, it is as though each chain were held in a kind of tube,
yet free to move along that tube, whose walls change over time under the effects
of thermal agitation in the system. This means that the chain can move on a time
scale that decreases as the thermal agitation increases, i.e., as the temperature gets
higher. The whole thing can be likened to a basketful of writhing snakes. Each one
1.6 Polymers 25
can move under its own free will, independently of the others, but along a path
that is nevertheless constrained by the presence of its neighbours. As a result of the
little movements of each relative to the others, it can eventually cover a significant
distance. Pursuing this analogy, it is natural enough to consider that the chain moves
by a process of reptation, as suggested by P.G. de Gennes.
In this context, a key feature of the material is its relaxation time, which corre-
sponds roughly to the time required for a chain to move a distance equal to its own
length. The type of response displayed by the material depends on the ratio between
the time the force is applied from rest to the relaxation time. If the latter is shorter, the
material will have time to reach its liquid regime because the chains will have been
able to move a certain distance through the network, allowing it to evolve signifi-
cantly. But if the relaxation time is longer than the duration of the applied stress, the
network will not have time to change noticeably and the chains will have to deform.
Given the effect of temperature on the thermal agitation which facilitates reptation,
an increase in temperature will tend to reduce the relaxation time of the system. As a
result, regarding the apparent properties of the material, a rise in temperature will be
equivalent to an increased duration of the applied force. There is thus an equivalence
between the time effects and the inverse temperature (see Sect. 4.7).
The structure can also be rigidified by linking the chains together at several points
using sulfur atoms in a process known as vulcanisation. As long as these bonds
or the chains themselves are not broken, the material remains solid, but it will be
deformable because its very structure is made up of deformable chains (see Fig. 1.15c
and Sect. 4.5). Natural rubber is a liquid polymer that has been solidified by this
process. Naturally, if too great a force is applied, associated with too large a defor-
mation, the structure will permanently come apart by breaking either the chains
themselves or the crosslinks.
1.6.4 Other Properties of Polymers
Other characteristic properties of polymer materials are observed in flow.

A Newtonian liquid has the same viscosity whatever deformation it undergoes.
In contrast, the viscosity of polymers in an elongational flow, which corresponds
to stretching a cylinder along its axis, can be much higher than their viscosity
under simple shear. This can lead to surprising effects since the material may seem
liquid under shear, but rigid when stretched. The happens because resistance to shear
arises mainly through the chains slipping one over the other, while elongational flow
involves not only relative slipping of the chains, but also an extension of these chains
in the direction of elongation, and the latter may require significantly greater forces
than the slipping.
Simple shear gives rise to another novel phenomenon, once again due to the elastic
properties of the chains, which tend to stretch out and orient themselves in the flow
direction. The result is that a polymer jet emerging from a duct will swell up, i.e., its
diameter will quickly grow bigger than the diameter of the duct. Indeed, in the duct,
26 1 Introduction
the fluid is being sheared and normal stresses must be applied, that is, perpendicular
to the flow direction, in order to maintain the relative sliding motion of the stretched
chains. At the duct outlet, the jet comes into contact with the air and these normal
stresses disappear, whence the chains will tend to contract, thereby causing the jet to
swell outward. The effects of developing normal stresses can also be observed when
a cylinder is rotated in a tank of viscoelastic polymer. For a Newtonian fluid, the
centrifugal force would tend to lower the level of the free surface near the rotating
cylinder, but for a viscoelastic polymer, the fluid will tend to climb up the cylinder.
This is the Weissenberg effect. In this scenario, the normal stresses increase with the
shear rate, which itself increases as one approaches the inner cylinder. The vertical
force on the fluid is thus maximal along the wall of the rotating cylinder, and this
induces the vertical motion of the fluid.
1.7 Emulsions
In many practical situations, one would like to spread a liquid over a solid surface or
have it penetrate gradually into some porous medium. This is the case, for example,
with medicines administered by injection. If this liquid has low viscosity, it will not
be so easily manipulated, spread over a surface, or given some arbitrary shape. One
solution is to use a process which places the liquid in a mixture that is globally much
more viscous, that is, an emulsion. The idea is to disperse the liquid in the form of
droplets within another liquid. The mixture obtained in this way is a suspension of
droplets whose viscosity will generally increase with the concentration, and which is
in fact a yield stress fluid for sufficiently high droplet concentrations. This technique
is used in cosmetics (moisturising creams), in pharmaceutics (balms, lotions), in civil
engineering (drilling fluids), in the food industry (mayonnaise, vinaigrette, butter),
and in explosives (explosive emulsions).
In fact, it is not at all obvious how to prepare an emulsion. Dividing a droplet into
two smaller droplets increases the total area of their interface with the surrounding
liquid. Making droplets out of a compact volume of liquid thus involves a very
significant increase in the interfacial area between the two liquids, and this in turn
requires a supply of energy to the system, since a very large number of molecules of
each phase must be brought into contact. Furthermore, when an emulsion is prepared
by vigorous mixing, the energy needed to divide the droplets into smaller ones is
transmitted via the viscous forces within the surrounding liquid. At a given shear rate,
a critical droplet size is reached after a certain time, and this size will depend on the
shear rate and also on the ratio of the viscosities of the liquid and the droplets. Beyond
this size, the droplets can no longer be divided up under the action of viscous forces
alone (see Sect. 6.3). The main problem with this kind of mixture is its instability,
since any two droplets that meet up will rapidly merge to form a larger droplet, thereby
minimising the surface energy. This process is known as coalescence (see Sect. 6.4).
Stabilising products known as surfactants must therefore be added to hinder this
phenomenon. These position themselves on the interfaces and thus prevent the liquids
1.7 Emulsions 27
Fig. 1.17 a A typical emulsion: mayonnaise. b Emulsion of oil (dodecane) in water at a moderate
concentration (75 %). The droplets are still spherical. c The same but at a higher concentration
(85 %). The droplets are now squashed up against one another. Credit L. Ducloué
within two droplets from coalescing. In reality, there will always be a residual effect
in which liquid from the largest droplets will diffuse very slowly to the smallest, but
if the surfactants are doing their job correctly, this process will be extremely slow.
Mayonnaise is an emulsion (see Fig. 1.17a) produced by dispersing oil droplets
in an aqueous solution of egg yolk comprising 50 % water, and mustard. If the basic
ingredients (water and oil) are whipped up in a bowl alone, large oil droplets can be
produced in a matter of minutes, but they will immediately join back together again,
floating up to the surface of the mixture. The egg yolk and mustard, containing
lecithin in particular, play the role of surfactant here, stabilising the emulsion. The
consistency of the final product depends on the volume ratio of oil and aqueous
solution. A small proportion of oil yields a very fluid mixture with only slightly
different viscosity from water. This is because the oil droplets are relatively far apart.
In order to increase the viscosity to any extent, the volume fraction of oil in the total
volume must be of the order of 80 %. In this case, the droplets are close enough
to one another to ensure that they must deform in order to fit into the available
volume of the sample (see Fig. 1.17b, c). Each droplet is held between its neighbours
as though caged, so the arrangement of droplets ends up forming a solid structure
28 1 Introduction
which can only be broken by applying a sufficient stress to force a certain number
of droplets out of these cages. By maintaining this stress, the process will involve
more and more droplets, and in the end this will induce an inexorable macroscopic
deformation of the material. Put another way, it will begin to flow, so a concentrated
emulsion is a yield stress fluid, i.e., solid below a certain level of stress, and liquid
beyond. The microstructural origins of this behaviour are analogous in some ways
to those described in the case of a colloidal suspension. However, the systems we
are discussing here are essentially repulsive in the sense that the elements do not
aggregate but repel one another as far as they can. For this reason, the restructuring
processes evoked to explain the behaviour of colloids are almost instantaneous and
only slightly influenced by any Brownian motion of the droplets. Finally, ‘pure’
emulsions, i.e., without any particular additives than surfactants, are apparently not
thixotropic.
The yield stress plays a determining role when the mayonnaise is used, because
it determines the force required to manipulate it and the form it will take under the
action of gravity. If the yield stress is low, deposits will not keep their initial form
and the mayonnaise will form a pool on the plate. The value of the yield stress can
be adjusted through the size of the droplets. However, in contrast to a commonly
held opinion, the concentration of droplets cannot be modified by reducing their
size, in other words, by stirring the mixture for longer or more energetically. Indeed,
observing each system on a length scale proportional to the size of the droplets it
contains, we recover the same geometric characteristics and hence the same volume
ratios. On the other hand, by reducing their size, we increase the elastic resistance of
each droplet, which is directly related to the macroscopic behaviour of the system.
We thereby raise the yield stress of the material.
Let us look a little more closely at the origin of this phenomenon. The solid struc-
ture of the emulsion is associated with the existence of an arrangement of droplets,
each one trapped among the others. To break this arrangement, a certain number
of droplets must be released momentarily. This can only be achieved by deforming
them to some extent. The deformation of a droplet induces an increase in the area
of its interface with the surrounding liquid which will be proportional to the square
of the diameter of the droplet. The gain in surface energy is proportional to this
increase. For its part, the relevant volume of emulsion will be of the order of the
cube of the droplet diameter. Computing the ratio between the above surface energy
and this volume, we deduce that the energy required per unit volume is inversely
proportional to the droplet diameter. This energy corresponds to the work that must
be done to bring about the flow of the emulsion. It is thus proportional to the applied
force, which is itself proportional to the yield stress. When the droplets are smaller,
the yield stress thus increases as the reciprocal of the droplet size (see Sect. 6.5.5).
Our mayonnaise will be all the more ‘solid’, i.e., its yield stress will be all the higher,
the longer and the harder we are prepared to beat it, thereby reducing the droplet
size. But it should be remembered that the thickest mayonnaise is not necessarily the
best!
1.8 Foams 29
Fig. 1.18 a Internal structure of a foam. b Shaving foam. c Chocolate mousse
1.8 Foams
Foams are made by trapping pockets of air in a liquid. They have an analogous
structure to emulsions. Indeed, we still have fluid inclusions, air in this case, in a
liquid (see Fig. 1.18a). They are prepared by shaking and mixing the liquid in order
to get the air pockets into it. There then follow the same stabilisation problems as
for an emulsion, and as before, an appropriate surfactant must be distributed over
the air–liquid interfaces. In cakes based on any kind of mousse (see Fig. 1.18b), it is
lecithin from egg yolk which once again plays this crucial role.
Foams are used for a whole range of applications, but often enough fluid foams
are required to form a light medium that is able to remain rigid under its own weight.
This is the case for example with shaving foam. The aim is to have soapy water on
the skin, and the foam provides a way of stopping the water from running down
one’s face during the shaving process. When such materials are used in foodstuffs
like chocolate mousse, meringue, soufflés, and Chantilly cream, for example, the aim
is to spread the food over a volume that looks much bigger than it really is, giving
it also a smooth or creamy texture which better spreads its taste around the mouth
and facilitates grinding and digestion. To achieve this, the mousse must have a high
enough yield stress to ensure that the various added ingredients remain harmoniously
dispersed until use.
As for emulsions, the behaviour of foams depends mainly on the concentration of
bubbles and their size when the concentration is high enough. When egg whites are
beaten, we first create a few inclusions which slightly increase the viscosity of the
mixture. In contrast to an emulsion (see above), during the preparation of a foam or
mousse, the amount of liquid remains constant while the volume of air trapped in it
increases. After a certain time, the bubble concentration is sufficient to guarantee a
yield stress behaviour. At this juncture, we obtain the desired effect in the form of a
foam that is no longer liquid but which on the contrary looks like a solid at rest. In
this case, the yield stress is explained by an arrangement of bubbles, each trapped
between its nearest neighbours. The mechanical strength of this foam can then be
increased by changing the size of the bubbles. As for the droplets in an emulsion, the
smaller the bubbles, the greater the surface to volume ratio and the higher the yield
stress. At this stage, the idea is to beat the foam without introducing any more air.
30 1 Introduction
In the end, the process will find its own equilibrium because the yield stress of the
foam will increase until air can no longer enter the mixture.
An interesting feature of foams is that one can reach a concentration very close to
100 %, producing what is then called a dry foam. In this situation, the films are very
thin and the bubbles no longer really exist in their own right. The structure of the
foam is based on an arrangement of neighbouring films obeying very precise rules
known as Plateau’s laws (with reference to the Belgian physicist who first discovered
them). These concern the angles along the line of contact between films and the angle
between two lines of contact. When the system flows, the problem is no longer to
release each bubble from its cage as it would be for lower concentrations, but to
deform the films, getting rid of some of them if necessary, while always respecting
the laws found by Plateau.
1.9 Granular Materials
Another kind of material with properties midway between those of a liquid and those
of a solid is the granular medium, which can be defined as a material comprising
a large number of solid particles whose direct interactions play a basic role in the
mechanical behaviour of the material. Concerning the mechanical behaviour, it is not
easy to classify these granular materials into the usual categories. Indeed, a powder
can flow like a liquid, as happens in an hour glass (see Fig. 1.19a). On the other hand,
a pile of sand seems to behave like a solid, maintaining its conical shape with quite
a steep slope, despite the action of gravity and without the need for walls to hold
it up. This is how sand dunes can form (see Fig. 1.19b), and it is also observed in
the manipulation and storage of gravel (see Fig. 1.19c). We know that we can walk
on the beach without sinking right into it, which would not be possible if sand was
a simple liquid. Since we are less dense, we would only sink in to a certain depth,
rather like an iceberg. Compared with yield stress fluids, the surprising thing about
granular materials is that, under certain circumstances they can remain in a solid
state under very high stress, but then behave like liquids when subjected to very low
stresses (a simple vibration can be enough). Finally, sufficiently fast granular flows
have similar characteristics to the propagation of a gas (see Fig. 1.19d).
These original features can be explained by the fact that, on the scale of the basic
components, the grains, a granular medium has two key properties that distinguish
it from other fluids:
• Two elements (particles) can push on one another with a very large normal force,
so the medium has certain properties of a simple solid on the local level.
• To set a granular medium in motion from rest, the structure must first expand
slightly, so that a grain can be extracted from the cage formed by its neighbours.
In contrast to a colloidal suspension, an emulsion, or a foam, it is impossible to
momentarily bring two neighbouring elements closer together. So when there is
no way for the material to dilate, it cannot possibly flow. Unless we impose a large
1.9 Granular Materials 31
Fig. 1.19 Different kinds of granular flow. a In an hour glass. b Sand dunes at rest. c Piles of gravel.
d Pyroclastic flow of volcanic ash on the flanks of a volcano
enough force to break some of the grains, the medium remains stuck where it is
(see Sect. 7.3.2). On the other hand, when a grain is extracted from its cage, no
force is actually acting on it and it may gain enough energy to ensure that it could
no longer get trapped in this way.
These are the effects which allow us to walk on a pile of sand while at the same time
being able to produce very fast avalanches of grains on its surface.
When a granular medium is not flowing, or is flowing very slowly, the particles
form a network of contacts which changes all the time as the medium deforms itself.
Locally, the relative motion of two particles involves a tangential force proportional to
the normal force applied at the point of contact (see Sect. 7.2.2). This is the Coulomb
model. It turns out that this model, valid for two solids in contact and in relative
motion, is also valid for a slowly flowing ensemble of grains (see Sect. 7.5). At first
glance, the fact that a yield stress must be exceeded in order to induce a flow implies an
analogy with yield stress fluids like colloids, concentrated foams and emulsions, and
gels. However, this behaviour differs significantly from what is observed for liquids
and all the complex fluids reviewed above through the fact that the apparent viscosity
depends here on the normal pressure. This is indeed the signature of a granular
medium. The material flows less and less easily as the normal force increases.
32 1 Introduction
This leads to a novel situation for flows under certain conditions. In a cylindrical
container filled with liquid and with a small outlet at the bottom, the flow will be
faster when the depth of liquid is greater. The pressure close to the outlet, which
determines this speed, is in fact simply related to the weight of liquid located above
it. If this container is filled with sand or small beads, it is observed that, above a
certain value, the depth of material has no further effect on the flow rate, as though
the pressure close to the outlet were constant (see Sect. 7.5.3). This is effectively what
happens, because the granular material is held back by a friction force along the walls
that is roughly proportional to the weight of material above the given region. It thus
holds onto the walls rather as a mountaineer would do to climb up between two rock
faces.
When a granular medium is flowing fast enough, there is no longer a network of
contacts as described above and energy is essentially transferred between particles
through collisions. In this situation, frictional stresses resulting from the motion are
analogous to those in a gas: under simple shear, each particle in a moving layer
exchanges energy with the particles in adjacent layers (see Sect. 7.6). However, this
flow regime is only found under very special conditions. The system must expand
sufficiently for the particles to be able to travel a certain distance between successive
collisions, and at the same time the inertia of the particles must be large enough to
ensure that collisional energy transfers predominate over other energy dissipation
phenomena such as the viscosity of the interstitial fluid, plasticity of the particles,
and friction between particles.
When the interstitial fluid of a granular medium is a liquid, the system is similar
to a highly concentrated suspension of non-colloidal particles. At low velocities, one
thus expects to have a frictional type of behaviour as above, because the hydrody-
namic interactions will be negligible. But at higher velocities, these hydrodynamic
interactions can begin to dominate over friction, thus yielding the more standard
behaviour of a concentrated suspension. The material then behaves essentially like
a high viscosity liquid (see Sect. 7.7.2). However, the presence of the liquid has a
significant impact on the behaviour of the granular medium at low velocities, in the
frictional regime. Indeed, we have seen that, when a dry granular medium undergoes
slow simple shear, the tangential stress is proportional to the normal force, where
the latter often results simply from the weight of material bearing down. When the
grains are immersed in a liquid, the buoyancy force exerted on each grain according
to Archimedes’ principle reduces the apparent weight of the grain, whereupon the
normal force in the granular network is decreased. This implies that the tangential
force required to induce motion will also be decreased, so the medium will resist less
if the grains are immersed in a liquid.
This effect can be further enhanced if the liquid moves vertically through the
porous medium constituted by the pile of grains. Such a flow then induces a pressure
gradient in the liquid (just as one must push the piston to get the liquid out of a
syringe), and this will have its effect on the pile of grains, further reducing the
normal forces. Combined with the bouyancy force, these effects can end up reducing
the apparent weight of the grains to the point where the granular network puts up no
resistance at all to shear. This is what happens in the so-called fox effect which occurs
1.9 Granular Materials 33
at the foot of earth dams: as water infiltrates slowly through the dam, it can weaken
the material situated downstream to the point where this material starts to flow. This
is also the phenomenon operating in certain quicksands, where water moves slowly
through a granular network made from very fine sand.
1.10 Real-Life Materials
In nature as in industry, materials rarely comprise a single type of element (polymer,

grain, bubble, droplet, etc.). Most real materials contain several types of element, each
with its own range of characteristics (size, shape, adsorbed ions, etc.), sometimes
all within the same category. Concrete is the archetypal example of this type of
mixture. It contains, in suspension in water, very large non-colloidal grains (sand and
gravel), particles (cement and fly ash) of intermediate size between the colloidal and
non-colloidal ranges, colloidal particles (e.g., nanosilica), and polymer additives.
Likewise, naturally occurring muds contain a very broad range of particle types,
including various classes of clay, silt, sand, and sometimes rocks. The same can
be said for most creams, mortars, paints, ceramics, and so on. This is generally a
consequence of the way the material has been designed and redesigned over decades
or centuries, as different ingredients were tried and tested to adjust the behaviour.
Despite this complexity, certain rheophysical properties of such materials can be
understood by noting that, if one type of interaction predominates, it will impose
much the same type of macroscopic behaviour on the mixture as it would when
operating alone. This works, for example, with naturally occurring muds. As long as
the fraction of non-clay particles is small enough, the paste formed by the water–clay
mixture constitutes a matrix which imposes its own typical behaviour on the whole.
The point is that the grains can be treated simply as suspended in this matrix, since
the colloidal interactions between clay particles then predominate. When the grains
have a high enough concentration, they engage in frictional contact and the mixture
begins to behave more like a granular medium. Depending on the material, it may
or may not be easy to work out which elements are actually playing the main role,
but it remains a worthwhile approach to the rheophysics of complex systems.
Reference
1. Robinson, D.J., Earnshaw, J.C.: Phys. Rev. Lett. 71, 715 (1993)
Chapter 2
Simple Materials
Abstract In this chapter we are concerned with materials comprising a single type
of elementary constituent, in the form of identical atoms or small molecules. These
constituents exert forces on one another which decrease with the distance between
them. When thermal agitation, which increases with the temperature of the system, is
stronger than the interaction forces which tend on the whole to bring the constituent
elements together, the matter is in the gaseous state. In this case, the mechanical prop-
erties of the material are associated with collisions between molecules and statistical
tools can be used to obtain exact relations between the viscosity and the physical
characteristics of the gas. When the attractive forces are strong enough compared
with thermal agitation, the basic constituents tend to form a compact cluster. This is
the liquid state. In this case, our understanding of the relationships between internal
forces and the dynamical evolution of such disordered structures in flow remains
incomplete. In certain cases, for example, at lower temperatures or higher pressures,
the material may become slightly more compact than in the liquid state and arrange
itself into an ordered structure. This is the solid state. The material can then be
deformed to a certain extent and a direct relationship can be established between the
force required and the local interaction forces.
2.1 Introduction
Here we shall be concerned with materials comprising a single type of elementary

constituent in the form of identical atoms or small molecules. We thus exclude
polymers for the moment. These materials are therefore simple with regard to their
composition, containing a single type of element which is not only undeformable
but also indestructible under ordinary conditions. It should be said that this is no
guarantee of simplicity in rheophysics, since mechanical behaviour does not depend
solely on intrinsic characteristics of the constitutive matter. It depends also and above
all on the interactions prevailing between these elements, that is, the different types

DOI: 10.1007/978-3-319-06148-1_2,
36 2 Simple Materials
of forces between them, and on the structure of the material, that is, the relative
spatial arrangement of the elements. Regarding structure, the simple materials fall
into several main categories, associated with different states of order and/or density
of the basic elements, and in which certain types of interaction are predominant.
The atoms or molecules exert forces on one another which fall off as the distance
between them increases. Another factor plays a crucial role in systems made up
of small elements, namely thermal agitation, which increases with the temperature
of the system. Each element is continually subjected to this phenomenon, which
tends to impart random motions to it in all directions. When this agitation dominates
over the interaction forces, which for their part tend to bring the elements closer
together, the elements will disperse as far as possible throughout the available volume,
occasionally colliding with one another. This is the gaseous state. In this situation,
the mechanical properties of the material will be associated primarily with these
collisions. The force required to compress it or cause it to flow will depend on what
is required to modify the number or strength of these collisions. Using statistical tools,
we can then obtain exact relations between the viscosity and physical characteristics
of the gas (see Sect. 2.2).
When the attractive forces are of the same order as thermal agitation, the elements
tend to clump together into a compact cluster. This is the liquid state. In this situation,
owing to thermal agitation, even though the elements remain close together, they can
nevertheless move relative to one another, provided that enough of their nearest
neighbours also change position, rather as happens in the celebrated Fifteen Puzzle.
This means that the structure of a liquid, like the structure of a gas, is not frozen in,
so to speak, but nevertheless remains statistically identical. However, we still have
only a limited understanding of the relationship between the internal forces and the
evolution of this disordered structure under flow conditions. So even for this simple
material, we reach the limits of present day rheophysics, which, except in certain
special cases, has great difficulty explaining the behaviour of condensed matter under
flow conditions (see Sect. 2.3). We shall come up against this problem again when we
discuss the liquid regime of other, in principle more complex, disordered materials,
such as colloidal suspensions, foams, and emulsions.
In certain cases, e.g., at lower temperature or higher pressure, the material may
become slightly more compact than in the liquid state and organised itself into an
ordered structure, the solid state. At this point, internal interaction forces dominate
over thermal agitation, which is no longer able to induce relative motions among the
elements of the structure. The latter is now effectively frozen in. However, the material
can still be deformed to some extent, and a direct relation can be established between
the force required to do this and the local interaction forces (see Sect. 2.4). From the
rheophysical standpoint, much information is provided by this situation. In a similar
way, we shall be able to understand the rheophysical behaviour of various condensed
systems such as foams, emulsions, and colloids in their solid state. Although these
do not have a crystal structure, they do form a ‘jammed’ structure, from which the
constitutive elements are unable to escape under the action of thermal agitation alone.
2.1 Introduction 37
Finally, there is an intermediate state between the solid and liquid states, known as
the glassy state. Here the structure is disordered as in a liquid, but thermal agitation
is not sufficient to allow spontaneous motion, as in a solid (see Sect. 2.4).
2.2 Interactions Between Elementary Components

and States of Simple Matter
2.2.1 Elementary Components
All matter is made up of some ensemble of atoms. To each chemical species there
corresponds an atomic species. An atom comprises a nucleus and a certain number
of electrons which move around this nucleus. Given the very strong forces needed
to remove one of these electrons, we shall assume that, in all the physical transfor-
mations dealt with here, the atom is effectively indestructible. Atoms are usually
associated together in groups known as molecules. In a molecule, the atoms are
bound together by so-called valence forces which arise when they effectively share
the electrons of their outermost electron shells. These forces are also strong enough
to ensure that a molecule will not be destroyed during ordinary physical transfor-
mations. The electron clouds of two atoms or two molecules cannot penetrate one
another because a very strong repulsive force builds up whenever such clouds come
within range. We may therefore treat an atom or a small molecule of a given chemi-
cal species as an entity with a definite undeformable volume, assumed spherical for
simplicity, whenever it is isolated. In the following, we shall use ‘molecule’ as a
generic term when the relevant physical phenomena are independent of the internal
structure of the elements (atoms or molecules). These molecules interact in various
ways depending on the separation between them, and the mutual interaction forces
may differ qualitatively.
2.2.2 Thermal Agitation
When a molecule is far removed from the other molecules in the system, it will not feel
any force from them. Its motion in the vacuum is then governed by Newton’s second
law which, when there are no external forces (we neglect gravity here), implies that it
will move with constant velocity. But for the molecule to have acquired this velocity,
there must have been an impulse of some kind at an earlier time, imparting a certain
energy to it. This is the energy of thermal agitation. If we consider systems made up
of many molecules, we observe that these molecules all have different speeds and
directions, i.e., they all have different velocities. The magnitudes of these velocities
will be denoted here by c.
In order to quantify this agitation, we consider the average kinetic energy of the
molecules. For simple systems (pure ideal gases), this has the form
1 3kB T
m◦c2 = , (2.1)
2 2
where T is called the temperature and kB = 1.38 × 10−23 J K−1 is Boltzmann’s

constant. In this expression, the mass m of a molecule is assumed constant for all
molecules in the system.
Equation (2.1) shows that the temperature provides a quantitative measure of the
agitation of the molecules in a system where they are widely separated from one
another. In fact, this idea is quite general. For any system, whatever the arrangement
and proximity of the molecules, the quantity kB T can be used to estimate the internal
energy of thermal agitation of the elementary components. This is equal to about
0.6 × 10−20 J at typical temperatures. Thermal agitation tends to disperse a system.
The internal interactions must be stronger than thermal agitation, therefore, in order
to keep the molecules close to one another, as happens in a liquid or solid.
2.2.3 Interaction Potential
To describe the interaction forces between arbitrary objects such as molecules, it is

convenient to phrase things in terms of energy. Indeed, a specific energy function
can be defined for each force in such a way that the force can then be derived from it.
Consider a system comprising two otherwise isolated interacting bodies, i.e., exerting
a force F on one another which depends on the distance x between them and which is
of course zero when the two bodies are infinitely far apart (see Fig. 2.1). When there
are no other forces, external to the two-body system, we can define the interaction
potential energy of the system as the energy required to bring the two bodies from
infinity to a separation x. In the rest frame of the body at the origin, the idea is to bring
the other body toward it from infinity to the position x. Throughout this operation,
we must apply a force −F(ξ ), where ξ is the distance between the two bodies. The
required energy is then the work done during this transfer, viz.,
x
=− F dξ .
∞
Differentiating this expression, we obtain
d
F=− . (2.2)
dx
2.2 Interactions Between Elementary Components and States of Simple Matter 39
Fig. 2.1 Interaction force F

and interaction potential (x)
F
defined as the energy required Φ(x)
to bring the particles from
infinity to a given distance x
from one another
x
2.2.4 Van der Waals Forces
Whatever the type and structure of the molecules and their constituent atoms, there
is always a short range attractive force between two molecules. This force arises
because, even though the electrons are distributed uniformly on average within an
atom or molecule, their instantaneous distribution will always be asymmetrical. It
follows that the particle will behave as an instantaneous electric dipole, and this dipole
will induce an electric field in the neighbouring atom which will in turn acquire a
dipole moment, whereupon the two dipoles will attract one another.
We can use the following highly simplified argument to get some idea of the form
of this interaction. Consider an atom in which the charge distribution constitutes a
dipole with moment p, which is the sum of the products of the charges with their
distances from a reference point located at the centre of the charges. At a distance
r which is large compared with the size of the dipole, it will induce an electrostatic
field E going as r −3 . Another dipole lying at this distance will then be polarised, that
is, it will acquire a dipole of moment p = αE. The potential energy of interaction
between these two dipoles can be written = p E, which is in this case αE 2 . Finally,
we find that is proportional to r −6 .
This expression is no longer valid when the molecules are too far apart (more
than about ten nanometres), because the time taken by the electric field to act on the
other particle is then of the same order as the typical time taken by fluctuations to
vary the dipole in the first particle. This retardation effect implies that the potential
goes rather as r −7 .
For atoms or molecules ‘in contact’, the interaction energy associated with these
van der Waals forces corresponds to separations r roughly equal to the molecular
radius, hence typically of the order of a tenth of a nanometre. The value obtained, of
the order of 10−20 J, is close to the thermal agitation energy at room temperature. Put
another way, these attractive forces, generated by effects that one might well have
neglected at first glance, can play an important role in the equilibrium and structure
of the system. However, this force decreases very rapidly with distance. In fact, at
a separation of the order of the size of the molecules, it is already a hundred times
smaller than when the molecules are in contact. Van der Waals forces thus arise
mainly to ensure cohesion between molecules in contact.
2.2.5 Chemical Bonds
Within a crystal structure, molecules give up something of their independent exis-

tence in favour of ordered arrangements and strong bonds between the atoms making
them up. The atoms can then bind together by valence bonds, as within a molecule.
The molecules can also bind together by ionic bonds. When two atoms exchange an
electron by emptying an incomplete shell of one of the atoms to fill a different level
of the other atom, they then interact like two ions of opposite sign. The interaction
energy between the two ions is of Coulomb type and generally much greater than
the thermal energy, whence the atoms are strongly attracted to one another. To a
first approximation, the corresponding interaction potential falls off essentially as
the reciprocal of the distance between the elements. The order of magnitude of the
energy of a covalent bond or an ionic bond is 50 × 10−20 J, hence significantly
greater than the energy of van der Waals forces or thermal agitation.
2.2.6 Born Repulsion Force
This effect accounts for the impenetrability of the electron clouds of atoms or mole-
cules. There is no general expression describing this force. Various empirical models
have been developed to describe it. Their common feature is that they all predict that
the force tends to infinity more quickly than all other known forces when the distance
between the two elements tends to zero. The model most widely used to represent
this effect is a power law, viz., (r) = β/r m , where m takes a value greater than 7
to yield a potential much greater than all the other possible forces within a certain
range. To simplify, the hard sphere approximation is used in some cases. This takes
the potential to be infinite below some critical separation d of the molecules, whence
d can be viewed as specifying their effective diameter. When there are no attractive
forces, this potential is (see Fig. 2.2a)
r < d =⇒ (r) = ∞, r > d =⇒ (r) = 0 . (2.3)
2.2.7 Balance of Forces
When several types of force are at work between two molecules, the total interaction
potential is the sum of the corresponding potentials. In particular, when these forces
are just van der Waals forces and the Born repulsion, a model commonly used to
represent the total potential simply sums these two potentials, taking m = 12 for
the repulsive potential (see Sect. 2.2.6). We then obtain the so-called Lennard-Jones
potential (see Fig. 2.2c):
β α
(r) = 12 − 6 . (2.4)
r r
(a) (b) (c)

Φ Φ Φ
d r d r d r
Fig. 2.2 Different models for the balance of interactions between molecules. a Hard-sphere poten-
tial. b Hard-sphere potential with van der Waals attraction. c Lennard-Jones potential
This model can be simplified by representing the repulsive part by a hard-sphere

potential within some critical distance. This is taken to be the equilibrium distance
d of the molecules, associated with the minimum of the actual potential energy (see
Fig. 2.2b). Hence,
6
d
r < d =⇒ (r) = ∞, r > d =⇒ (r) = −w , (2.5)
r
where w is the maximum attractive potential, or adhesion potential, obtained when

the molecules can be considered to be in contact, i.e., when r = d. The models
(2.4) and (2.5) are used in particular to describe the behaviour of liquids close to the
gas–liquid transition.
For solids, one can also use a more general expression for the total potential in the
form of a sum of an attractive potential and a short-range repulsive potential, viz.,
α β
(r) = − n
+ m , (2.6)
r r
where n = 1 and m = 9 for a solid with essentially ionic bonds between atoms,
n = 6 and m = 12 for a van der Waals solid, and n = 1 and m = 2 for monovalent
metals, i.e., with a single electron in their outer electron shell.
2.2.8 Hydrogen Bond and Hydrophobic Forces
The hydrogen bond occurs when a hydrogen atom is covalently bound to an elec-
tronegative ion such as oxygen or nitrogen. The electronegative atom will attract the
electrons of the hydrogen atom very strongly towards it, thereby inducing a highly
unbalanced charge distribution, to such an extent that the hydrogen atom will appear
to be positively charged. Since two charges of opposite signs attract, the hydrogen
atom can then interact electrostatically with another electronegative atom. The corre-
sponding interaction energy is of the order of 1.5–6.5 × 10−20 J, that is, somewhere
between the energies of a van der Waals interaction and a covalent bond.
One consequence of this phenomenon is that water molecules in solution will tend
to arrange themselves in such a way as to form as many hydrogen bonds as possible.
For this reason, when a different molecule is immersed in water, the water molecules
can react in different ways depending on the size of this molecule and its affinity
with water, i.e., its ability to develop hydrogen bonds. For example, a molecule that
has no particular affinity with water will cause the molecules around it to arrange
themselves in such a way as to preserve as many hydrogen bonds as possible between
them. This will reduce the number of possible arrangements of the molecules, and
hence also the entropy of the system (see Sect. 2.4). When several elements of this
type are placed in water, it is more favourable in terms of the entropy of the system
for these elements to be in contact with one another, since this will reduce the area
over which the water molecules must arrange themselves in some specific way. This
amounts to introducing attractive forces between such elements. Interactions of this
kind can destabilise a two-phase mixture by tending to make the elements of each
phase gather in certain regions of the sample.
2.2.9 States of Simple Matter
Here we consider a system made up of molecules that start out far removed from one
another. Such a system is in a gaseous state, i.e., the elements are in an excited state
and only encounter one another on rare occasions (see Fig. 2.3). Disorder reigns in
this state. When two molecules meet, there will be an attractive force between them,
but as long as the thermal agitation remains great enough, they will not be able to
hold on to one another for long. If we now reduce the temperature or the volume
available to the system, the kinetic energy of the particles will decrease or collisions
will become more frequent, so the particles will remain together for longer.
Below a critical temperature or volume, a condensed phase will arise, namely the
liquid state (see Fig. 2.3). In this phase, the molecules are still agitated and the density
is not yet optimal. The molecules are held very close to one another thanks to the
van der Waals forces, but thermal agitation is still sufficient to maintain spontaneous
relative motions of the molecules.
Below a certain temperature, the elements will organise themselves in a regular
manner, in the ordered arrangement of a crystal. This generally allows the substance
to obtain its optimal density. In this solid state, the particles are almost held motionless
in their positions (see Fig. 2.3), since thermal agitation is now much weaker than the
energy associated with the chemical bonds that have become established.
Finally, there is another state of matter, intermediate between liquid and solid,
which can be reached by certain materials under certain conditions, in particular
when we try to obtain a solid somewhat too quickly by reducing the agitation of
its constituent elements without leaving them the time to order themselves into a
GAS LIQUID GLASS SOLID
Fig. 2.3 Structure and mobility of the constituent elements in different states of matter. Molecules
are represented here by black disks and their trajectories by straight line segments. The particle
trajectories in the solid and glass have been exaggerated for clarity
crystalline arrangement. We then obtain an amorphous or glassy state. In this state, the
resulting glass has a disordered structure similar to that of a liquid, but its constituent
elements remain almost fixed in place, as in a solid.
2.3 Gaseous State
A gas is made up of widely separated molecules, in fact at separations much greater

than their own dimensions. The molecules have velocities with a range of different
directions and magnitudes. Each encounter between two molecules or with a solid
wall gives rise to a collision. There is no other means of energy transfer within the
system. The mechanical properties of the gas, that is, the way the material reacts
on the macroscopic scale to the forces applied to it, are thus related to the energy
exchanges through these collisions. In order to establish this relationship, it is useful
to begin by characterising the state of the system in terms of the velocities and
relative positions of the molecules. We will then be able to determine the mechanical
properties of the material.
2.3.1 Velocity Distribution
Assuming simply that the agitation of the molecules in a given system is uniform in
some statistical sense, one can establish the average characteristics of the molecular
velocities without making further physical assumptions. This statistical uniformity
states that the average velocities of the molecules, measured over volumes containing
a large enough number of molecules or over long enough observation times, are
identical in all directions and throughout the volume of the sample.
The velocity distribution is described by a probability density function P(c), such

that the probability of finding a velocity with magnitude between c and c+dc is equal
to P(c)dc. We must take into account the fact that the velocities can have different
directions. The velocity vector can be represented by its three components u, v, and
w in a Cartesian coordinate system. The probability of the velocity having its three
components in the ranges from u to u+du, v to v+dv, and w+dw, respectively, is then
f (u)f (v)f (w)du dv dw. Since the velocity distribution is independent√ of the direction,
the function f (u)f (v)f (w) will depend only on the magnitude c = u2 + v2 + w2 of
the velocity, and we shall write it F(c). Note that this is not the same as P(c) because
here we are only considering particular forms of velocity vectors. The direction of
the velocity is independent of its magnitude and the components of the velocity are
mutually independent. Therefore, changing variable and using the expression for c
as a function of u, we have
∂ ln F u ∂ ln F
= ,
∂u c ∂c
and noting also the relationship between F and f , which implies that
∂ ln F d lnf
= ,
∂u du
it follows that
1 ∂F 1 df (u)
= . (2.7)
cF ∂c uf du
The same can be done for the other velocity components, with analogous results.
Now, since c and u can in part be varied independently, each side of (2.7) must
be constant. Writing this constant in the form −m/B, where B is a constant, and
integrating the resulting differential equation
df m
= − uf ,
du B
we find
mu2
f (u) = A exp − , (2.8)
2B
where A is a constant. If B were negative, the probability of having a velocity in a

specific direction would tend to infinity when the magnitude of this velocity tends to
infinity, which is not realistic. B is therefore positive. Note also that the function f is
symmetric in the velocity, i.e., it does not favour any particular direction of motion.
So starting from the simple assumption that this kind of agitation does not favour any
particular direction, we have shown that the velocity distribution in one direction is
Gaussian, centered on zero. More detailed theories of statistical physics confirm this
result.
2.3 Gaseous State 45
We may now calculate the probability P(c)dc that the magnitude of the velocity
should lie between c and c + dc, using the fact that this is the sum of the probabilities
that the velocity vector should have magnitude c and arbitrary directions θ and ϕ,
where the latter range from 0 to π and from 0 to 2π , respectively. We then have
du dv dw = c2 sin θ dθ dϕ dc, whence the probability distribution for the magnitude
of the velocity can be written

P(c) = f (u)f (v)f (w)c2 sin θ dθ dϕ .
u2 +v2 +w2 =c2
This then implies

mc2
P(c) = 4π A3 c2 exp − .
2B
The two constants A and B can be determined by using the fact that the total probability
must be equal to 1, viz.,
∞
P(c)dc = 1 ,
0
together with the fact that the average value of the kinetic energy is given as a function
of temperature by (2.1), viz.,
∞
3kB T
◦c2 = c2 P(c)dc = .
m
0
After several integrations by parts and using the standard result
∞ √
π
e−x dx =
2
,
2
0
we find that B = kB T and

1/2
m
A= .
2π kB T
Finally, the velocity distribution is given by

3/2
m mc2
P(c) = 4π c2 exp − . (2.9)
2π kB T 2kB T
We can now calculate the average value of any quantity depending on the velocity,
such as the kinetic energy. In particular, the average value of the magnitude of the
velocity is
∞
kB T
◦c = cP(c)dc = 2 . (2.10)
m
0
2.3.2 Mean Free Path
The molecules of a gas are moving around all the time, so even though they are
widely separated from one another, they occasionally end up in collisions. In fact,
these collisions are needed to maintain the state of statistical equilibrium which
ensures uniformity of the velocity distribution. This agitation effectively determines
the transport properties of the gas (viscosity, diffusion, thermal conductivity) which
are associated with energy transfer from one point of the system to another. In this
context, an important quantity is the characteristic time for exchange of momen-
tum between two molecules. This time is equal to the ratio of the distance between
collisions and the speed of the molecules. Since we already know the velocity distri-
bution and the average velocity (2.10), it remains only to identify the typical distance
between two successive collisions of a given molecule, referred to as the mean free
path.
For a collision to occur, the molecules must have a nonzero effective diameter d.
Then any other molecule on its path with centre a distance less than d from the centre
of the first in the direction perpendicular to the motion will enter into collision with
it (see Fig. 2.4). Let us follow the path of a molecule, assuming that the others are not
moving on average. When this molecule has travelled a distance L, it will have swept
out a volume Lπ d 2 . If n is the number of molecules per unit volume, the number of
encounters with other molecules is thus nLπ d 2 . The average distance between two
collisions is the distance allowing a single encounter, viz.,
1
λ= . (2.11)
nπ d 2
In fact, (2.11) does indeed give the exact value up to a multiplicative factor close to
unity, but one must take into account the velocities of the other molecules and the
changes in direction induced by each collision.
2.3.3 Entropy
It makes no sense to try to describe the spatial distribution of the molecules directly
because, given the agitation prevailing within the system, each configuration is
Fig. 2.4 Estimating the mean

free path. This is the average
distance travelled by a mole-
cule (black) before entering
into a collision with another
molecule (grey) in a given d
d
direction
equiprobable. However, the number of degrees of freedom in positioning the elements

within the available volume can be used to distinguish one system from another. To
describe this idea precisely, we calculate the number Z of microscopic states that
can occupy the system when it is in a given macroscopic state. We then define the
entropy, a function of this number of microscopic states with the form S = kB ln Z.
As shown in Appendix B, the entropy is related to the free energy of the system. This
will prove particularly useful when describing the evolution of molecular systems or
more complex systems.
In the case of an ideal gas, assuming that the internal states of the molecules are
constant, we can calculate the entropy by counting up the various microscopic states
specified by the positions and velocities of the molecules in the available space. We
begin by counting the number of possible spatial configurations for the molecules of
a gas comprising N molecules in a volume Ω. In an arbitrary volume, the centre of
each molecule can of course sit at infinitely many different points. To simplify the
calculation, we first assume that the centres of the molecules can only sit at a finite
number of positions in a space divided up into the same number of small elementary
volumes ν associated for example with the typical volume of a molecule. Note that,
to obtain a better estimate of the real situation, we could choose these volumes to be
much smaller. Neglecting the volume occupied by the other molecules in comparison
with the volume available, which amounts to assuming that Nν Ω, there are to
a first approximation Ω/ν ways of placing each molecule in the given volume. The
number of possible spatial configurations for the N molecules is thus (Ω/ν)N . For
identical (in fact, indistinguishable) molecules, one cannot distinguish two states that
differ only by a permutation of the molecules. This means that one must divide the
above number by the number N! of permutations of these molecules. The number of
distinct spatial configurations is therefore (Ω/ν)N /N!.
The number of possible configurations for the velocities of the molecules could be
calculated from the velocity distribution. Since the characteristics of this distribution
are related to the thermal agitation, it will suffice here to note that this number is
a function f of the energy U of the system and the number of molecules. We then
obtain
f (U, N)(Ω/ν)N
S = kB ln .
N!
This expression can be simplified using Stirling’s formula for ln N!, which gives
ln N! ≈ N ln N to first order. Finally, we arrive at the expression

Ω
S = kB N ln + ln f (U, N) + C , (2.12)
N
where C = −N ln ν. In (2.12), the parameter ν thus only induces changes in the

additive constant C. For a given system, the number N is fixed and we are concerned
with the relative changes in the variables S, Ω, and U of the system. These changes
do not depend on the initial choice of volume ν.
2.3.4 Ideal Gas Law
2.3.4.1 Volume and Pressure
Consider a sample of gas placed in a solid box and hermetically sealed, except that
one of its faces is in fact a movable piston. The most natural variable to characterise
the constitutive material of this sample is its volume Ω, which is simply the volume
within the box here because, given the thermal agitation of the molecules, it is natural
to expect the gas to spread over the whole of this space. If we now inject more gas
into this box, either the piston moves, thereby allowing the volume of the box to
increase, or else the piston is held in place, in which case the force F required along
the piston axis in order to keep it in its initial position is found to increase during the
injection. There is therefore a relationship between the force and the volume of gas.
We also note that, in such a system, if we manage to increase the surface area
of the piston in contact with the gas while keeping the volume of gas constant, the
force increases in proportion to this area. It is thus natural to define a new variable,
the pressure p = F/A, which does not depend on the area A of the surface in contact
with the gas and thus characterises the state of the system.
For such a system, if each face of the box were made in the same way from a
movable piston, the same pressure would have to be applied on each of these faces.
In fact, this same pressure applies to each face of the box, and this would also be
true for a box with polyhedral shape of any kind at all. This implies that, whatever
surface element we may consider within the gas, the box could be deformed in such a
way that one of its faces corresponds to this element and we would recover the same
value of the pressure. For this reason, we may define the pressure at any point of the
gas, and this pressure will be uniform throughout the gas. Then given this pressure,
we can at last write down the force exerted by the gas on a virtual surface element of
area χA located inside the gas. This force is equal to −pχA n, where n is the unit
vector normal to the surface element (see Fig. 2.5).
Fig. 2.5 Forces induced by

the change in volume of a gas F
A
Ω
ΔA
2.3.4.2 Temperature
Another physical characteristic of such a system is its temperature T . In practice,

this variable can be assessed directly by our senses. If we compare two systems, we
are able to say whether one is at a higher temperature than the other. The remarkable
property of a gas is that, no matter how we manipulate the box discussed a moment
ago, the product of the pressure and the volume is proportional solely to the number N
of gas molecules in the system, provided that the temperature of the system remains
the same. If we just define the temperature to be proportional to some measure of
this product, we arrive at
pΩ = NkB T . (2.13)
This is an equation of state of the material since it expresses a general relation between
the fundamental physical variables associated with the material. It is also a first step
toward rheophysics since it expresses the pressure (which describes the essential
forces within the system) in terms of the physical characteristics of the system, and
in particular the temperature. In the following, we shall see that it is effectively
possible to quantify these different phenomena in terms of the physical properties
of the matter on the microscopic scale. In particular, this will allow us to establish
the consistency of this macroscopic approach and the microscopic description of
temperature using (2.1). Note also the consistency of the equation of state of an ideal
gas and the thermodynamic approach. Indeed, using (B.14), which tells us that

p ∂S
= ,
T ∂Ω U
and inserting the expression (2.12) for the entropy, we obtain (2.13) directly.
Fig. 2.6 Collision between a u

particle with initial velocity u u
and the wall P
u'
P
2.3.5 Kinetic Theory
To establish the relationship between the forces exerted on the gas and the macro-
scopic motions of this gas, we must pay more careful attention to the interactions
between the molecules and between the molecules and a solid surface. To tackle this
problem, we make the key assumption that these interactions are all elastic collisions.
In other words, we assume that, when one molecule encounters another, or when it
encounters a solid surface, there is no clustering effect resulting from short-range
interactions. In addition, the elastic nature of these collisions implies that energy
dissipation is negligible. These collisions can then be characterised by conservation
of momentum, and also by the conservation of kinetic energy in the system.
2.3.5.1 Pressure on a Solid Surface
Consider a particle arriving with velocity u at a solid wall P with total area A. This
velocity will have a normal component u along the unit normal n to the wall and
a tangential component which can be decomposed into two components v and w
relative to two perpendicular axes. In an elastic collision, since the wall does not
move, the particle will bounce off with a velocity u that has the same tangential
components v and w but normal component −u (see Fig. 2.6). The particles colliding
with the wall over a time lapse χt will be located in a layer of thickness χx = uχt
above the wall. If there are nu particles per unit volume with velocity between u and
u + du, there will be nu Aχx particles in this layer during the time χt, and there will
therefore be nu Au collisions per unit time.
By Newton’s second law, the resulting force on the wall is equal to the total
momentum imparted to the wall per unit time, i.e., f (t) = mdv/dt. For a collision
like the one described above, the molecule has a constant velocity u up to the time
when it hits the wall. Its velocity will then change quickly to a new value u . As
a function of time, the force associated with this collision is thus sharply peaked
about the moment of contact. Before contact, the force is zero, then at the beginning
of contact, the speed is reduced, inducing a positive force up until the time when
the molecule comes to a halt (zero speed). The force reaches its maximum value at
this time. Subsequently, the motion is reversed and the force will drop off more or
less symmetrically. As a result of this collision, the average force on the wall in the
normal direction n, between two times before and after the collision separated by a
time lapse χt, is thus
χt χt
1 1 2mu
◦f = f dt = m dv = . (2.14)
χt χt χt
0 0
The wall is struck many times by particles with the same velocity and the total
average force is thus the sum of the average forces (2.14) associated with each
collision, viz., F = nu Auχt◦f . Typical pressure measurements are made over areas
and time lapses such that the number of impacts taken into account is extremely
high. For this reason, one would never notice the fluctuations due to the succession
of collisions at different speeds, and the effective force measured is very close to the
total average force calculated above. Finally, the pressure exerted by all the collisions
taken together is given by
F
p = = 2nu mu2 . (2.15)
A
We now calculate the kinetic energy of a molecule. We have c2 = u2 + v2 + w2

and denote the number of particles per unit volume with speed c by nc . This set
of particles comprises subsets with different values of the velocity component u.
However, the square of each velocity component will have the same average over
all the molecules with speed c, because no direction is favoured. Each will thus be
equal to 1/3 of c2 , whereupon we obtain
1 c2
nu u2 = .
nc u 3
We can now calculate the pressure exerted on a wall by all the molecules with speed
c. In this ensemble, only those molecules with velocity directed toward the wall will
actually collide with it. This corresponds to half of the ensemble. The total pressure
exerted by the particles with speed c is thus given by

1 1
p= 2m nu u = mnc c2 .
2
(2.16)
2 u
3
Since by definition
1
◦c2 = nc c2 ,
n c
the total pressure due to the impacts of all the particles is given by
1
p= mn◦c2 .
3
Fig. 2.7 Simple shear. The A

material is sheared between F, V
the two solid surfaces (grey).
The fluid layers slide one over
the other, parallel to these H
surfaces
For a volume Ω containing a total of N particles, we have n = N/Ω and the above
equation becomes
1
pΩ = mN◦c2 . (2.17)
3
Using the expression for the average kinetic energy of a molecule due to thermal
agitation as given by (2.1), we note that (2.17) is nothing other than the equation of
state (2.13) of an ideal gas. We have thus found a direct relation between the pressure
in a gas and the kinetics of its constituent molecules.
2.3.5.2 Viscosity
In the above discussion, the gas as a whole was at rest, i.e., any motions occurring
within it did so without changing the shape of its apparent volume. We now consider
what happens when certain regions of the gas are in motion relative to others. In this
context, the simplest situation is when one plane layer of gas moves in a direction
lying within this plane and relative to two adjacent layers. To maintain such a relative
motion at some speed v between two layers, it turns out that one must exert a tangential
force F in the direction of motion, and this whatever material we may consider. If
we imagine the system divided into a large number of identical layers of thickness
χy and apply a force F on the upper layer, this force will also apply to all the layers,
displacing each at a speed χu relative to the layer just beneath it. This kind of motion
is called a simple shear (see Fig. 2.7).
In this kind of flow, the relative speed between the two solid surfaces separated
by a distance H can be expressed as the sum of the relative speeds of the H/χy pairs
of adjacent layers, whence
H
V= χu .
χy
Repeating with a series of layers with another thickness χy , we would obtain the
same result. The ratio χu/χy of the relative speed of the layers and their thickness
is thus constant and equal to V /H. This quantity is called the shear rate and we write
du
γ̇ = . (2.18)
dy
The ratio between the tangential force and the area of the layers is the tangential
or shear stress τ = F/A. It has physical dimensions of pressure and is thus given
in pascals (Pa). One would expect this variable, which expresses the resistance to
friction between the sliding layers, to depend on the relative speed of the layers, and
hence on the shear rate. We can thus define the apparent viscosity of the material by
τ
η= . (2.19)
γ̇
In practice, this type of flow can be obtained by putting the fluid between two parallel
solid walls and imposing a relative parallel motion of these two walls (see Fig. 2.7).
The gas layers close to the walls will tend to move a the same speed as the walls, and
this induces a relative motion of the different layers. The tangential force applied
to the walls is transmitted to the other layers of the material and, in the stationary
regime, one expects a uniform simple shear.
It may seem strange to have to exert a force in order to shear a gas. In fact, this can
be understood using the following picture. Two neighbouring gas layers behave like
two trains travelling in the same direction but at different speeds V1 < V2 , each train
being full of excited travellers (the molecules) running in all directions inside their
train (due to thermal agitation) and occasionally jumping from one train to the other.
Even if each train always carries the same number of travellers, some travellers will
arrive in the faster train with speed V1 and others will arrive in the slower train with
speed V2 . Under these conditions, the faster train will tend to slow down unless it is
supplied with some extra energy, and the slower train will tend to accelerate unless
some energy is removed from it. This is why a tangential force has to be applied
between the two fluid layers to hold their relative speed constant.
With the help of this mechanism, we may now calculate the viscosity of a gas
using the kinetic theory developed earlier. A detailed calculation would involve a
rather sophisticated formalism, but we shall simplify here. We represent the gas in
simple shear flow by plane layers sliding one over the other, each layer exchanging
energy with its neighbours, like the two trains in our analogy. The thickness of these
layers is of the order of the mean free path, since a distance of this order is required by
each molecule to exchange energy with another molecule. We assume that, as soon as
one molecule arrives in the neighbouring layer, it imparts its momentum to this layer
through a collision, thus neglecting the possibility of the molecule actually crossing
the layer without collision. Furthermore, we assume that the velocity distribution is
the equilibrium distribution we determined earlier, in other words, that the momentum
exchanges are instantaneous.
Consider two layers of gas of thickness λ (the mean free path) in relative motion
(see Fig. 2.8). Viewed from a frame moving with the lower layer, the latter is of
course fixed while the upper layer moves at a speed V = γ̇ λ. Each layer ‘ejects’ and
‘absorbs’ molecules all the time at a rate q which is just the number of molecules
crossing the interface per unit time. A molecule coming from the lower layer and
entering the upper one has a speed equal to its agitation speed within gas that is
macroscopically at rest, while each molecule leaving the upper layer and entering
Fig. 2.8 Momentum V

exchanges due to molecules λ
moving from one layer to the
next during simple shear
the lower one has this agitation speed plus the speed V of the layer as a whole. The
change in the momentum of the upper layer per unit time in the direction of relative
motion of the layers is thus −qmV . By a similar calculation to the one we did for
the pressure, it follows that the force exerted by the lower level on the upper level
is −qmV . We deduce that the shear stress that must be applied to the upper layer to
maintain this motion is τ = qmV /A.
We now calculate the flow q of molecules through a surface per unit time. Accord-
ing to the discussion in previous sections, we know that nu Au particles of speed u
cross the area A per unit time. The total number of molecules crossing A per unit
time is thus obtained by summing over all possible magnitudes of the velocity, which
yields nA◦u+ , where
1
◦u+ = nu u .
n
u>0
The quantity ◦u+ can be determined directly from the velocity distribution (see
Sect. 2.3.1), and in particular using (2.8), with the result
∞
1
◦u+ = uf (u) du = ◦c . (2.20)
4
0
It follows that q = nA◦c/4. Finally, the apparent viscosity τ γ̇ = qmV /Aγ̇ of the
gas is given by
2α
μ = αmnλ◦c = mkB T , (2.21)
π d2
where α is a coefficient equal to 1/4 according to this simplified calculation. If we take
into account the more complex reality of momentum exchanges within the gas, we
find α = 1/2. At room temperature, the order of magnitude of the viscosity of a gas is
10−5 Pa s. According to (2.21), we note that the viscosity of an ideal gas increases with
the temperature. This is quite the opposite of what is generally observed for liquids
(see Sect. 2.4.5). It arises because the internal friction mechanisms are directly related
to thermal agitation, which of course increases with temperature. Another remarkable
property is that, to a first approximation, according to (2.21), the viscosity of the gas
does not depend on its density, something which confirms that the physical origin of
viscous friction is essentially the agitation energy of the molecules, however many
of them there are. A final important point is that the viscosity coefficient obtained
above does not depend on the shear rate γ̇ . The ideal gas is a Newtonian fluid.1
2.3.5.3 Viscous Dissipation
It is useful to calculate the energy that must be supplied to maintain the relative
motion under simple shear as described above. Given that the motion is maintained,
we can calculate the energy per unit time, that is, the power that must be supplied to
the system. The power required to displace two neighbouring layers relative to one
another is the product of the applied force and the relative velocity, viz., τ Aγ̇ λ. The
total power needed to shear a volume of gas of thickness H is thus the sum τ Aγ̇ λ
over all the layers of thickness λ, namely, τ Aγ̇ H, or again,
P = τ γ̇ Ω, (2.22)
where Ω = AH is the volume of the sample under shear.

The power supplied here is often considered to be dissipated power and referred
to as viscous dissipation. In practice, the corresponding energy must effectively
be supplied continuously to maintain the motion despite the friction between the
layers as they slide over one another. According to the first law of thermodynamics,
this energy contributes to increasing the internal energy of the system, and hence
eventually to increasing the temperature of the gas. On the other hand, heat exchange
with the surroundings, and in particular with the solid walls, may allow the system to
remain at the same temperature. In any case, these effects are usually negligible for
gases because the shear rates encountered in practice are actually very low compared
with the kinds of speeds attained by molecules within the gas at macroscopic rest,
and which characterise the internal energy of the material. However, these effects
may nevertheless become significant for viscous liquids under high shear rates.
2.4 Liquid State
2.4.1 Transition from Gaseous State to Liquid State
2.4.1.1 Possible Existence of a Condensed State
When describing the gaseous state using the kinetic theory for ideal gases, which was
consistent with the equation of state (2.13), we assumed that the molecules could not
congregate together and that they remained on average rather far from one another.
1 In this book, we shall use μ for the viscosity when discussing Newtonian fluids, and η for the
apparent viscosity when the latter is not necessarily constant.
If volume effects or interaction forces play a significant role, this is no longer justified.
The probability of collisions between molecules increases with their density, and
when the molecules of a gas encounter one another, there is a certain probability that
they will stick together. This probability goes up as the temperature drops, since the
kinetic energy of the molecules then decreases. To get a better understanding of this
phenomenon, it is instructive to consider the changes in the energy of a system when
the molecules are brought closer together.
Consider a material made up of non-polar molecules, that is, molecules such as
oils (hydrocarbons or silicones) with no net electric charge and no permanent dipole
moment. Under these conditions, the only forces between molecules are the van der
Waals attraction and repulsive forces. In order to describe the corresponding mutual
interaction potential, we shall use the modified hard-sphere potential defined by (2.5).
We now calculate the total interaction potential energy T of a molecule with
all the other molecules of the system. To do this, we must sum the potentials, but
taking into account the spatial distribution of the surrounding particles. However, we
know that the mutual interaction potential drops off very quickly with the relative
distance between molecules. To a first approximation, we may therefore simply take
into account the molecules in the immediate neighbourhood of the chosen molecule,
that is, at a centre-to-centre distance equal to their size d. Treating the molecules
as spherical, there will be on average 4π d 3 n/3 molecules in contact with the given
molecule, whence the total potential can be written
4
T = − π d 3 nw .
3
Adding up all the potentials calculated in this way over the whole ensemble of
molecules, we would obtain a total potential with a double count of the potential
associated with each mutual interaction. The average potential per molecule is thus
given by T /2, which we write more simply using α = 2π d 3 w/3 and Ωm = 1/n,
the average volume available per molecule:
α
=− . (2.23)
Ωm
If the molecules are stacked up on top of each other as in a compact disordered pile of
grains (see Sect. 3.2), the volume fraction φ, i.e., the ratio of the volume of the mole-
cules to the total volume, is of the order of 64 %. The volume available per molecule
is therefore the volume of one molecule divided by 0.64, or Ωm = π d 3 /6 × 0.64.
The energy per molecule is then of the order of 6w, or a few times kB T . In this
situation, the total attractive potential of a molecule with its neighbours is distinctly
greater than its energy of thermal agitation. The latter will not therefore be suffi-
cient for it to break away easily from its set of neighbours. This shows that a stable
condensed state, in which all the molecules are very close to one another, is quite
feasible. We shall now investigate the conditions under which the transition to such a
2.4 Liquid State 57
state becomes possible. To do this, we shall examine the evolution of the free energy
of the system, which depends in particular on the entropy (see Appendix B).
2.4.1.2 Entropy of the System
Since we are now considering the possibility of condensed states, we may no longer
assume as we did when calculating the entropy of a gas (see Sect. 2.3.3) that the
volume of the molecules is negligible compared with the volume of the system.
We must now take into account the reduction of accessible volume within the sys-
tem that results simply from the presence of the other molecules. We shall further
assume that the second term on the right-hand side of (2.12), relating to the velocity
configurations, is not for its part significantly affected by the increase in the den-
sity of molecules or their interactions. (A more detailed investigation does in fact
demonstrate the general validity of this assumption.)
Imagine once again that the molecules are placed successively within the volume
Ω. The number of possible positions for the first molecule is still equal to the number
of volume elements, viz., Ω/ν. Now that this molecule occupies one of the volume
elements, the volume available for the second molecule is Ω − ν. However, one must
also take into account the so-called excluded volume around a molecule, due to the
fact that two molecules cannot come closer than a certain centre-to-centre distance
equal to their diameter. For a spherical molecule of volume v, it is straightforward to
show that the total excluded volume when positioning the centre of another molecule
is 8ν. For the second molecule, the number of possible positions is therefore only
(Ω − 8ν)/ν. Under these conditions, if we continue to add molecules in this way,
for the k th molecule, there will remain only
Ω − 8βk (k − 1)ν
ν
possible positions. In this expression, βk accounts for the fact that the excluded
volumes calculated in each step may overlap and hence have less impact on the
reduction of the available volume. This factor will be equal to 1 for the first few
molecules, but will then fall off as the number of added molecules increases.
Taking into account the possible permutations of the N molecules in a given
configuration, it follows that the number of configurations is
1 Ω − 8βk (k − 1)ν
N
Z∝ .
N! ν
k=1
To estimate this product, we take a kind of average value of the factors, given that
they vary between Ω and Ω − 8βN (N − 1)ν. Hence, to a first approximation, we
assume that the above expression can be rewritten in the form
N
1 Ω
Z∝ − βN , (2.24)
N! ν
introducing a kind of ‘average’ factor β. Using the available volume per molecule
Ωm /N, the entropy per molecule Sm = S/N becomes, up to a constant,
Sm = kB ln(Ωm − βν) . (2.25)
2.4.1.3 Equation of State
In order to find the equation of state of the system, we cannot consider entropy changes
alone, because the changes in available volume now have an impact on the internal
energy. We must in fact calculate the Helmholtz free energy (see Appendix B).
The total internal energy U of the system is the sum of the kinetic energies of the
elements making up the system, i.e., 3kB T /2 per molecule, and the internal potential
energy given by (2.23). The second term in the free energy can be deduced from
the expression (2.25) obtained for the energy, whence the average free energy per
molecule is
3 α
Fm = kB T − − kB T ln(Ωm − βν) . (2.26)
2 Ωm
By (B.15), viz.,
∂F ∂Fm
p=− =− ,
∂Ω T ∂Ωm T
we then obtain
α kB T
p=− + . (2.27)
Ωm2 Ωm − βν
In this relation, we may insert the total volume to obtain the most usual form of the
so-called van der Waals equation of state:
NkB T αN 2
p= − 2 . (2.28)
Ω − βNν Ω
This provides a good qualitative description of the behaviour of an ensemble of

molecules over a rather broad range of states. Let us consider some of its predic-
tions, concerning in particular the dependence of the pressure on the volume at
different temperatures. When the temperature is high enough, the second term in
(2.28) becomes negligible for all values of the volume, because the volume cannot
be made to tend to zero. In this case, the pressure therefore falls continuously as
the volume increases (see Fig. 2.9): the system variables do not give any sign of
a change of state. Note, however, that the pressure tends to infinity when the vol-
ume of the system tends to βNν, so there is no possible state for a smaller volume.
2.4 Liquid State 59
Fig. 2.9 Typical isotherms

T = const. obtained from the
p
van der Waals equation (2.28).
Upper curve high temperature.
Lower curve low temperature.
These curves can be used to
define the liquid and gaseous
states of the material (see text)
P
P'
M
p0 N M'
O
1,5Nυ Ω0 Ω
We may take it that the densest possible disordered state is reached for this value.
Now we know that the maximal stacking concentration of a disordered ensemble of
beads is 64 %. Here the concentration is the ratio of the effective volume Nν occupied
by molecules to the available volume Ω. It thus makes sense to take β = 1.5, which
leads to a divergence of the pressure for a concentration Nν/Ω ≈ 0.64.
When the temperature is high enough, there is a region between the points O and
P in the example of Fig. 2.9 where the pressure grows with increasing volume. This
region does not correspond to a stable state. Indeed, if there are slight fluctuations in
the characteristics of the system within the sample, these will quickly degenerate, i.e.,
they will increase in amplitude, thereby inducing significant non-uniformities into
the state of the system. For example, consider a point N with coordinates (p0 , Ω0 )
in this region of the curve. If the volume available per molecule is at a given time
slightly greater than Ω0 /N in part of the system, the pressure y is greater than p0 .
In the rest of the system, the volume per molecule is then on average smaller than
Ω0 /N and the pressure y is less than p0 . To try to reestablish pressure equilibrium in
the various parts of the system, the region of higher pressure will tend to expand even
further, climbing well beyond the point N on the curve, while the region of lower
pressure will tend to shrink, falling well below the point N. The fact that the slightest
local fluctuation in the variables is amplified in this way means that the system is
unstable.
The two other regions of the lower curve in Fig. 2.9, where the pressure decreases
with increasing volume, correspond to stable regimes. The stable region associated
with the first part of the curve (up to the point O) is what we shall call the liquid
state of the material. The volume available per molecule is small, of the same order
as the volume of the molecule, and hardly sensitive to changes in pressure. In this
regime, the term associated with interactions and the entropy term in (2.28) both
play an important role in the expression for the pressure. The second stable region
(beyond the point P ) is associated with the gaseous state. The volume is well above
the total volume of the molecules and varies rapidly with changes in the pressure.
In this case, the interaction term is negligible, cohesive effects are almost zero,
thermal and entropic effects dominate, and the density is low.
Given the instability of the intermediate region between the two states, there is no
way of going continuously from the liquid to the gaseous state. When the pressure
is gradually reduced from a point in the upper part of the curve, we remain in the
liquid state and the volume increases slowly. But as soon as we reach a state (for
example, at the point M) below the point P associated with the maximum pressure
of the unstable region, the system can go to the gaseous state (at the point M in our
example) situated at the same pressure on the other stable part of the curve. The exact
point at which this phase transition occurs depends on the rate at which the pressure
is reduced. If it is reduced very slowly, the transition will occur around the point P,
but if it is reduced quickly, the point O can be reached before the transition occurs.
Between the points P and O, the liquid is said to be metastable.
2.4.2 Structure
A liquid consists of a dense but disordered stack of more or less spherical molecules.
At short range, that is, considering only the nearest neighbours of a molecule, there
is an organisation close to what we see in a crystalline solid, for there is then only
a limited number of stacking possibilities. However, at longer range, the small suc-
cessive deviations from an organised structure lead eventually to a totally disordered
one.
In the liquid state, the molecules are characterised on the one hand by the fact that
they are in some sense ‘stuck together’ by the attractive van der Waals force, and
on the other, by the fact that they are subject to permanent thermal agitation, which
allows them to move around their neighbours and even ‘unstick’ themselves from
time to time.
2.4.3 Deformation
Let us consider what happens when a simple shear deformation is imposed on a

liquid. Suppose first that this deformation is imposed extremely quickly, in fact so
quickly that thermal agitation has no time to play any role. Under such conditions,
for a small deformation, a given molecule (see Fig. 2.10) will retain the same set of
neighbours, while their relative distances will increase slightly. Some neighbouring
molecules will come closer and others will move away. This implies that the relative
distance of the molecules is no longer the distance corresponding to the potential
energy minimum associated with the interaction forces (see Fig. 2.2), whence the
total interaction potential energy will rise. As in the case of a solid subject to a small
deformation (see Sect. 2.6), the stress required is proportional to the deformation and
the system returns to its initial position as soon as the stress is released. The liquid
can be considered in this regime to behave in an essentially elastic way.
2.4 Liquid State 61
t<θ t>θ
Fig. 2.10 Deformation and relaxation within a liquid subjected to a stress from time t = 0
(schematic view). Left the liquid is at rest. Centre the liquid deforms without rearrangement after a
short time. Right rearrangement occurs
This regime is only observed if the above forces are applied for extremely short
times. Indeed, under typical conditions, in contrast to a crystalline solid, a liquid
has time to rearrange itself. Indeed, the fluctuating motions of the molecules due to
thermal agitation allow them to explore a whole range of spatial configurations in a
very short space of time. Such a rearrangement is able to ‘relax’ the internal stresses
resulting from the potential energy stored up during the tiny relative displacements of
the molecules. Finally, in practice, the force required to maintain a deformation falls
to zero after a short time, whereupon the liquid recovers a structure equivalent to its
initial structure before deformation. There is therefore a characteristic time known
as the relaxation time θ beyond which we may consider that thermal agitation will
have allowed the molecules to explore the various configurations close to their initial
state. The elastic regime can only be observed over times shorter than θ . For simple
liquids, we find that θ takes values between 10−12 and 10−10 s. Consequently, these
materials only exhibit an apparent solid behaviour for extremely fast external inter-
vention, behaving like simple fluids on the time scales relevant to most experimental
situations. On the other hand, materials with relaxation times comparable with typical
observation times will have more complex rheological properties than liquids under
these same conditions. This is in fact the case for all the materials to be discussed
subsequently in this book. We shall see later on that the fast relaxation of simple liq-
uids also underlies their simple mechanical (Newtonian) behaviour, which suggests
that the non-Newtonian behaviour of complex fluids results from slow relaxation
phenomena.
2.4.4 Flow
When a large enough deformation is imposed, there is no other solution but to force
neighbouring molecules to move apart permanently. For the moment, we assume that
thermal agitation can be neglected. We shall assume that the molecules are aligned
in parallel layers and remain so. We then impose a simple shear in the direction of
one of these layers (see Fig. 2.11). This shear will induce a relative motion of the
layers in this direction, and during this motion, some molecules will begin to come
Fig. 2.11 Modelling the flow (b)

behaviour of a liquid in the Φ
absence of thermal agitation. (a)
Relative motions of a molecule
and two of its neighbours and (a) (c)
changes in the interaction
potential energy and the
γ
force required to impose this
motion. Different relative F
positions of the molecules (a,
(b)
b, c) (right) associated with
different levels of interaction
energy and force (left)
γ
(c)
closer, while others will move apart (see Fig. 2.11). This induces an increase in the
potential energy associated with short-range interactions, since the initial equilibrium
position (a) probably corresponded to a minimum potential energy. Beyond a certain
displacement associated with a maximum value, located at (b), the potential energy
begins to decrease, then returns to zero at (c), when the system recovers an analogous
configuration to the original one. We can use the gradient of the potential energy as
given by (2.2) to estimate the changes required of the applied force in order to follow
such a development (see Fig. 2.11). When the force required is negative, there is no
need to apply any force at all, for the system will return to its original configuration
of its own accord. The average force to be imposed is the integral over all positive
force values. Note that, in this argument, only the interaction potential energies are
relevant. The resulting average force does not therefore depend on the relative speed
of the layers, i.e., the shear rate. This corresponds to a (plastic) behaviour of type
τ = const. which obviously differs from the viscous behaviour we expect to find in
a liquid: the higher the shear rate, the greater the force required to maintain the flow.
This comes about because we have not taken thermal agitation into account. This
is what allows the molecules to rearrange themselves very quickly at each instant of
time. Indeed, it is not necessary to supply all the energy associated with the separation
of neighbouring molecules since their own kinetic energy can contribute significantly
here. To get a better understanding of this phenomenon, we first investigate the effect
of agitation on the structure of the liquid at macroscopic rest (without external forces,
in a non-deformable container). Due to the high concentration of the system, each
molecule is as though emprisoned in a cage formed by its neighbours. However,
the position of the cage walls fluctuates as time goes by due to the agitation of the
molecules around it, whence the local density can be reduced by rearranging the
system slightly. In this way, from time to time, a ‘hole’ appears in the cage, large
enough for the molecule to escape. If at this time it has enough kinetic energy to
2.4 Liquid State 63
overcome the attractive force of its neighbours, it can then completely escape from
its cage. When the liquid is at macroscopic rest, such movements of one molecule
in one direction or another tend to balance out.
On the other hand, when a stress is imposed on the system, it will favour motions
in one particular direction by reducing the energy needed by the molecules to get out
of their cage in that direction. In this situation, the flow is like a destabilisation of
an equilibrium situation, favouring one specific direction. The extent of this desta-
bilisation increases with the magnitude of the stress which reduces the height of the
energy barrier, and it can be expressed via the relative displacement speed of the
layers, that is, via the shear rate. In practice, Newtonian behaviour is observed for
simple liquids, as for gases:
τ = ν γ̇ , (2.29)
where ν is here the viscosity of the liquid.

Apart from the increase in the stress with the shear rate discussed above, (2.29)
also expresses the fact that a flow at constant velocity will set up instantaneously
when a given stress is applied. Given what was said above, this is valid as long as the
characteristic flow time, i.e., the time 1/γ̇ required to reach a deformation of 100 %,
is significantly longer than the relaxation time θ of the system. This is of course true
most of the time. Another feature expressed by (2.29) is the fact that the behaviour
of the material does not depend on the history of the flow. The shear rate achieved
depends only on the stress applied at the given time. These properties all arise due
to the very fast relaxation of the liquid which, thanks to the thermal agitation of
the molecules, tends to forget almost instantaneously the deformations it has just
undergone.
2.4.5 Rheophysical Model
The Eyring model is based on the qualitative principles discussed above. It assumes
that, through its interactions with its neighbours, each molecule behaves at each
instant of time as though it were in a potential well of average depth ε and trying at
regular intervals to escape from this well with the help of thermal agitation. To do so,
the molecule must have greater kinetic energy than ε. We thus seek the probability
of a molecule within the system having such an energy. We may assume that the
velocity distribution of the molecules is the same as in a gas. The desired probability
is then
P(c) dc ,
mc2 /2>ε
which, according to (2.9), is proportional to exp(−ε/kB T ). The frequency with which

a molecule leaves its well, i.e., the number of jumps made by a given molecule per
unit time, is proportional to this probability and to the frequency C of attempted
jumps:
τb3/2
Fig. 2.12 A molecule escapes from its potential well with the help of thermal agitation. Left equally
probable motions in all directions. Right asymmetrical motions resulting from modification of the
potential barrier in a specific direction due to application of a stress

ε
C exp − . (2.30)
kB T
Eyring suggested taking C to be a vibration frequency approximately equal to kB T /,

where = 6.63 × 10−34 m2 kg s−1 is Planck’s constant. Note that, since ε is the
energy needed to get the molecule out of its liquid environment, we should expect
it to correspond to the latent heat of evaporation per molecule ε (see Sect. 2.4.6).
In practice, for many liquids, the relation is rather of the form ε ≈ 0.4ε . Note also
that (2.30) can be interpreted as the reciprocal of a characteristic time required by
the system to undergo an elementary change of configuration. It is thus also the
relaxation time θ of the system, beyond which the system forgets the deformations
it has undergone (see Sect. 2.4.3).
When no stress is applied to the system, the probability of a molecule leaving
a well is equal to the probability of a molecule turning up there, which means that
the system is at rest macroscopically. Suppose now that a shear stress is applied to
the system. The jumps in the direction of the force corresponding to this stress are
no longer balanced out because it is easier to leave the well in the direction of the
force than in the opposite direction. Let b be the average distance between the centres
of two neighbouring molecules. During an elementary displacement of the typical
length required to get to the top of a well, i.e., b/2 (see Fig. 2.12), the work done on a
molecule is the product of the applied force τ b2 and the displacement, that is, τ b3 /2.
The energy barrier that must be overcome to accomplish such a jump is reduced by
this much. On the other hand, the energy barrier in the opposite direction is increased
by this same amount, whence the frequency of jumps per unit time in the direction
of the applied force, which is the difference between the frequencies of jumps in the
two directions, can now be written

kT ε τ b3 τ b3
f = exp − exp − exp − .
kB T 2kB T 2kB T
When τ b3 2kB T , this expression simplifies to first order, yielding

τ b3 ε
f ≈ exp − . (2.31)
kB T
2.4 Liquid State 65
We now consider two parallel plane layers of molecules in relative motion at a

distance b from one another, under the action of a shear stress τ in the direction of
the planes. The instantaneous motions of the various molecules in each layer are not
identical, but their average velocity is uniform and can be found from (2.31). Indeed,
the speed of one of the layers relative to the other as induced by this motion can be
written V = bf . We deduce that the shear rate is V /b = f . Finally, we obtain the
apparent viscosity of the liquid in the form
τ 2 ε
η= = exp , (2.32)
γ̇ Ωm kT
where Ωm ≈ b3 is the volume available for each molecule. The right-hand side
is independent of the shear rate so the apparent viscosity η is constant and this
model effectively predicts that a simple liquid will have Newtonian behaviour. Note,
however, that (2.32) is not necessarily valid unless the viscous energy associated
with a unit deformation is much less than the thermal energy, i.e., τ Ωm 2kB T .
This is true in most cases for liquids made up of small molecules, i.e., with diameter
of the order of a few angstroms, and at temperatures that are not too low.
Equation (2.32) agrees quite well with the observed temperature dependence of
liquid viscosities. In contrast to gases, the viscosity decreases with temperature.
In other words, a liquid is fluidified by increasing the thermal agitation. This is
because we thereby increase the frequency with which the elements jump from one
cage to another. As an example, water has viscosity 1.787 × 10−3 Pa s at 0 ∇ C and
0.295×10−3 Pa s at 100 ∇ C, a value about a hundred times greater than for a gas. The
values are of the same order of magnitude for ethyl alcohol and mercury. In contrast,
glycerol has viscosity 12 Pa s at 0 ∇ C and 1.5 Pa s at 50 ∇ C.
Note also that the pressure does not appear in the above expression for the vis-
cosity. In practice, this is usually the case. The pressure has little influence on the
viscosity of liquids because an increase in pressure would induce a slight reduction
in the intermolecular distance and hence in the interaction energy, but it would not
significantly modify the frequency of jumps determined above.
2.4.6 Interfacial Tension
Energy is required to create a liquid–gas interface. This phenomenon is due to the

cohesive (van der Waals) forces between the liquid molecules. A molecule immersed
in the liquid, hence surrounded solely by liquid molecules, has a total interaction
potential energy (cohesive energy) n0 w, resulting from its interactions with its n0
nearest neighbours. In practice, it is simpler to use the cohesive energy per unit area
wL = n0 w/s. In addition, the molecules sitting on the liquid–air interface interact
on average (over the ensemble of local arrangements) with only half as many liquid
molecules. We may neglect their interaction energy with the molecules of the gas,
since they only rarely encounter any of them. The cohesive energy of the molecules
on this interface is thus wL /2.
Note in passing that wL is the energy one must supply to separate a molecule
from all its neighbours, hence to evaporate the liquid. In fact, since each elementary
separation involves two molecules, the energy required per molecule, that is, the
latent heat of evaporation per molecule, is ε = wL /2.
When the area of the liquid–gas interface is increased, the number of molecules
situated on this interface naturally increases too, while the other molecules remain
completely immersed in the liquid. On average, each molecule initially immersed in
the liquid and arriving at this interface loses a cohesive energy wL /2, this being the
elementary work that must be done on the system in order to achieve this operation.
The total work that must be done to increase the area of the interface by an increment
dA is dW = (wL /2)dA. Defining the interfacial tension, usually known as the surface
tension for a liquid–gas interface, by σLG = wL /2, the surface energy that must be
supplied to the system becomes
dW = σLG dA . (2.33)
The surface tension of water in air at 20 ∇ C is 0.073 Pa m. It changes by less than

10 % around this value when the temperature goes from 0 to 50 ∇ C. For other liquids,
it varies between 0.02 and 0.08 Pa m. This approach can be extended to two other
arbitrary phases A and B in contact. It may then be useful to take into account the
interactions between molecules in the two phases, which alters the definition of the
interfacial tension σAB (see Sects. 3.2.3 and 6.2.1).
2.5 Solid State
2.5.1 Structures and Interactions
When we lower the temperature of a liquid, we thereby reduce the thermal agitation
and hence also the possibilities for spontaneous relative motions of the molecules.
In some cases, the structure remains disordered and we then obtain a glass (see
Sect. 2.6). But in the most common situation with a simple body, the molecules
arrange themselves into an ordered structure within which they continue to move
slightly as a result of thermal agitation, but sit on average at some fixed position. In
general, with the exception of water, the structure thereby obtained is denser than
the liquid phase and the interaction energy of each molecule is significantly greater
than in the liquid state. For a given pressure, this relatively sudden transition occurs
at a specific temperature. However, in order for the transition to come about, it must
start from a ‘seed’ that grows to take over the whole sample. In the same way, it is
impossible to obtain the regular tiling of a mosaic by randomly pushing an ensemble
of tiles around on the floor. One must start with a small set of tiles arranged according
2.5 Solid State 67
to the chosen pattern. This structure can then be made to grow with the same pattern
by successively placing the remaining tiles around the outside of this seed, an exercise
that would soon get faster with the growth of the structure. Such a seed will often
evolve close to a solid surface across which the molecules are more ordered, but in
a liquid, it can also simply form around a suspended impurity.
The many possible characteristics of the resulting crystal structures are well known
and we shall not go into the details here. For the record, the main ones are the
hexagonal close-packed (hcp) and face-centered cubic (fcc) structures, which are
the densest (74 %), with n0 = 12 nearest neighbours for each atom, and the body-
centered cubic (bcc) structure, which is less dense, with n0 = 8.
In the above description, we only consider the special case of a solid formed by
orderly arrangement of the same molecule as in the liquid phase. However, there are
crystals in which the particles are atoms or ions, while in the corresponding gas, one
finds molecules. The main kinds of interaction within a crystalline solid are:
• Simple van der Waals attractions, as in solid hydrogen, the noble gases, and alkanes.
In this case, the atoms are simply juxtaposed.
• Ionic interactions, which are stronger, as in salt crystals like NaCl. Ions of opposite
signs are arranged in such a way as to preserve charge neutrality.
• Covalent interactions, as in diamond or silica. These are giant molecules with ori-
ented bonds. Their arrangement is determined by the valence number and direc-
tions.
• In metals, atoms release their valence electrons, leaving the ions in a sea of elec-
trons. The forces between ions and electrons are key here, leading to a close
packing with strong attractions.
Under such conditions and in order to simplify the discussion, we shall hereafter
systematically use the term ‘atom’ to refer to the particles making up the basic
structure of any solid.
2.5.2 Microrheology in the Solid Regime
In the solid state, the atoms are in equilibrium positions as regards their interactions
with all the surrounding atoms. When a force is imposed on the material, the atoms
are slightly displaced from these equilibrium positions. In this way, energy is stored
in the system. When the force is removed, the atoms will naturally return to their
original equilibrium position, i.e., the deformation is reversible. This is therefore
essentially elastic behaviour.
To simplify here, we assume that each atom is in an equilibrium position with
regard to the mutual interactions with each of its z nearest neighbours. Put another
way, if the mutual interaction potential is , this will have a relative minimum at a dis-
tance b equal to the separation between neighbouring atoms, whence (r = b) = 0.
When a force is applied to the solid, its atoms are slightly displaced relative to one
another and the distance between the two atoms considered above is now r such that
|r − b| b. The interaction potential thus becomes
1
(r) = (b) + (r − b)2 (b) + O (r − b)3 . (2.34)
2
The force associated with this potential is
F = (r) ≈ (b − r) (b) . (2.35)
An arbitrary deformation of the material will stretch or shorten the separation between
the atoms by an amount r − b that is proportional to this deformation. The constant
of proportionality will depend only on the crystal structure and characteristics of the
deformation. As a consequence, according to (2.35), the force required to impose a
deformation will be proportional to this deformation. The total stress to be applied,
equal to the sum of forces of this kind with different coefficients of proportionality,
will also be proportional to the deformation. The material is thus linearly elastic
in the limit of small deformations. In the following, we shall focus on the relation-
ship between local physical characteristics and macroscopic properties of solids for
specific simple deformations.
2.5.3 Elongation
Consider a solid cylinder with cross-sectional area A and length l to which a force
F is applied at each end of the cylinder axis. The cylinder will then extend by an
amount χl. The deformation or strain is defined as the relative elongation ε = χl/l
in the principal direction. The normal stress in the principal direction is the ratio
σ = F/A of the force to the cross-sectional area of the cylinder. Since we know that,
for a small deformation, σ is proportional to ε, we define the Young’s modulus of
the material by
σ
E= . (2.36)
ε
When the material is deformed in this way along a specific axis, it will also deform
in the plane perpendicular to this axis. Indeed, the radius of the cylinder will change
from R to R + χR. If its volume is conserved, we have π R2 l = π(R + χR)2 (l + χl).
For small deformations ε 1, it follows that χR/R = −χl/2l = −ε/2.
For convenience, we shall assume that, in this crystalline solid, the atoms are
arranged in planes parallel to the cylinder axis and lined up in parallel lines a distance b
apart in cross-sections perpendicular to the cylinder axis (see Fig. 2.13). There are l/b
atoms along the cylinder axis and, as long as the cylinder deforms uniformly, no cross-
section is favoured. Consequently, each atom is affected equally by the extension
and thus moves a distance x relative to its neighbours, where (l/b)x = χl. In the
perpendicular plane, the atoms move a distance y such that (R/b)y = χR, whence
y = x/2. There are 1/b2 atoms per square metre in a cross-section perpendicular to
the cylinder axis and the force applied to each atom is f = Fb2 /A.
Consider now the volume element bounded by two surface elements, as shown
in Fig. 2.13 and separated by a distance b. When this volume is deformed by ε as
2.5 Solid State 69
b
b
Fig. 2.13 Change in position of the atoms relative to a central atom (grey) under elongation. In this
case, the atoms move closer together along the vertical axis and further apart in the perpendicular
plane
described above, the particles are brought together in the direction of the x axis and
moved apart in radial directions. The energy needed to do this is thus equal to the
sum of the energy stored by bringing two atoms closer together (a central atom and
four quarter atoms in the corners of the surface element) by a distance x, and by
moving four atoms further away. According to (2.34), the total potential is to first
order
x2 y2
Ψ = 6(b) + 2 (b) + 4 (b) .
2 2
We thus deduce the magnitude of the force in the x direction to be
f = Ψ (x) = 3εb (b) .
Given the area 2b2 over which it applies, the Young’s modulus is equal to
3 (b)
E= . (2.37)
2 b
2.5.4 Behaviour Under Simple Shear
Here we consider a simple shear inducing a deformation γ . For small deformations

γ 1, we know that the shear stress τ is proportional to the deformation, so we
define the shear modulus by
τ
G= . (2.38)
γ
Fig. 2.14 Displacement of (a)

atoms distributed in parallel
planes under a simple shear in b
the direction of one of these
planes. a Initial configuration.
b Configuration after a small
displacement y
(b) y
r- b
r+
Suppose, for example, that the atoms are arranged in parallel planes and shifted
through an angle π/4 (see Fig. 2.14a). The structure is made up of atoms arranged
in a similar way in planes parallel to the one shown in Fig. 2.14, these planes being a
distance b apart. This is the same structure as the one discussed
√ in Sect. 2.5.3. Since
the distance between two layers in√ the shear plane is b/ 2, the deformation induces
a relative displacement x = γ b/ 2 of the molecules in the principal direction. The
distance between two neighbouring molecules thus changes from r = b to

b2 b2 (1 ± γ )2
r± = + ,
2 2
where the plus and minus signs correspond to atoms moving apart or coming closer,
respectively (see Fig. 2.14b). For a small deformation γ 1, we then obtain r± ≈
b(1 ± γ /2). The macroscopic shear stress is the same√ as the shear stress obtained by
each atom by the area 2b×b associated with it within
dividing the force f applied to √
its layer, which gives τ = f / 2b2 . If we only take into account the interactions of
the atom with its neighbours in the plane of observation, then according to (2.34),
the total interaction potential energy becomes
1
Ψ (x) = (r+ ) + (r− ) = 2(b) + x 2 (b) .
2
The magnitude of the force that must be applied to each molecule is then
f = Ψ (x) = x (b) ,
whereupon the shear modulus is
1 (b)
G= . (2.39)
2 b
2.5 Solid State 71
We thus find that the shear modulus is 1/3 of the Young’s modulus. This result, shown
here for a specific crystal structure, turns out to be quite general, applying to any
incompressible solid material.
2.5.5 Compressibility
In fact, this approach is not completely general because, when a material is deformed,
it may succeed in minimising the energy supplied by collapsing in on itself to some
extent. In particular, this implies that the material can be compressed when a uniform
force, or pressure, is applied to it. This effect is described by introducing the uniform
compression modulus, defined as the ratio between the imposed pressure and the
relative reduction in volume ω = χΩ/Ω:
p
K= . (2.40)
ω
Let us assume that the pressure causes all the atoms to approach one another by
the same distance x. The force exerted on each atom is pb2 and the work done is
pb2 x. The energy stored by a similar displacement of all the n0 neighbours of the
given atom and the interactions between the atoms is n0 (1/2)x 2 (b). The energy
associated with the volume Ωm ≈ b3 available around an atom is half of this since
each interaction occurs in the volume associated with each atom. It follows that
p = n0 x (b)/4b2 . In addition, the volume occupied by an atom can be written in
the form Ω ∝ 4π r 3 /3, where α is a coefficient depending on the atomic arrangement.
Hence, χΩ/Ω = dΩm /Ωm = 3x/b and finally,
n0 (b)
K= . (2.41)
12 b
Note the similarity between the expressions for K and E obtained by the microscopic
approach. In fact, there are quite general relations between these two parameters and
they can be deduced using a macroscopic approach based on linear relations between
stresses and strains (deformations):
E E
K= , G= , (2.42)
3(1 − 2ν) 2(1 + ν)
where the Poisson coefficient ν introduces a correction that takes compressibility into
account. For an incompressible material, we have ν = 1/2, which implies that K is
not defined since no compression is possible, and G = E/3. This is the situation for
elastomers and most fluids. However, when the Poisson coefficient differs from 1/2,
this means that the material can expand or contract. For most solids, ν lies between
1/4 and 1/3, implying a reduction in volume during elongation.
F
Plastic regime P Ductile
material
Y
F Brittle
Elastic Fracture material
regime
Δε
Ο O' ε
Fig. 2.15 Relation between force and deformation within the two main types of solid, namely,
ductile and brittle
2.5.6 Maximal Mechanical Strength
The behaviour of a solid under significant deformation is not generally linear. One
reason is that changes in the interaction potentials are more complex for large relative
motions of the atoms. Another is that, for imperfect crystals, some bonds may actually
break. It is not easy to treat this regime in a completely general way. When a solid
is deformed still further, e.g., under traction, two kinds of behaviour are observed in
practice (see Fig. 2.15):
• Ductile Materials. These can be quite significantly deformed without fracture,
the deformation increasing with the applied stress. For small deformations (OY
in Fig. 2.15), the behaviour is linear elastic to begin with. The deformation is
reversible if the stress is removed, but beyond a critical deformation (associated
with the point Y), further deformation is in part irreversible. This is the ductile
(or plastic) regime. If the stress is removed, for example beyond the point P, the
system does not return along the curve PO, but instead the deformation decreases
along the curve PO . If the stress is once again increased starting from the point
O , the system climbs back along the same curve, indicating that the behaviour is
indeed elastic in this regime, but that the material has undergone an irreversible
(plastic) deformation χε.
• Brittle Materials. These deform elastically up to a critical value of the deforma-
tion (associated with point F in Fig. 2.15), or equivalently, a critical value of the
stress beyond which the material fractures, i.e., it separates into two pieces whose
elements no longer interact as they did in the uniform material. This behaviour is
represented by the dashed horizontal straight line in Fig. 2.15.
It is interesting to estimate the critical stress corresponding to the transition to plastic
deformation for a ductile material or corresponding to fracture in the case of a brittle
2.5 Solid State 73
material, since this critical stress represents the maximum resistance of a material to
deformation. Even though the two phenomena look different macroscopically, one
would expect them to occur for similar reasons, precisely when we induce a structural
modification which takes it beyond the point of no return.
2.5.6.1 Ductile Solid
Consider the situation depicted in Fig. 2.14. A simple shear is imposed on a material
made up of atoms distributed in plane layers parallel to the shear direction. As we
have seen, the deformation causes some atoms to come closer together and others to
move further apart, and the interaction potential energy increases. This phenomenon
continues until the displaced atom reaches the level of the neighbouring atom in the
lower layer and continues on its way toward a position equivalent to its initial position,
but situated between the two atoms of the lower layer. In this last step, the potential
energy drops back down to its minimal value again. There is therefore a critical
deformation γc , of the order of 1/4 in our example, beyond which it is no longer
necessary to apply a force to maintain the deformation. The structure subsequently
evolves on its own toward a new configuration associated with the shift between the
two layers. If we maintain the force required to achieve this critical deformation, we
can thus in principle displace one layer of atoms indefinitely relative to the other layer,
by a succession of jumps like the one just described. The corresponding critical stress,
which is in fact associated with the point of inflection of the potential energy curve,
can be estimated roughly if we assume that the material has the same shear modulus
(determined in the linear regime) right up to the critical deformation, whence
τc ≈ Gγc . (2.43)
In practice, it turns out that this seriously overestimates the actual value, by a factor
of the order of 100 or 1,000. The above analysis can be corrected by taking into
account the decrease in the shear modulus when the deformation increases, but this
cannot reduce the value of the critical stress as much as necessary. We must therefore
look for another explanation for this discrepancy, namely, localised weak points in
the material called dislocations which facilitate collective movements of the atoms.
These dislocations take the form of atomic planes partially inserted between two
layers. For a shear in a direction perpendicular to these planes, a much smaller stress
than the one needed for extraction from a potential well is sufficient for the inserted
plane to slide along and position itself opposite some other plane. Moreover, the
resulting lateral shift is nevertheless of order b, which means that small stresses
can generate significant deformations. However, the adaptation of this argument to
explain the full deformation of the material is a complex matter that goes beyond the
scope of this book.
Fig. 2.16 Irregularity or

‘hole’ at the surface of a solid, σ
leading to the formation of a
fracture under traction
ρ
l
2.5.6.2 Brittle Solid
For a brittle material, fracture occurs when two atomic layers literally come apart.
In this case, the simplest thing to consider is the effect of traction. Once again, the
applied force must first increase with distance, but then drops off rapidly beyond the
inflection point of the potential. Assuming as before that Young’s modulus remains
constant over this broad range of deformations, we find that the critical stress is of the
order of σc = γc E, with γc of the order of 1/4. Taking into account the dependence
of Young’s modulus on the deformation, we would obtain a slightly smaller critical
stress, but nevertheless much greater than the actual value, by a factor of 10–100.
At this point, one must consider local weak points within the material in order to
explain this result.
It is irregularities on the outer surface of the solid that lead to these weak points.
For example, if there is a small hole at the surface of the solid (see Fig. 2.16), the
local stress near the bottom of the hole is much higher than the macroscopic stress.
A detailed calculation assuming the material to be linearly elastic shows that, for a
hole of radius r and depth l, the stress at the bottom of the hole is
1/2
l
s=σ , (2.44)
r
where σ is the stress applied to the sample. Since the ratio l/r is generally large, this
can lead locally to a very high stress, close to the theoretical value expected from
the above estimates, and hence capable of generating a fracture that subsequently
propagates through the material. And all this while the macroscopic stress remains
rather low.
2.5 Solid State 75
2.5.7 Solid–Liquid Transition
As the temperature of a solid is increased, so also is the agitation of its constituent

molecules. The molecules at the surface are in the shallowest potential wells since
they are bound to fewer molecules than their counterparts within the material, so
it is naturally these molecules that are first to leave the solid state. This happens
at a slightly lower temperature than the temperature referred to as the solid–liquid
equilibrium temperature. A liquid layer thus forms at the free surface of the solid. At
the solid–liquid equilibrium temperature, the liquid phase moves gradually through
the material as more heat is supplied to the system. By thus increasing the temperature,
we increase the amplitude of agitation of the molecules about their equilibrium
position in the solid state, until this amplitude is such that the ordered structure can
no longer maintain itself. For a crystal, this happens when the amplitude of agitation
reaches about 20 % of the distance between the closest molecules within the structure.
2.5.8 Solid–Gas Transition
The latent heat of sublimation, that is, the energy required to vaporise unit mass
of solid material, can be related to a first approximation to the cohesive energy
wS = n0 w between the atoms. The calculation is similar to the one for the latent
heat of evaporation (see Sect. 2.4.6). The energy needed per molecule to completely
separate all the molecules from one another, that is, the sublimation energy, is
1
LS = wS . (2.45)
2
For materials like neon, argon, and krypton, this calculation gives a value very close
to the actual value. For other materials, things are more complicated. For example, for
ionic solids, sublimation preserves the interaction between certain atoms, e.g., Na and
Cl, and for metals, the interactions between electrons must be taken into account.
It turns out that LS usually lies between 1/3 and 1/6 of the interaction energy wS
between the basic elements of the structure.
It is interesting to note that the latent heat of fusion is roughly 1/10 of the latent heat
of sublimation, which suggests that fusion leads to a slight decrease in the number
of bonds. As it happens, there is no precise physical justification for this result. We
may just say that, qualitatively, a certain energy is needed to provoke the molecules
to the point where the system liquefies, but much more energy is needed to remove
them completely.
2.6 Glassy State
2.6.1 Glasses
Most mineral elements form liquids with rather low viscosities when they melt.
Conversely, when the temperature is lowered, these liquids crystallise rapidly once
below the melting point, thereby solidifying, and this even if the cooling rate is very
fast. However, there are materials which give liquids with relatively high viscosi-
ties when they melt, of the order of 104 –106 Pa s. When such materials are cooled
quickly enough, any crystallisation can be totally avoided. The viscosity of the liq-
uid increases steadily to reach such values that the material may be considered as
a solid. Such materials are glasses (or amorphous materials) and the phenomenon
leading to this type of material is called the glass transition. Various materials have a
glassy phase, including oxides such as SiO2 and Na2 O, sulfides, phosphorus, organic
molecules like toluene, methanol, glucose, or sucrose, polymers (see Chap. 3), and
metallic glasses if the cooling is fast enough.
The structure of a glass is similar to that of a liquid in that the molecules or atoms
are very close to one another and there is no long-range order. On the other hand,
as in a solid, the atoms or molecules are not free to move very much relative to one
another, their displacements being limited to the tiny motions around their average
positions resulting from thermal agitation. In terms of its internal structure, one may
thus view a glass as a rigidified liquid.
2.6.2 Glass Transition
Experimentally, this transition can be monitored via changes in the volume of the
material when the temperature is varied at constant pressure. When a liquid is cooled,
its volume first decreases steadily, following branch A corresponding to liquid behav-
iour. When the crystallisation temperature TS is reached, the volume drops suddenly
and the system evolves along branch B corresponding to solid behaviour. In some
cases, the liquid can nevertheless be cooled below TS without it crystallising, either
because it has been cooled very quickly, or because its molecular characteristics pre-
vent it from crystallising. In this situation, the liquid evolves along branch A. Then
at a certain temperature Tg1 , there is a sudden change in the slope of the curve due
to a change in the behaviour of the system. This is the glass transition temperature.
In fact, for a given liquid, this temperature is not unique, but depends on the cooling
rate. For example, cooling the system more slowly, the transition will occur at a tem-
perature Tg2 < Tg1 . Note that, if we carry out experiments with slower and slower
cooling rates, the glass curve ends up coinciding with the crystal curve. This occurs
at a critical temperature Tk called the Kauzmann temperature, which is the lowest
glass transition temperature than can be reached for the given system.
2.6 Glassy State 77
Volume
or entropy
Liquid
(A)
Glass (1)
Glass (2)
Crystal
(B)
Tk Tg2 Tg1 TS Temperature
Fig. 2.17 Glass transition. Temperature dependence of the entropy or volume for different types
of material or different cooling rates
In practice, the glass transition is usually studied by monitoring changes in the

specific heat. The glass transition is then characterised by a drop in the specific heat.
Since the specific heat at constant pressure is related to the entropy by

∂S
cp = T ,
∂T p
the temperature dependence of the entropy of the system can be found by integration.
This dependence is qualitatively similar to that of the volume (see Fig. 2.17).
It is thus observed that the entropy of a glass remains finite even when the temper-
ature tends to zero. Glasses therefore have a residual configurational entropy which
reflects the level of disorder. It is also observed that the entropy of a glass is not a
simple thermodynamic state function, since it turns out that it depends on the tem-
perature and pressure history of the sample. This means that, in a glassy state, the
material is no longer able to explore all possible microscopic states, and one speaks
of ergodicity breaking.
2.6.3 Mechanical Behaviour Associated with the Glass Transition
Regarding the question of mechanical behaviour, a glass cannot simply be considered

as an extremely viscous liquid. Indeed, under ordinary observation conditions, glasses
have certain properties that are commonly found in solids, and in particular a nonzero
elastic modulus. This type of behaviour has already been mentioned for liquids, but
over extremely short time scales (see Sect. 2.4.3). When considering glasses, one
Fig. 2.18 Viscoelastic γ

behaviour. Two phases in
the time dependence of the
deformation, for fixed stress
γc
η
θ t
must thus envisage some kind of intermediate form of matter that can behave under
ordinary conditions either like a solid or like a liquid, depending on the circumstances,
that is, depending on the boundary conditions or conditions of observation. In a
physically consistent description, they can conveniently be considered to exhibit
viscoelastic behaviour.
When subjected to a stress τ from some initial time (and assuming that it was at
rest up until then), a viscoelastic material reacts to begin with like an elastic solid. The
resulting deformation is finite, increasing with the applied stress, and the initial shape
is recovered when the stress is removed. However, if the applied force is maintained
beyond some characteristic time (which we shall soon find to be the relaxation time
θ ), the material will deform more easily and in fact the deformation will end up
increasing linearly in time (see Fig. 2.18), i.e., the material will begin to flow like a
liquid. If the solid regime is roughly characterised by a constant elastic modulus G
and the liquid regime by a Newtonian viscosity μ, the transition between the two
regimes occurs somewhere near the intersection between the deformation plateau
γ = τ/G of the solid regime and the straight line γ = γ̇ t = τ t/μ corresponding
to flow in the liquid regime. It follows that this solid–liquid transition occurs after a
time equal to a characteristic time θ such that τ/G = τ θ/μ, i.e.,
μ
θ= . (2.46)
G
In fact, this is the behaviour of a liquid such as we have described it qualitatively,
with a very short characteristic time. Extending the analogy, this suggests that the
relaxation time θ is the characteristic time for spontaneous reorganisation of the
material through thermal agitation.
The above relation is particularly interesting because it shows that, if the elastic
modulus does not change too much with temperature, and this seems realistic enough
since the thermal agitation is not directly involved in the behaviour of the system in
its solid regime, the viscosity of the material must exhibit the same variations as the
relaxation time under changes of temperature.
2.6 Glassy State 79
It should nevertheless be borne in mind that this is a very rough description of

the reality since it does not take into account a range of features observed in glasses.
For example, it seems that several relaxation times must be taken into account to
describe the time dependence of a deformation. In addition, fracture phenomena and
localisation phenomena suggesting plastic behaviour have been observed.
2.6.4 Viscosity of Glasses
Given a reduction in thermal agitation, one must expect the relaxation time of the
system, and hence also its viscosity, to increase when the temperature is lowered. In
practice, it turns out that the relaxation time increases suddenly above the temperature
Tg . In fact, relaxation phenomena in glasses can rarely be described in terms of a
single relaxation time. As a consequence, the simplified description above is not
strictly valid. In reality, the time required to reach equilibrium, corresponding to
the longest relaxation time, is significantly longer than the experimental time scale,
which means that we are dealing with a non-equilibrium system. However, in the
largely qualitative discussion below, we shall assume that the system has only one
relaxation time.
Empirically, it is found that the viscosity of glasses, when represented in a plot
of log η versus Tg /T , follows a curve that seems to diverge at Tg . In this particular
context, Tg is arbitrarily defined as the temperature at which the viscosity reaches a
value of 1013 Pa s. In this same kind of plot, different glassy materials do not give the
same curves. For the strongest materials, typically composed of tetrahedral lattices
like SiO2 , the curve can be quite well represented by a straight line, whence the
model used to describe them has the form
Ea
μ = μ0 exp . (2.47)
RT
For brittle materials, usually ionic or molecular liquids, the curve has a slope that
increases steadily and tends to a vertical asymptote when Tg /T → 1. One model used
to describe this behaviour is the so-called Vogel–Fulcher–Tammann–Hesse model:
B
μ = μ0 exp . (2.48)
T − T0
The existing rheophysical description of glasses (free volume, cooperative motion,

and mode-coupling models) is still largely based on rather speculative or relatively
technical theories which go beyond the scope of this book.
Further Reading
1. Cabane, B., Hénon, S.: Liquides—Solutions, dispersions, émulsions, gels. Belin,

Paris (2003)
2. Cottrell, A.H.: The Mechanical Properties of Matter. Wiley, New York (1964)
3. Guinier, A.: The Structure of Matter: From the Blue Sky to Liquid Crystals.
Butterworth-Heinemann Ltd, Oxford (1984)
4. Israelachvili, J.: Intermolecular and Surface Forces, 2nd edn. Academic Press,
London (1991)
5. Jancovici, B.: Statistical Physics and Thermodynamics. McGraw-Hill, New York
(1973)
6. Jones, R.A.L.: Soft Condensed Matter. Oxford University Press, Oxford (2002)
7. Loeb, L.B.: Kinetic Theory of Gases. McGraw-Hill, New York (1927)
8. Tabor, D.: Gases, Liquids and Solids, 3rd edn. Cambridge University Press, Cam-
bridge (1991)
9. Zarzycki, J.: Glasses and the Vitreous State. Cambridge University Press, Cam-
bridge (1991)
Chapter 3
Suspensions
Abstract In this chapter, we consider the simplest possible situation, namely, (rigid)
solid particles that are much larger than the constituent elements (atoms or molecules)
of the liquid in which they are immersed. A mixture of such particles and liquid is
able to flow by virtue of the flow of interstitial liquid and the relative motions of
the particles. As long as the particles are not in direct contact with one another,
viscous energy dissipation is of hydrodynamic origin, which is to say, related to the
flow of interstitial liquid. We begin by discussing the mechanical characteristics of
a homogeneous and stable suspension of particles in a Newtonian liquid. We then
review the specific effects of particle concentration, the orientation of anisotropic
particles, a non-uniform spatial distribution of particles, and indeed the structure
of this distribution. Finally, we discuss the case of particle suspensions in yield
stress fluids.
3.1 Introduction
As discussed in the last chapter, simple liquids made up of identical small molecules
are Newtonian. In nature as in industry, a great many materials are produced by adding
to a simple liquid something whose basic elements are bigger than the molecules of
the liquid. The various chapters of this book are devoted to the main categories of this
type of material, each associated with a certain type of element, namely, polymers,
colloids, droplets, bubbles, and grains. The behaviour of such materials depends on
the mutual interactions of the elements and their interactions with the molecules of
the liquid, as well as the deformability of these elements, and any collective structures
that may be induced by their presence in the liquid. In this first chapter devoted to
such materials, we shall be concerned with the simplest situation, where the particles
are solid (not deformable) and much larger than the constitutive elements (atoms or
molecules) of the liquid in which they are immersed.

DOI: 10.1007/978-3-319-06148-1_3,
82 3 Suspensions
Interface
Fig. 3.1 Microscopic view of a suspension. A particle is surrounded by a large number of liquid
molecules (left) which form a continuous medium from the mechanical point of view as far as
the particle is concerned (right). The interactions between the two phases occur mainly between
molecules located on either side of the interface
When a solid particle is immersed in a liquid, this gives rise to specific interactions
between the molecules of the liquid and those of the solid particles. In general, ionic,
valence, or hydrogen bonds will not form between the molecules of the solid and the
liquid. The interactions between solid molecules and liquid molecules will thus be
mainly van der Waals over the interface between the two phases. There may also be
long-range interactions between the solid particles, known as colloidal interactions,
resulting from van der Waals forces or from the presence of various other species in
solution or adsorbed on the surface of the particles, such as ions or polymers. Finally,
collisions between the liquid molecules and solid particles due to thermal agitation
may induce spontaneous agitation of the particles (Brownian motion). In this chapter,
we shall assume that the solid particles are large enough to justify neglecting colloidal
interactions and Brownian motion. These phenomena and their consequences for the
behaviour of the system will be the subject of Chap. 5. There may be other types of
interaction between particles due to direct contacts when the particles are very close
to one another. This will be discussed in Chap. 7 on granular materials.
Since the liquid molecules are much smaller than the particles (see Fig. 3.1), the
liquid looks like a continuous medium on the scale of the particle, and the mechanical
properties of this medium can be assumed identical to those it would have had if the
particles had not been there. Since any space between two particles is occupied by
liquid, we speak of the interstitial liquid. A mixture of particles and a liquid flows
thanks to the flow of the interstitial liquid and the relative motions of the particles.
When the particles do not come into direct contact with one another, viscous energy
dissipation is solely of hydrodynamic origin, i.e., it relates solely to the flow of
interstitial liquid.
Up until now (see Chap. 2), we have been concerned with pure bodies made up
of large numbers of a single type of molecule in a certain volume. But with suspen-
sions, we move on to a new type of material comprising several different types of
elements. In this context, we have so far described the characteristics of the material
at the local level, on the scale of the particle. Now we must describe the mate-
rial on the macroscopic scale, i.e., on the scale of the sample considered by the
experimenter. We thus begin by discussing the main features of a homogeneous and
stable suspension of particles in a Newtonian liquid (see Sect. 3.2). We shall see how
3.1 Introduction 83
Ω
ΩL ΩS
ω0
ω1
A
Fig. 3.2 Suspension of spherical objects (centre) in a liquid (left). The resulting mixture (right) is
not homogeneous on the local scale. For example, the density varies significantly from one small
volume ω0 to another identical one located somewhere else. Beyond a certain observation scale,
e.g., a volume ω1 , if the suspension is homogeneous, the density is roughly constant, whatever
region of the sample is considered for such a volume
the presence of particles can influence the behaviour of the mixture (see Sect. 3.3).
We then review specific effects of the particle concentration (see Sect. 3.4), the ori-
entation of anisotropic particles (see Sect. 3.5), and any non-uniformity in the spatial
distribution of the particles (see Sect. 3.6), but also effects due to the structure of
such a distribution (see Sect. 3.7). Finally, we discuss suspensions of particles in a
yield stress fluid (see Sect. 3.8).
3.2 Preparing a Suspension

3.2.1 Geometry
The first step in making a suspension is to place an object in a volume of liquid. Such a
system cannot be treated as homogeneous, that is, as having the same properties at all
points, since it comprises two contrasting regions, viz., a liquid volume surrounding
a solid volume. A large enough number of particles must therefore be mixed into the
liquid volume to obtain a homogeneous suspension (see Fig. 3.2). Furthermore, these
particles must be well spread out through the available space. And finally, since the
system is never homogeneous on the scale of the individual particle, we must specify
a minimum volume for observing the properties of the system when we need to
demonstrate its homogeneity. We shall call this the representative elementary volume
or representative volume element. The material can then be treated as homogeneous
if it has the same characteristics from one such volume element to another.
More precisely, in mechanics, one would like to have a continuous medium, in the
sense that its physical characteristics vary continuously in space and time. This notion
of continuity is somewhat delicate to define quantitatively, but the representative
volume element plays a crucial role. It is a volume element on the scale of which
we can begin to make measurements relevant to some given property. We thus study
the spatial and temporal variations of the relevant physical characteristics on a scale
greater than this volume. It follows that the dimensions of the sample must be large
compared with those of the representative volume element in order to justify the
continuity hypothesis. Since each volume element contains at least one particle, the
dimensions of the sample must be distinctly greater than those of the particles.
84 3 Suspensions
Volume
ω0 ω1 Ω
Fig. 3.3 Measuring the average value of a variable q over a certain volume. When the volume is
small, this value will fluctuate significantly, but above a large enough volume, it reaches a plateau.
For even larger volumes, it may depend on macroscopic variations of q in the sample. The points
ω0 , ω1 , and ξ correspond to the volumes indicated in Fig. 3.2
Let us consider for example the variable density α and calculate it for larger
and larger volumes (see Fig. 3.2). Starting with a much smaller volume than the
volume of a particle, α varies significantly, in fact as much as the difference in
density between the two phases, when the given volume is increased (see Fig. 3.3) and
depending on the chosen initial position, because this volume is sometimes located
entirely within the liquid, sometimes entirely within the solid, and then encompasses
differing volumes of both phases. But as the volume is increased to the point where
it encompasses many particles, α tends to level out at an average value of the density
(see Fig. 3.3). A representative elementary volume is thus a volume for which we
come close enough to this plateau.
We may extend these considerations to other variables like stress, temperature,
velocity, and so on, but in some cases it is more difficult to specify conditions jus-
tifying the continuity hypothesis. Indeed, for materials with complex (nonlinear)
behaviour, the spatial variations of some variables conditioned by the behaviour of
the material may also prove to be highly nonlinear. A typical example is provided by
the yield stress fluid, which only flows beyond some critical stress. For slow flows
of this kind of material in ducts with diameters much greater than the dimensions of
the constitutive elements of the material, the solution of the flow problem based on
the assumptions of continuum mechanics (see Appendix A) can predict a localised
shear in a very thin layer, with a thickness of the order of the elements making up the
material. This dimension is smaller than the size of the representative volume ele-
ment, whence the continuous medium assumption is not justified. As a consequence,
it is difficult in practice to make a strict statement of sufficient conditions for this
hypothesis and its validity must be justified a posteriori.
3.2.2 Volume Fraction
The primary physical characteristic of a suspension is the amount of particles placed

in a given volume of liquid (see Fig. 3.2), as specified by the solid volume fraction
3.2 Preparing a Suspension 85
β. This is equal to the ratio of the total volume ξS occupied by the particles to the
total volume ξ = ξL + ξS of the suspension sample, i.e., the sum of the volumes
of the solid phase and the liquid phase:
ξS
β= . (3.1)
ξ
In some fields, it is usual to characterise the amount of suspended solid using other
variables such as the density, or again the ratio of the solid mass to the liquid mass.
Since these variables depend on both β and the densities of the various components,
they only characterise the ratio of the volumes occupied by the different phases in a
relative way and in a specific situation where the components of the material have
roughly constant densities. As we shall see later on, the mechanical behaviour of a
stable suspension depends solely on the volume and shape of the elements making
it up, and not on their density. As a consequence, the key variable characterising a
suspension in any area of rheophysics is the solid volume fraction.
3.2.3 Energy Involved in Creating the Suspension
When solid particles are placed in a liquid, certain interfaces are modified and this
changes the potential energy of interaction of the system. As discussed in Chap. 2,
the molecules in a liquid or solid phase have an interaction energy that is essentially
due to interactions with their nearest neighbours. These interactions can be described
in terms of the cohesive energy required to separate to infinity two parts of the same
material initially in contact over a unit area. We denote this by wL for the liquid
and wS for the solid (see Sect. 2.4.6). When a solid phase is placed in contact with
a liquid phase, the two phases have a new interaction energy over the interface. The
interaction energy per unit area wLS between the two phases, once again as required to
move them infinitely far apart, is then called the adhesion energy. The relevant solid
or liquid phases are usually in contact with a gas, but the interaction energies with
gas phases are generally negligible compared with those associated with interactions
over liquid or solid interfaces.
Consider now an ensemble of solid particles initially in the open air and such
that the total area of the air–solid interface is A (see Fig. 3.2). When these particles
are immersed in a liquid, one must first create openings with total area A. This
involves separating the liquid over a total area A/2, because when two molecules
are separated, this generates a total interface equal to twice the cross-sectional area
of the molecules (see Fig. 3.4). The energy that must be supplied to achieve this
is AwL /2. In a second step, the solid particles must be placed in contact with the
liquid over an interface of area A, which requires an energy −AwLS . In addition, one
would expect the area of the interface between the liquid and its container to increase
slightly during this operation, but this change in area is negligible if the volume of
the suspension is large compared with the volume of a particle, which is generally
the case. Finally, the total energy required to set up this suspension becomes
86 3 Suspensions
(a) (b) (c)
+wL /2 -wLS
Fig. 3.4 Energy requirements to immerse a solid particle in a liquid. The process is split into three
steps. a Initial state in which the particle with unit surface area and the liquid are separate. b Creation
of an opening with half unit area. c The solid particle is placed in contact with the interface created
by the opening in the liquid
w ⎡
L
∂W = A − wLS . (3.2)
2
The suspension is easy to establish if ∂W is negative and much more difficult if
∂W is positive since energy must then be supplied to set the two phases in contact.
A simple experiment can establish the sign of wL /2 − wLS , which itself determines
the sign of ∂W . A small volume of liquid deposited on a plane solid surface takes
the form of a truncated sphere. The angle θ between the plane tangent to the sphere
and the solid surface, along the line of contact between the gas, the solid, and the
liquid, is determined by the cohesive and adhesive energies of the various phases
according to the so-called Young–Dupré relation which follows from the balance of
forces along the triple line (where the three phases meet):
1
wLS = wL (1 + cos θ ) .
2
This angle characterises the wetting of the solid by the liquid (see Fig. 3.5): good
wetting is associated with a low value of θ , in which case the droplet spreads out
significantly, while poor wetting is associated with a value of θ greater than ϕ/2, in
which case the droplet retains a rather compact form, thereby reducing the area of
the liquid–solid interface. We thus arrive at a new expression for the energy required
to get the particles into a suspension:
1
∂W = − AwL cos θ . (3.3)
2
We see that contact between the solid elements and the liquid will be favourable
when the contact angle is less than ϕ/2, since a negative energy must be supplied,
and unfavourable when θ > ϕ/2.
We can write the total surface area of the solid particles in the form A = N s,
where N is the number of particles and s the average surface area of the particles. The
total volume of the solid particles can also be written in the form ξS = βξ = N v,
where v is the average volume of the particles. The total surface area of the solid
particles can thus also be written in the form
θ π−θ
Fig. 3.5 Central cross-section of a droplet with different contact angles. Left Good wetting. Right
Poor wetting
βξ
A= , (3.4)
d
where d = v/s is a characteristic dimension of the particles. For example, in the
case of identical spheres of radius R, we have d = R/3, and for particles in the form
of square blocks of side a and thickness b, we have d = v/s ◦ b/2. According
to (3.4), for a given solid concentration, the absolute value of ∂W [see (3.2)] thus
increases as the reciprocal of a characteristic dimension of the particles. So the ease
or difficulty in getting particles into a suspension increases depending on whether
the contact angle is greater than or smaller than ϕ/2, respectively, or as the particle
size decreases.
3.2.4 Dispersing the Particles
When they have just been placed in a container with the liquid, the particles will
not immediately be uniformly dispersed. A simple shear or elongational flow are not
sufficient to disperse the particles correctly because they tend to remain on distinct
trajectories associated with their initial positions. In order to disperse the particles
in the best possible way throughout the liquid and obtain a uniform suspension, one
must generate a complex flow known as mixing. The trajectories must cross if one is
to bring regions with few particles into regions with many particles. This corresponds
to the qualitative idea we have of mixing. The standard approach is to displace some
object through the liquid in a looping motion, but going beyond these qualitative
considerations, there is no simple enough way to present the principles of mixing in
the present discussion.
3.2.5 How Many Particles can be Put in Suspension?
The answer to this question must of course be phrased in terms of the solid volume
fraction. We are thus looking for the maximum packing fraction βm , that is, the
maximal value of β that can be reached for a given type of solid particle. This
maximum packing concentration is also a useful point of reference at much lower
concentrations because the particles are already likely to get in each other’s way, and
it will be convenient to express this hindrance effect by a function of the ‘distance’
between β and βm .
88 3 Suspensions
(a) (b) (c)
φ=φ c φ=φ m φ=φ m
Fig. 3.6 Distribution of disks in a square using different procedures. a Successively, one after the
other and at random until it is no longer possible to continue without moving the others. b By piling
them up naturally under the effects of gravity, until they cannot move any further. c By distributing
them optimally in a crystal arrangement. Analogous results are obtained by distributing spheres
in a volume, but the resulting critical concentrations (mentioned in the text) are lower than in two
dimensions
The presence of interstitial liquid is irrelevant to the problem, which can be formu-
lated as follows: what is the maximum concentration of particles that can be placed in
a given volume? For the moment, we shall assume that all the characteristic lengths
of the given volume ξT are much greater than the particle size so that the exact shape
of the volume plays no role. The first and simplest approach would be to try to fill
the space with particles until it was no longer possible to add another. For example,
in the case of spheres, a perfect crystal arrangement such as hexagonal close-packed
or face-centered cubic leads to βm = 74 % (see Fig. 3.6c). This is the optimal filling.
On the other hand, if each ball is placed successively at random in the volume left
free by the previous balls, the space will be filled in a way that is far from optimal,
leading to a fraction of only a few tens of percent (see Fig. 3.6a). So by placing the
balls at random, the space is not filled efficiently.
During the preparation of a suspension, the various configurations are explored
in search of an optimal configuration by some mechanical action such as mixing,
vibration, stacking, the aim being to bring the particles closer together so that they
can better occupy the available space. We may thus consider the situation where balls
are simply stacked up in a container under the effects of gravity, e.g., by pouring them
in from another container. The value of βm thus obtained is then of the order of 55 %,
but this depends to some extent on how the powder is poured in (see Fig. 3.6b). More
generally, the value obtained depends significantly on the history of the preceding
flow. For example, if the container is shaken sufficiently after pouring the mixture in,
the particles will arrange themselves still better, yielding maximum packing fractions
of the order of 64 %.
We have talked here about the maximum packing value under the effects of gravity,
but what is the physical significance of such a value? It corresponds to the formation of
a network of direct contacts between grains that is able to support its own weight. This
is interesting because it gives us a physical criterion for the maximum concentration:
if we extend to an arbitrary suspension without the action of gravity, we can imagine
another critical concentration βc such that a continuous network of contacts forms
for the first time across the whole sample. Once this network has formed, we may
expect the mechanical behaviour of the suspension to change significantly compared
with what is observed before it comes into being. However, even though this physical
idea seems promising from the rheological standpoint, we nevertheless find that, as
for βm , the associated critical value is not unique. In fact, it depends to a large extent
on the configuration of the particles, which itself depends on the history of the flow
through the preparation of the material and subsequent processing.
Note that the maximum packing also depends on the granulometric distribution of
the particles. Quite generally, the broader this distribution, i.e., the broader the range
of particle sizes, the greater the value of βm . This is easy enough to understand. If we
first place the largest particles in the volume, we can add the next smaller particles
in the spaces between them, still smaller ones in the tiny remaining spaces, and so
on and so forth. Taking the argument to the extreme and assuming a granulometric
distribution extending to infinitely small particles, we would then be able to fill the
whole space with solid particles. In practice, of course, this would not be possible,
but even with a reasonably broad granulometric distribution, of the kind found in
concretes and natural muds (particle sizes from less than the micron up to the cen-
timetre or metre), it is soon possible to obtain solid fractions of the order of 95 %.
Obviously, βc also increases with the spread of the granulometric distribution.
3.2.6 Resistance of the Liquid to Particle Displacement
Since the preparation and flow of a suspension induce relative motions of the solid
particles and the carrier liquid, it is useful to begin with a much simpler question,
namely, what happens when a single particle is made to move through a macroscop-
ically stationary liquid, i.e., one in which the average displacements of the liquid
elements relative to the container are zero (see Fig. 3.7). Even if the liquid remains
globally at rest in the container, the displacement of the particle will induce defor-
mations of the liquid around this object: the liquid exerts a viscous resistance or drag
that tends to slow the motion of the object. This force depends on the viscosity of
the liquid, the shape and size of the particle, and the relative velocity of the motion.
We assume that the particle remains far from the walls of the container so that the
presence of these walls plays no role in the flow around the particle. The motions
induced within the liquid by this movement are more complex than those resulting
from a simple shear. Consider a particle moving through the liquid under the action
of a constant force F. The resulting motion of the liquid can be described by its
velocity distribution (or field), i.e., the value of the velocity at each point of space.
After the initial (inertial) phase when the particle is set in motion, it will move at a
certain constant speed V . The applied force must then be exactly equal to the viscous
resistance. We thus expect the velocity field around the particle to remain constant
in a frame moving with the particle. This velocity field is associated, through the
constitutive law of the material, with a certain stress field described by the (see
Appendix A). In a stationary state, the sum of the forces on the object is zero, implying
90 3 Suspensions
Fig. 3.7 Displacement of an

object (here a sphere) at speed
V through a macroscopically μ
stationary liquid. The flow of
the liquid around the object
induces a viscous resistance F R F
that the external force (usually related to gravity) exerted on the particle to maintain
its motion exactly balances the stresses exerted by the liquid over the outer surface
A of the solid object:
⎢
F= π · n ds ,
A
where n is the unit normal vector to the small surface element ds.
In the above expression, the stress tensor can be replaced by the sum of the pressure
and the stress deviator tensor T [see (A.12)]. This gives a first force term associated
strictly with the pressure in the fluid and generally independent of the speed. This is
in fact the buoyancy, i.e., the thrust due to Archimedes’ principle, discussed further
in the next section. This adds to the external force exerted on the object. The second
term, i.e., the viscous drag force, is strictly related to the viscous friction on the
object. For a Newtonian fluid, it has the form
⎢
F= 2μD · n ds ,
A
where D is the strain rate tensor which describes the local shear of the fluid in all
directions.
Suppose now that a force λ F is imposed on the particle, where λ is some arbitrary
factor. Consider the velocity field obtained by multiplying all the local velocities of
the above velocity field by the factor λ. Then all the shear rates (ratios of a speed and
a length) are multiplied by the same factor, since the geometry of the system has not
changed. Under such conditions, all the local stresses, which are proportional to the
local shear rate for a Newtonian fluid [see (A.10)], are multiplied by the factor λ. This
is true in particular for the liquid located along the solid–liquid interface A, whence
the force exerted on the object is effectively λ F. As we know that there is only one
solution for the flow characteristics of the fluid around the particle, we deduce that
this solution, in terms of velocity and stress fields, is the unique solution. As the local
velocities over the particle surface have been multiplied by λ, the particle velocity
for this velocity field is λV . Since this calculation is valid whatever the value of λ,
we may conclude that F is proportional to V .
Since the stresses within the liquid are proportional to its viscosity, we may show
by similar arguments that the force is proportional to the viscosity. If we now increase
all the dimensions of the particle by a factor λ without changing its shape, the solution
for the velocity field is of course obtained by multiplying all the velocities in the initial
velocity field and all the distances by this same factor. In this case, the local shear
rates at equivalent points under homogeneous dilation do not change. Likewise, the
stresses at equivalent points do not change. However, the total force exerted on the
particle and computed using the above integral increases as λ 2 , like the area of
the solid object. Since the speed has simultaneously increased by a factor λ, we
deduce that the force also contains a velocity-independent factor proportional to a
characteristic dimension d of the particle. The only thing we have not taken into
account so far is the shape of the particle. For the same characteristic dimension, the
viscous drag will depend on a further, dimensionless factor k called a form factor.
Taking into account all the changes discussed above, we deduce1 that the drag force
resisting the motion of the particle can be written
FD = kμV d. (3.5)
In order to compute k, one must solve the equations of motion giving the velocity
field (see Appendix A). From this we may deduce the stress field and use it to
calculate in particular the total force due to viscous friction exerted on the object.
For a sphere moving through a Newtonian fluid, such a calculation leads to k = 6ϕ
(with d = R). For a disk of negligible thickness and radius R moving along its axis,
we find k = 5.1ϕ . For the same disk moving along one of its diameters, we find
k = 3.4ϕ (still with d = R). Finally, for a fibre of characteristic thickness b and
length L, we have k = 4aϕ/ log(2L/b), where d = L and a = 1 for motion in a
direction perpendicular to the principal axis of the fibre and a = 1/2 for motion in
the direction of the fibre axis.
Naturally, since the surrounding liquid would be set in motion by rotation of
the object, a drag torque would in this case result from viscous friction. By similar
arguments to those put forward for translational motion, it can be shown that this
torque will have the form
1 A more direct mathematical demonstration is obtained by noting that the velocity can be expressed
in the form u = V u+ and the length terms in the form x = dx+ . The area terms can thus be
written s = d 2 s + and the shear rates (components of D) Ω̇i j = (V /d)Ω̇i+j . The drag force is then
⎣
F = 2μV d A+ D+ · n ds + . In this expression, the integral is computed in terms of dimensionless
variables x , u+ , etc., and so depends only on the shape of the object. When one of the system
+
parameters is modified, e.g., the speed or size of the object, we obtain a solution to the problem which
is the unique solution by using the solution in terms of dimensionless variables and multiplying the
lengths or velocities by the appropriate factor.
92 3 Suspensions
M = k d 3 μξ , (3.6)
where k is a form factor for the object. For example, for a sphere, we have k = 8ϕ ,
where d = R.
3.2.7 Stability
The density αS of the particles generally differs from the density αL of the liquid.
One may naturally wonder whether they could remain suspended in the liquid. To
find out, we must examine the balance of forces on a particle immersed in a liquid.
To begin with, there is the force of gravity αS ξg, which is simply proportional to
the volume of the particle and directed downwards. Then there is the force due to
the contact between the solid surface and the liquid. When the system is at rest, the
liquid molecules interact with the solid surface in a way that does not pick out any
particular direction, as for a gas in contact with a wall (see Sect. 2.3). The force per
unit area exerted by the liquid on the solid is thus a pressure p and the total force
resulting from this kind of contact is given here by
⎢
F=− pn ds .
A
This force is not zero because the pressure is not generally uniform in a liquid at rest.
In fact, it varies with the depth owing to the weight of fluid situated above.
Since this force would be the same in the liquid in the vicinity of the object,
whatever kind of object it may be, we can calculate it by imagining the object to be
replaced by a volume of liquid of exactly the same shape. Consider a large container
in which the liquid is at rest and imagine a cylinder of liquid of cross-sectional area A
extending between vertical coordinates y and y + dy (see Fig. 3.8). We now consider
the balance of forces on this volume. We have:
• the vertical force −αL g Ady due to gravity,
• the pressure force on the vertical walls of the cylinder which has zero resultant
since each local component is balanced by the one on the symmetrical point at the
opposite side of the cylinder, and
• the forces p(y + dy)A and − p(y)A, respectively, resulting from the pressure on
the lower face, which tends to push the liquid upwards, and the pressure on the
upper face, which tends to push it downwards.
The balance of forces thus gives d p/dy = αL g. This equation can be integrated
between the depth y and the free surface where the pressure is just the atmospheric
pressure p0 . We thereby obtain the pressure as a function of the depth in the form
p(y) = p0 + αL gy . (3.7)
Fig. 3.8 Pressure distribution p0

on the outer surface of a liquid
or solid cylinder immersed in y
a liquid
p(y)
p A p
p p
p(y+dy)
This is known as a hydrostatic pressure distribution. Such a distribution will generally

be found in liquids at rest in open containers, but also in free-surface flows for which
the liquid depth varies slowly.
A cylindrical particle of length h and cross-sectional area A immersed in a liquid
is thus subjected to the force of gravity −αS gξ and the resultant of the pressure
forces on the upper and lower faces, viz., αL gh A, which can also be written αL gξ.
The latter result has only been shown above for a simple case but is generally true
for a particle of arbitrary shape and volume ξ. The total force is thus
F = (αL − αS )gξ . (3.8)
The ordinary gravitational force is thus reduced by a force equal to the weight of
liquid displaced when the particle is immersed. This is the buoyancy force, always
opposed to gravity, and (3.8) expresses Archimedes’ principle.
For this reason, a particle whose density differs from the density of the liquid
cannot remain suspended in this liquid. For sufficiently slow speeds, we may assume
that, if the particle is moving vertically through the liquid, the surface forces exerted
by the liquid on the solid are a superposition of pressure forces and the drag force
given by (3.5) which slows the particle down in all its motions relative to the liq-
uid (when the latter is macroscopically at rest). Furthermore, we assume that the
hydrostatic distribution of the pressure is maintained around the particle despite
the movement of the liquid. Under these conditions, for free fall through a liquid,
the balance of forces is expressed by
(αL − αS )gξ + kμVfall d = 0 .
The steady displacement speed is thus given by
∂αgξ
Vfall = , (3.9)
kμd
where ∂α = αS − αL . For a spherical particle, we obtain Vfall = 2g R 2 ∂α/9μ.

94 3 Suspensions
The speed of fall is directed downwards when the density of the particle is greater
than the density of the liquid, i.e., when ∂α > 0, as happens in the vast majority of
solid particle suspensions. We then speak of sedimentation. It is directed upwards
when the particle has lower density than the liquid, i.e., when ∂α < 0, as happens in
bubble suspensions, i.e., foams, and in certain emulsions and suspensions, in which
case we speak of creaming.
Sedimentation and creaming are actively sought in certain applications, but in
general such phenomena tend to denature the material owing to the heterogeneities
they induce. In practice, it is thus important to know whether a suspension sediments
(or creams) to any significant extent during an experiment. For example, we may
consider this to be the case if the time required to sediment (or cream) through a
distance equal to 10 % of the characteristic height H of the sample is less than the
duration ∂t of the experiment, for this would mean that more than 10 % of the sample
close to the upper surface (the lower surface) would no longer contain any particles
by the end of the experiment. Under these conditions, the maximum duration of an
experiment before sedimentation becomes significant is of the order of
0.1kμH d
∂tc = . (3.10)
gξ|∂α|
For example, for polystyrene beads of radius 100 µm and density 1.05 in oil of
viscosity 100 times the viscosity of water in a container of depth 10 cm, we find
∂tc = 15 min.
These considerations provide a good description for isolated particles in a liquid.
But as soon as two particles come within a distance of the order of twice their diameter,
they interact ‘hydrodynamically’, i.e., the velocity field of the liquid around either of
them is significantly different from what one would have if it were alone in the liquid.
The drag force can then differ considerably, and since this effect is not the same for
each of the particles, which never occupy precisely symmetrical relative positions,
such a pair can evolve in a much more complex manner, e.g., with alignment effects,
mutual drag, etc. For high particle concentrations, greater than about βm /2, there are
many interactions involving more than two particles, resulting in even more complex
effects. Quite generally, as long as the material remains homogeneous, one expects the
overall sedimentation (of many particles) to be slowed down when the concentration
increases, because viscous dissipation increases owing to the fact that the liquid must
follow ever more tortuous paths to get past the particles, while the force exerted on
each particle remains roughly constant (if we assume that the pressure distribution
remains close to a hydrostatic distribution). Finally, using dimensional arguments
like those given above for a single object alone in a liquid, it can be shown that the
speed of fall of the suspended objects has the form
Vfall (β) = λ(β)Vfall (0) , (3.11)

F
V
H
.
γapp
Fig. 3.9 Simple shear of a suspension containing particles of arbitrary size and shape. The average
velocity field shown on the right is not the same as the local velocity field because the flow induced
in the liquid by the relative motions of the planes is perturbed by the presence of particles
where λ is a factor depending on the particle concentration, equal to 1 when there is

only one particle. At low concentrations, for a disordered suspension, theory predicts
that λ = 1 − 6.55β. There are also empirical relations predicting the sedimentation
speed as a function of the concentration, e.g., λ = (1 − β)n , where n = 5.1, which
provides a very rough description of the sedimentation speed up to high concentra-
tions [1].
However, many perturbing effects are likely to occur when the suspension is no
longer dilute, including preferential flow of the liquid over the walls of the container,
collective motion of groups of nearby particles, and so on. Finally, note that in
practice, the sedimentation speed in a non-dilute suspension depends heavily on the
boundary conditions, i.e., on the size and shape of the container.
3.3 Effect of Particles on the Behaviour of the Mixture
Consider a suspension contained between two parallel solid planes in relative motion
in a direction parallel to the planes. We neglect the boundaries of the planes, which
amounts to assuming that they are infinite. The presence of particles in the liquid
means that a simple shear cannot be obtained locally, so the flow cannot take the
form of a relative sliding of plane liquid layers (see Fig. 3.9). We may nevertheless
estimate the apparent viscosity of the system by computing the ratio of the averages
of the shear stress and the shear rate. In practice, if the continuous medium hypothesis
is valid, these averages are the apparent stress νapp , defined as the tangential force
applied to one of the two planes per unit area, and the apparent shear rate Ω̇app , defined
as the ratio of the relative speed of the two planes and the distance between them.
The apparent viscosity μ of the suspension is then equal to νapp /Ω̇app .
To establish the dependence of the apparent viscosity on the system parameters,
we can adopt a similar approach to the one used to examine the displacement of a
single particle in a liquid (see Sect. 3.2.6). Applying a force F to one of the planes, a
velocity field is induced within the liquid. This field will depend on the shape and size
distribution of the particles (see Fig. 3.9), and it will produce a relative displacement
96 3 Suspensions
of the walls at speed V . If a different force λ F is now applied, we obtain the solution
to the flow problem by multiplying each stress and each velocity by the factor λ,
giving the relative speed of the walls as λV . If we increase the viscosity, we obtain the
same velocity field by increasing all the local stresses in the same proportion. Finally,
we deduce that the shear stress is proportional to the relative speed of the walls, and
hence to the apparent shear rate, and to the viscosity of the interstitial liquid. If the
shape and size distribution of the particles is fixed, we thus find μ ∞ μ0 .
Furthermore, for an analogous system in which all length scales, and in particular
all particle sizes, are multiplied by a factor λ, the initial velocity field multiplied by λ
solves the problem with a stress field that is identical at all equivalent points under a
homogeneous dilation. The relative speed of the walls is now λV and the separation
between the walls is λ H , whereupon the apparent shear rate remains unchanged.
Since the shear stress is unchanged, the apparent viscosity of the suspension is the
same. We thus deduce a fundamental property of all suspensions: if we can justify the
continuous medium hypothesis, the viscosity will be independent of the size scale
of the suspended entities.
The viscosity of a suspension therefore only depends on the size distribution,
shapes, and relative positions of the particles. Assuming that the particles all have
the same shape, these characteristics can be expressed in terms of:
• the distribution of the volumes of each of the N particles relative to their average
volume, viz., {ξi /ξ}i=1,...,N , where ξ is the average volume of the particles,
• the distribution of their positions xi and orientations θi , viz., {xi , θi }i=1,...,N , and
• the volume fraction occupied by the particles in the mixture, viz., N ξ/ξT , which
is simply the volume fraction β.
The apparent viscosity of the suspension then takes the form

μ = μ0 F {ξi /ξ}i=1,...,N ; {xi , θi }i=1,...,N ; β . (3.12)
Since the variables describing the particle volume distributions are fixed for a given
suspension, we deduce the following fundamental result: a suspension is Newtonian,
i.e., it has constant viscosity, if and only if the distribution of positions and orienta-
tions of the particles does not depend on the flow history. In fact, we have obtained
this result under specific boundary conditions, namely, simple shear, but if the suspen-
sion is to be Newtonian, the spatial distribution of the particles and their orientations
must be independent of the boundary conditions, i.e., independent of the type of flow
imposed upon it. This is only possible if this distribution is isotropic and remains
so [2].
In the general case, the main task in the rheology of suspensions is to calculate F as
a function of the characteristics of the solid particles and any changes in their positions
and orientations as time goes by. Given the broad range of possible characteristics of
these particles, there is of course no direct general relation of this type. In order to get a
good grasp on the problem, we shall review a series of situations focusing on specific
features of the relative distribution of the particles: volume fraction and distribution
of sizes, shapes, and orientations. In each case, we shall consider a situation allowing
3.3 Effect of Particles on the Behaviour of the Mixture 97
us to isolate a single aspect, i.e., in which the potential effects of the other aspects
is negligible. We shall see in certain cases how to identify the cumulative impact of
several such aspects.
3.4 Effect of Concentration
3.4.1 General Considerations
To begin with, we shall consider the simplest case of a suspension of spheres of the
same size. In this case, ξi /ξ = 1 and the orientation of the particles plays no role,
whence F depends only on β and the spatial distribution of the particles. If the latter
does not vary on average, the suspension is Newtonian and its apparent viscosity is
a function of the volume fraction of the particles.
Since the velocity field within such a suspension remains complex under simple
shear, we shall estimate the viscosity of the suspension by assuming that the solid
particles are distributed in such a way that the resulting velocity field is much simpler.
In fact, we replace the spheres by a solid layer of thickness h parallel to the planes,
centered on the median plane at y = H/2 (see Fig. 3.10). Then by symmetry, the flow
in the two liquid regions between each plane and the solid inclusion is a simple shear
with shear rate Ω̇ . On the other hand, there is no shear in the solid layer. Assuming
that there is no slip along the various solid walls, the velocity is continuous at the
interfaces and the x component of the velocity field is thus given by
⎧
⎪ H −h
⎪
⎪ Ω̇ y, 0<y< ,
⎪
⎪ 2
⎪
⎨ H −h H −h H +h
vx = Ω̇ , <y< ,
⎪
⎪ 2 2 2
⎪
⎪ H −h H +h H +h
⎪
⎪
⎩ Ω̇ + Ω̇ y − = Ω̇ (y − h) , <y<H.
2 2 2
Given this velocity field, we must have Ω̇ = V /(H − h) if the boundary condition
vx (H ) = V is to be respected, and this implies that the shear stress over the solid
planes is μ0 V /(H − h). The apparent viscosity of the mixture is thus
νapp 1
μapp = = μ0 . (3.13)
Ω̇app 1 − h/H
Equation (3.13) already picks out several important features that remain true in a
more complete analysis:
• The effect of a solid inclusion is to increase the apparent viscosity of the mixture,
and increasingly as the solid fraction is larger.
98 3 Suspensions
Fig. 3.10 Effect of solid

particles on the behaviour of
a suspension (upper). It is H φ
assumed that the particles are
contained at the maximum
packing fraction βm in a layer
of thickness h (centre). The
velocity profile (lower) is not
linear (dotted line), but is
piecewise linear (continuous h φm
line), due to the fact that the
central layer remains rigid
.
γ
• When the volume of solid tends to occupy all the available volume (h/H → 1),
the viscosity of the mixture tends to infinity.
We can now get an approximate idea of the viscosity of a suspension of spherical
particles as a function of the solid volume fraction using the above approach and
assuming that the particles are all located at the centre of the flow to form a dense
cluster of grains at the maximum packing fraction βm . Assuming that the liquid is
blocked by this granular packing, i.e., that it cannot flow through the porous structure
formed in this way, this granular layer can be treated as a solid block of thickness
h. This thickness is such that the total volume of particles is conserved, whence
βm h A = β H A and h = H β/βm . Under these conditions, the above discussion
leading to (3.13) gives here

β −1
μapp = μ0 1 − . (3.14)
βm
This shows that the viscosity of a suspension of objects tends to infinity when the
volume fraction of these objects tends to the maximal packing fraction βm .
The above calculations have shown us the main trends and helped us to understand
why they come about. We shall now focus more precisely on the impact of the particles
on the viscosity of a suspension. However, to achieve this, we shall have to accept
certain results that can only be obtained by relatively complex computation.
3.4 Effect of Concentration 99
Dilute Semi-dilute Concentrated Compact

φ<<1 0.01<<φ<φc φc<φ<φm φ−>φm
Fig. 3.11 Internal structure of a suspension in the main concentration regimes
3.4.2 Concentration Regimes
By successively adding fluid inclusions of the same size, we pass through three
concentration regimes associated with different types of interactions between the
inclusions.
3.4.2.1 Dilute Regime
When the solid volume fraction is not too great, i.e., when β is at most of the order of
a few percent, the solid particles will remain on average far enough apart to ensure
that the perturbations induced by the presence of one particle will have a negligible
effect on the velocity field in the vicinity of its neighbours (see Fig. 3.11). We may
then consider that the particles do not interact hydrodynamically and everything
happens as though each particle were alone in the liquid. This is possible because
the perturbation of the flow induced by the presence of a particle falls off with the
distance from the particle and becomes second order beyond a certain distance. If
the particles are sufficiently spaced out, these perturbations should not ‘interfere’.
3.4.2.2 Semi-dilute Regime
When the concentration of inclusions is not low enough (β ⇒ 1 %) but is neverthe-

less not too high (see below), the distance over which the liquid flow is perturbed
by the presence of an inclusion is much greater than the distance separating two
neighbouring inclusions (see Fig. 3.11). In this case, everything happens as though
the inclusions interact hydrodynamically. The velocity field can no longer be deter-
mined by a calculation with an isolated particle. The inclusions must be taken into
account as a whole in the calculation. In this regime, the motion of the particles is
strictly governed by hydrodynamic phenomena.
100 3 Suspensions
3.4.2.3 Concentrated Regime
As long as the solid fraction is not too high, the trajectories of the particle centres
are close to what would be induced by the same macroscopic flow in a fluid that is
continuous on their scale, e.g., a molecular liquid. But at high enough concentrations,
their trajectories will tend to fluctuate as a result of hydrodynamic interactions with
their neighbours and the simple presence of impenetrable solid volumes around them.
Above the critical concentration βc , if the particles follow exactly the trajectories
induced by the same macroscopic flow in a fluid that is continuous on their scale,
jamming will occur. This is due to the formation of a continuous stretch of particles
in contact with one another from one side of the sample to the other, capable of
blocking the induced motion. The critical concentration βc is in principle less than the
maximum packing fraction discussed above, since it is associated with the possibility
of jamming as different configurations are explored during the flow, whereas βm is
associated with static jamming due to a single configuration. However, the definition
of βc is no stricter than the definition of βm since its value depends in particular on
the characteristics of the given flow.
In this regime, hydrodynamic interactions no longer necessarily play the dominant
role. The behaviour of the system depends on how the particle network evolves during
the flow. Various phenomena can occur here, including dilatancy (dilation of the
apparent volume of the particle ensemble), a network of direct interparticle contacts,
or an order–disorder transition. The compact regime β = βm is the extreme case
where a change in apparent volume of the sample is required for flow to occur. We
shall consider this kind of material in the context of granular materials (see Chap. 7).
3.4.3 Dilute Suspension
In this case one can determine the velocity field around each sphere analytically using
the equations of motion and assuming that the sphere is alone in the liquid, hence
taking the boundary conditions to be those at the boundaries of the macroscopic
flow. In practice, we therefore assume a simple shear u = Ω̇ y, v = 0 = w, at an
infinite distance from the centre of the particle and continuity of the velocity at the
liquid–solid interface. Difficult calculations are required to solve the equations of
motion, but they lead eventually to the following solution for the velocity field:
u = Ω̇ y − λy + χx, v = −λx + χy, w = χz , (3.15)
where

Ω̇ R 5 5Ω̇ x y R 3 R2
λ= , r = (x 2
+ y 2
+ z )
2 1/2
, χ = −1 .
2r 5 2r 5 r2
The velocity field is of course highly perturbed just around the sphere when r ◦ R,
but this perturbation falls off rapidly with the distance from the centre of the sphere.
We may now calculate the stress field in detail using the constitutive law for a
3D Newtonian fluid. To calculate the viscosity μ of the suspension, one approach is
to compare the total viscous dissipation P(β) in a given volume of the suspension
containing a sphere with the dissipation P(0) in an equal volume of liquid alone,
undergoing the same macroscopic simple shear. For a Newtonian fluid, we know
that the viscous dissipation per unit volume can be written in the form P = μΩ̇ 2
(see Sect. 2.3.5). The dissipation increases with the solid fraction because, in a given
volume, the amplitude of the perturbation of the velocity field increases with the solid
fraction. Comparing the expressions for the dissipation at concentration β and when
there are no particles, we find that μ/μ0 = P(β)/P(0). With the above velocity
field and the resulting stress field, we obtain finally
μ = μ0 (1 + 2.5β) . (3.16)
This expression, which was first obtained by Einstein [3], is in fact only an approx-
imation, becoming more accurate as β gets smaller. The error compared with the
exact result for μ/μ0 − 1 is less than 10 % as long as β is less than 4 %.
3.4.4 Non-dilute Suspension
When the concentration is higher, it is much more difficult to calculate the exact
value of the viscosity owing to the complexity of the velocity field. This is of course
a consequence of the presence of the particles and their mutual hydrodynamic inter-
actions. However, there are various ways to estimate the viscosity of a non-dilute
suspension.
The first of these begins by noting that the way the viscosity is modified by adding
particles is the same whatever the composition of the initial mixture. Put another way,
if we start out with a suspension with concentration β and apparent viscosity μ(β)
and add a volume of particles associated with a concentration β √ 1 (see Fig. 3.12)
within the initial suspension, the final viscosity of the mixture will be μ(β)(1+2.5β ).
To reach a particular concentration, we may thus introduce successive increments,
increasing the total fraction by dβ in each step. To do this, we must be able to
take the particle fraction in the overall mixture from β to dβ. Such an increase in
the total concentration is not obtained simply by adding a fraction β to a mixture
of concentration β. Indeed, consider a total initial volume ξT of suspension at a
concentration β. If we assume that this suspension constitutes a homogeneous fluid
which we mix with a certain amount of particles, then in order to obtain a particle
fraction β , we must add a particle volume β ξT to the homogeneous fluid, where
ξT is the total final volume of the suspension. We thus have ξT = β ξT + ξT . If
we now take into account the fact that the suspending fluid is itself a suspension with
a concentration β, we can calculate the effective final solid fraction as
102 3 Suspensions
Fig. 3.12 Different ways to

view the addition of a small
amount of particles to a sus-
pension. Upper Concentration
increases from β to β + dβ.
Lower Dilute suspension at
concentration β in a homoge- φ φ +dφ
neous fluid
+φ'
βξT + β ξT
,
ξT
which we would like to be equal to β + dβ. From the two expressions, we deduce
the relation between the increment in the total concentration and the concentration
of the particles added to the mixture:
dβ
β = ,
1−β
We can then obtain the viscosity of the final suspension following an increment dβ
in the total concentration:

2.5dβ
μ(β + dβ) ◦ μ(β) 1 + .
1−β
This can be rewritten in the form

2.5dβ
d(ln μ) ◦ ,
1−β
which can be integrated from 0 to β to yield

⎢ β
μ 2.5dβ
ln ◦ , (3.17)
μ0 0 1−β
whence it follows that

μ
= (1 − β)−2.5 . (3.18)
μ0
This method is in fact based on a rather unrealistic hypothesis, because in fact, as

soon as the particle fraction reaches a certain level, newly added particles will interact
‘hydrodynamically’ with the particles already in the mixture and the effect on the
viscosity of adding a small amount of particles is greater than the simple increase
assumed above. This effect is all the more important as the initial concentration gets
higher, and becomes critical when the concentration approaches βc .
To take this effect into account, several authors have suggested modifying the
viscosity increment to the form 2.5dβ/(1 − β/βc ). This correction has no precise
physical justification, but it does lead to a correct description of one of the con-
sequences of the phenomenon described above, namely, the fact that the viscosity
diverges as we approach the critical jamming concentration, something that was com-
pletely missed in the first approach. This leads to the following much more realistic
expression for the viscosity of a non-dilute suspension:

μ β −2.5βc
= 1− . (3.19)
μ0 βc
This kind of expression, usually called the Krieger–Dougherty model [4], has the
advantage of agreeing well with theoretical predictions at low concentrations [the
Einstein model of (3.16)] and diverging when β → βc , in agreement with the trend
one would expect qualitatively. Many other empirical models sharing these two
essential features have been put forward, but they do not introduce any decisive new
physical input and they do not predict experimental results any more accurately. In
practice, there are no precise enough experiments to decide between the competing
models. Their predictions are indeed very close at low or moderate concentrations
and only differ noticeably at high concentrations, i.e., when β → βc , near the
concentrated regime for which hydrodynamic interactions are no longer necessarily
the dominant phenomenon and additional effects such as migration may occur (see
Sect. 3.6).
Note finally that it is when β → βc that the formulas are most sensitive to the
value of βc , owing to the fact that they tend asymptotically to infinity. This means
that an error of a few percent in this quantity may lead to an error of a factor of 10 in
the value of the calculated viscosity (see Fig. 3.13). Now as we saw earlier, it is in
principle difficult to determine the exact value of βc . In fact the most reliable way to
evaluate the value of βc associated with a given flow is to measure the dependence of
the viscosity of the suspension on the concentration and represent it using a Krieger–
Dougherty type of model, fitting the parameter βc .
Another advantage of this method is that it can be used with a broad granulometric
range, that is, with particles of non-uniform sizes. In this case, the maximum packing
concentration is higher than for particles of uniform size, and βc is also higher, but
the viscosity of the suspension can still be estimated using (3.19) with an appropriate
value for the critical concentration βc .
104 3 Suspensions
Fig. 3.13 Relative viscosity of a suspension according to the Krieger–Dougherty formula (3.19)
for two different values of βc : βc1 = 60 % (continuous curve), βc2 = 64 % (dashed curve)
3.5 Effect of Particle Anisotropy
We are interested here in effects that are specifically induced by the presence of
anisotropic, i.e., non-spherical, particles in suspension. Several of the parameters
characterising an anisotropic particle suspended in a liquid are likely to play a role:
the exact shape of the particle, its orientation relative to the main flow direction at a
given instant, and any spin it may have about its centre of gravity. There are many
possible shapes but we shall restrict discussion here to spheroids, i.e., flattened or
stretched spheres, fully characterised by two lengths. To begin with we shall examine
the impact of the presence of such particles on the viscosity of a dilute suspension,
and in particular its dependence on their orientation relative to the flow direction. We
shall then describe how an arbitrary particle suspended in a liquid under simple shear
tends to orientate itself in the direction of flow. This will be followed by a discussion
of the impact of this orientation on the behaviour of the suspension. Finally, we shall
investigate the probable changes in the behaviour of a suspension of anisotropic
particles when the solid fraction is increased.
3.5.1 Ideal Anisotropic Particles: Spheroids
A spheroid is an ellipsoid of revolution with axis I. This kind of object is fully specified
by two lengths a and b, where a is its maximal diameter in a plane perpendicular to
I and b is its length along this axis (see Fig. 3.14). These objects include things with
the shape of a rod when a > b (prolate), a disk when a < b (oblate), and even a fibre
3.5 Effect of Particle Anisotropy 105
θ
a
b
I I
Fig. 3.14 Spheroids. Left Prolate. Right Oblate
when we have a prolate spheroid with a ⇒ b. The aspect ratio of such an object is
defined by p = b/a.
3.5.2 Effect on Viscosity of Anisotropic Particles

with Constant Uniform Orientation
If the particle orientation is given, the viscous energy dissipation can be calculated
using a similar approach to Einstein’s for the case of a simple shear in a dilute
suspension [5]. The result has the same form as predicted by the Einstein model,
namely,
μ = μ0 (1 + γβ) , (3.20)
with a concentration factor γ which now depends on the shape and orientation of the
particles. Naturally, γ = 2.5 when p = 1, since we are then dealing with spheres
and must recover Einstein’s formula (3.16). If p = 1, we find that γ changes from
a maximum value, obtained when the direction of greatest length of the particle is
perpendicular to the flow direction, to a minimum when it is parallel (see Fig. 3.15).
This is just what one would expect: the extra viscous dissipation induced by the
presence of a particle increases as the orientation of its major axis tends to oppose
the flow.
3.5.3 Particle Rotation in a Fluid Under Simple Shear
Up until now, we have assumed that the particles have fixed orientation. However, in
reality, anisotropic particles cannot remain oriented in the same way. For example, a
106 3 Suspensions
Fig. 3.15 Maximum and minimum values of the coefficient γ as a function of the aspect ratio p.
The upper curve (continuous line) corresponds to a particle with large dimension oriented perpen-
dicular to the flow direction, while the lower curve (dashed line) corresponds to a particle with
large dimension aligned with the flow direction. For intermediate orientations, γ takes values lying
between the two extrema associated with the given value of p
rod immersed in a liquid under simple shear with its major axis lying perpendicular
to the flow direction will tend to rotate around its centre of gravity since, viewed
from its own frame, the surrounding liquid is flowing in opposite directions above
and below it (see Fig. 3.16). For analogous reasons, a sphere will rotate about an axis
through its centre and perpendicular to the shear plane: the stresses resulting from the
shear of the fluid above and below the sphere induce a couple which tends to make
the sphere spin around its centre. However, the fluid motions in front of and behind
the sphere are directed in opposite directions and tend to slow down this spinning
motion. The full solution of the equations of motion shows that this rotation occurs
at an angular speed of Ω̇ /2.
For a spheroid suspended in a liquid that is undergoing simple shear, the time
dependence of the angle θ between the axis I and the flow direction (see Fig. 3.14)
is described by the following equation (see Fig. 3.17):

1 p
tan θ = tan Ω̇ (t − t0 ) , (3.21)
p 1 + p2
where t0 is the time at which θ = 0. The spheroid thus swings around its axis with
a period given by
1 + p 2 2ϕ
P= . (3.22)
p Ω̇
When p = 1, the object is a sphere and we recover the result stated above, viz.,
θ = Ω̇ (t − t0 )/2, in other words, the sphere rotates at an angular speed of Ω̇ /2.
.
γ
Fig. 3.16 Rotation of a sphere suspended in a fluid under the flow of a simple shear with shear rate
Ω̇ , as represented in a frame moving with the centre of the sphere. Arrows represent the motion of
the fluid in the absence of the particle. The sphere ends up rotating with angular speed ξ = Ω̇ /2.
The rotation of a spheroid centered at the same position is also shown qualitatively
p =1 p =2,5 p =10 p =50
Fig. 3.17 Rotation of a (prolate) ellipsoidal particle under a simple shear flow when its principal
axis lies in the direction of flow, as observed in a frame moving with the centre of gravity of this
particle, for different values of the aspect ratio p. For clarity, the particle has been shown as though
it had the same shape in each case. Successive positions at times n P/60, where n varies from 0
to 59
When p = 1, things are slightly different because of the prefactor 1/ p in (3.21): the
rotation is irregular and the object spends more time in positions for which its axis
lies close to the flow direction than in positions where its axis is largely perpendicular
to this direction (see Fig. 3.17). The situation is the same for both oblate and prolate
spheroids, but shifted by ϕ/2, whence it is always the direction of greatest length
that spends most time close to the flow direction.
Generalising, we may deduce a very important result regarding non-spherical
objects, namely that they tend to remain along the flow direction for rather a long
time then swing round quickly to line up with the direction of flow. Quite generally,
a non-spherical object will thus tend to align itself in this direction, and those objects
with larger aspect ratios will spend on average proportionally longer like this.
108 3 Suspensions
Dilute Semi-dilute Concentrated

2
φ<<1/p 2
1/p <<φ<<1 φ≈1
Fig. 3.18 Different concentration regimes for fibre suspensions using a 2D representation. Note
that this is a projection of the image of the fibres on a plane that does not exactly reflect their relative
positions in space. In particular, in the semi-dilute regime, it is the fictitious spheres around each
particle that get entangled, while the particles themselves may not actually form a tangled network
in the way this representation might suggest
3.5.4 Effect of Concentration
Here we consider only particles with high aspect ratio. In this case, as indicated by
(3.21), we expect the particles to spend most of their time aligned with the flow
direction. The viscosity is then well described by (3.20) using the minimum value of
the factor γ. This is true for low enough concentrations (dilute regime) as long as most
of the particles are (geometrically) free with regard to their rotational motions in an
arbitrary plane. For this to be the case, the fictitious spheres described by the particles
in an arbitrary rotational motion about their centres, must form an ensemble with
much lower concentration than the maximum packing fraction of identical spheres
(see Fig. 3.18). A fictitious sphere of this kind has volume ϕ b3 /6, and if n is the
number of particles per unit volume, the dilute regime corresponds to ϕ nb3 /6 √ 1.
Since we also have β = nγ, where γ is the volume of one particle, we can rewrite this
condition in the form ϕ b3 β/6γ √ 1. For fibres γ ◦ ϕa 2 b/4 and the dilute regime
is obtained for
1
β√ 2 . (3.23)
p
As for suspensions of spheres, the mechanical behaviour of non-dilute suspensions

of non-spherical entities is poorly understood, and all the more so in that a further
factor comes into play here: the hydrodynamic interactions between particles affect
the orientation of the particles. For fibres, semi-dilute suspensions are defined as
suspensions in which the volume 1/n available per particle is less than the volume of
the sphere describe by the rotation of a particle about its centre, i.e., 1/n √ b3 , but
nevertheless much greater than the effective volume of a particle, i.e., a 2 b √ 1/n
(see Fig. 3.18). The semi-dilute regime thus corresponds to
1
√β√1. (3.24)
p2
In this regime, the particles can get in each other’s way but their motions are still to
some extent independent.
For the sake of completeness, let us mention some research results whose validity
remains to be proven. Empirical methods suggest that it should be possible to describe
the viscosity of a suspension of anisotropic particles approximately by simply assum-
ing that their anisotropy would affect the critical concentration βc . In this case, we
may make a rough estimate of the viscosity of the suspension using the Krieger–
Dougherty formula with values of βc that decrease according to βc ◦ 52 − 1.15 p
as the aspect ratio increases (for p < 30). Furthermore, sophisticated theoretical
techniques lead to the following formulas [6] for isotropic suspensions (random
orientations):

μ 4βp 2 ln(− ln β)
◦− 1+ , (3.25)
μ0 3 ln β ln β
and for suspensions of perfectly aligned fibres:
μ 4βp 2
◦ . (3.26)
μ0 3 ln(− ln β) − ln β
When the volume available for a particle is of the same order as its effective volume,
we enter the concentrated regime (see Fig. 3.18). In this case the particles form
a jammed network with complex mechanical behaviour. The complexity of this
behaviour comes about because it is hard to identify a clear physical origin for
this jammed state. Indeed, several phenomena seem likely to play a role: ‘steric’
jamming, frictional interactions between the particles, and colloidal interactions due
to their small size in certain directions.
3.6 Effect of Non-uniform Particle Concentration
In some cases an initially homogeneous suspension is observed to lose this homo-

geneity during flow. The particles migrate preferentially away from certain regions
and into others. This means that the particles do not remain on the average trajectory
described by the flow. There is on average a net displacement of the particles in
one of the directions of the plane perpendicular to the direction of the velocity. This
effect raises real practical problems because it occurs in most complex flows and can
sometimes quickly generate zones of higher concentration that flow more slowly and
are liable in the end to block the flow if they become too extensive. This problem
arises in particular in extrusion flow: the particles in a suspension tend to migrate to
the corners located just before the reduction in cross-section and accumulate there.
At constant pressure, this slows the flow down and such a blockage can then grow
until it completely jams the flow (see Fig. 3.19). The migration phenomenon has also
110 3 Suspensions
Fig. 3.19 Migration in a reduced flow section. Left to right Successive stages in the vertical flow
of a suspension with steady accumulation of particles in the regions where the flow is slowed down
in the corners just prior to the reduction of cross-section. Arrows show the effect of migration of
solid particles into the left-hand corner
been observed in simple configurations such as coaxial cylinders and capillaries,

but as yet it has not been fully and accurately quantified. The physical origin of
this phenomenon is not simple and its theoretical treatment remains in its infancy.
For present purposes we shall restrict discussion to a few pointers that should help
understanding, but without attempting to fully resolve the problem.
We now consider possible particle migration under simple shear, i.e., potential
motion in the y direction for a flow in the x direction (see Fig. 3.20). Consider first
a non-Brownian spherical particle immersed in a homogeneous fluid of the same
density which flows under simple shear far from this particle. The latter undergoes
rotation and motion in the flow direction, but there is no reason why its centre of
gravity should move in the y direction. Indeed, in a frame moving with the centre
of gravity in the x direction, the flows above and below the median plane parallel to
the shear planes are equal by symmetry (see Fig. 3.20). The viscous forces exerted
from above and below the particle are thus the same by symmetry with respect to this
plane. They therefore balance one another and induce no traverse motion. A simple
homogeneous shear does not therefore induce any migration of isolated particles.
We now consider two particles in a liquid undergoing simple homogeneous shear
(see Fig. 3.20). A particle which comes near another will follow different paths
depending on where it is situated in the fluid relative to the latter. The particles
then enter a motion which will enable them to avoid one another, or more precisely,
which will reduce energy dissipations, given that these increase as the particles come
closer together. So either they give each other a wide berth, or they skirt one another
close by if their centres are originally in proximate slip planes. Viewed from a plane
moving with the centre of gravity of the two-sphere system (see Fig. 3.20), we note
3.6 Effect of Non-uniform Particle Concentration 111
G G G G
G
.
γ
Fig. 3.20 Relative motion, observed in the frame of the centre of gravity, of two solid spheres
immersed in a liquid under simple shear
that the spheres are each immersed in an identical flow by the symmetry of the situa-
tion, and this remains so throughout the history of their motion. This means that their
approach, as they move together, will be identical up to sign with their separation,
as they move apart. Under such conditions, no migration will be induced by such
motion. After coming together and avoiding one another, the two spheres will revert
to their initial trajectories. We conclude that the presence of neighbouring particles
in a simple homogeneous shear will not induce migration.
The above considerations imply that more complex conditions are required to
induce a migration effect. Consider for example a heterogeneous simple shear, i.e.,
with a shear rate gradient in the direction perpendicular to the shear plane. In a frame
moving with the centre of gravity of the two spheres, the latter will be immersed
in regions where the shear rate is different, associated with different stress fields
with different resultant normal forces which can cause the particles to move apart or
come together. A similar situation arises when there is a viscosity gradient within
the suspension, since the stress field is then different in each zone. This also happens
for a gradient in the solid fraction, since the forces exerted on each particle due to
the relative motion of the particles within a region will then differ from the force
resulting from the relative motion of particles in the adjacent region (see Fig. 3.21).
Finally, one observes particle migration toward regions of lower shear rate, greater
viscosity, and greater solid fraction.
Migration can seriously affect the relevance of measurements, especially when
dealing with concentrated suspensions. Consider for example a layer of suspension
of thickness H , sheared with an apparent shear rate Ω̇ (see Fig. 3.22). When the
grains are distributed uniformly, the concentration is β. If the particles are gathered
in a thickness H − τ with a new concentration β , leaving a zero concentration over
a thickness τ, mass conservation implies that β = β/(1 − x), where x = τ/H .
Here we shall simplify by assuming that these concentrations are in fact reduced
concentrations, i.e., divided by the maximum packing fraction. Under these condi-
tions, the distribution of particles in space must respect the condition β < 1, which
implies that x < 1 − β. The viscosities μ(β) and μ(β ) can be estimated using the
Krieger–Dougherty formula (3.19). It follows that
112 3 Suspensions
(a) (b) (c)
Fig. 3.21 The main situations leading to a migration effect in a suspension under flow (here a
simple shear). a Concentration gradient. b Shear rate gradient. c Viscosity gradient. Vertical arrows
show the direction of the induced migration
AB
H
.
γ
ε
A B φ φ'
Fig. 3.22 Non-uniform concentration that may be induced by a migration effect in a suspension
undergoing simple shear. A Particles are distributed roughly uniformly. B Clearly non-uniform
distribution, including a zone containing no particles at all over one interface. The concentration
distribution induced in each case is represented in the diagram on the right

xnβ
μ(β ) ◦ μ(β) 1 + ,
1−β
where n = 2.5βc . Assuming that the shear stress ν is homogeneous throughout the
suspension, the difference in speed between the two solid planes can be expressed
in terms of the flows in each part, whence
ν ν
Ω̇ H = τ+ (H − τ) ,
μ0 μ(β )
which yields the apparent viscosity ν/Ω̇ of the material in the form
1 x 1−x
= + . (3.27)
μ μ0 μ(β )
The right-hand side of (3.27) is greater than 1/μ when

β −n nβ
1− >1+ ,
βc 1−β
3.6 Effect of Non-uniform Particle Concentration 113
Fig. 3.23 Dependence of μ

the viscosity on the shear
rate in a shear thickening
suspension, and probable
structures associated with the
two regimes
.
γ
which is usually the case. This shows that the apparent viscosity of the system is
lower after migration than the viscosity of the homogeneous suspension. This effect
is all the greater as the concentration is high. When β → βc , the viscosity of the
suspension tends to infinity, whence the second term on the right-hand side of (3.27)
tends to 0 and the apparent viscosity of the suspension is roughly equal to μ0 /x,
which is clearly much less than the viscosity of the homogeneous suspension.
This effect can also make the behaviour of the suspension look non-Newtonian
since in principle the value of x depends on the flow conditions, and hence in partic-
ular on Ω̇ , so that the apparent viscosity depends on the shear rate. For example, for a
very concentrated suspension, if x ∞ Ω̇ , we find that μ ∞ 1/Ω̇ , which looks like the
behaviour of a yield stress fluid, since ν = μΩ̇ → ν0 when Ω̇ → 0. Hence the yield
stress behaviour observed in certain highly concentrated non-colloidal suspensions
can in fact originate from an evolving migration phenomenon.
The migration effect also provides an explanation for a slipping phenomenon
observed on the container walls in certain cases. A fluid may appear to flow much
more quickly over a smooth solid surface than would be expected given its viscosity.
Indeed, one only needs a slight migration effect to occur close to the wall for a thin
layer of interstitial liquid to form there. Under shear, this then gives rise to most
of the overall motion. Considering what was said earlier, this effect is all the more
significant as the viscosity of the suspension is high.
3.7 Shear Thickening
A spectacular phenomenon is observed in certain suspensions of roughly spherical

particles at concentrations close to but less than βc : their viscosity is more or less
constant at low shear rates, then increases suddenly by several orders of magnitude
above a critical shear rate Ω̇c (see Fig. 3.23). This increase in the viscosity is sometimes
so great that it no longer seems possible to treat the material as a liquid beyond Ω̇c ,
since no flow is then observed. This phenomenon has been found in different kinds
of suspensions of small particles (with maximum size of the order of a few tens of
microns) for which it is generally considered that colloidal effects are negligible.
114 3 Suspensions
In any case, there is little doubt that, in the flow regime for which shear thickening
occurs, i.e., shear rates of the order of Ω̇c , colloidal interactions and the effects of
thermal agitation become negligible. In practice, this effect is easy to observe with
a suspension of corn starch (see Fig. 1.10). If mixed with water, it flows like a rather
low viscosity liquid when stirred slowly with a spoon, but it suddenly resists like a
solid if treated more roughly, e.g., by trying to push the spoon quickly through the
fluid.
Bearing in mind (3.12), such a phenomenon must necessarily result from a change
in the distribution of solid particles in the liquid. Since it can arise with spherical
particles, it cannot be due to any specific change in the orientation of the particles
in the suspension. It thus results essentially from a change in the distribution of
particle positions. However, given that shear thickening is observed in the semi-
dilute regime, we must exclude a jamming effect due to the formation of a network
of direct contacts between the particles. The only physical phenomenon that could
explain a significant increase in viscous dissipation under such conditions would
seem to be the formation of structures within which the particles are very close to
one another. This suggests that the distribution of average distances between the
particles may play an important role. This is effectively the case, because the energy
required to move a particle relative to its neighbour diverges when the two particles
are very close together.
Indeed, if we consider two objects immersed in a liquid and in relative motion that
tends to bring them together (at relative speed V ), when the distance between them
is of the same order as their size, the viscous forces exerted on each object result
from viscous friction over the whole surface of the particles, whereupon the viscous
dissipation is of the same order as would be generated by moving the particles alone
through the liquid. Since the viscous drag is of the order of μRV , the dissipated power
is in this case of the order of μRV 2 . On the other hand, when the distance between
the objects is much less than the size of the particles, the situation is different. The
liquid layer separating the particles and which is expelled from this region when they
move together flows mainly in the direction of the tangent plane and is more rapidly
sheared as h gets smaller. To induce this flow, a high pressure gradient must be
applied between the central zone and the outside of the liquid layer. The pressure in
the central zone of the liquid located between the two solid objects is therefore much
higher than outside this zone. It thus results in a normal force between the particles
which is of the order of μVR2 / h (see Sect. 8.5.1). Note that this force diverges when
the distance h between the particles tends to zero. In this situation, the dissipated
power is μV 2 R 2 / h. The ratio of the powers dissipated in the second and first cases is
therefore of the order of R/ h, which tends to infinity when the separation h between
the particles tends to zero.
The still rather qualitative picture that emerges from the above considerations is as
follows. At low enough shear rates, the spatial distribution of the particles is more or
less random and remains constant. The particles move nearer to one another during
the flow but the shear rate is low enough to ensure that the particle configuration can
evolve in such a way as to ‘relax’ these approaches. But above a certain rate, the
configuration no longer has time to relax the encounters occurring between particles,
3.7 Shear Thickening 115
and it tends to ‘freeze’ into a state where any further shear would induce many relative
motions of highly proximate particles, which would require very high stresses. The
various explanations put forward up to now [7] are based largely on the general
notions of order–disorder transition and so-called hydrocluster formation. However,
exact predictions of the way these systems evolve remain problematic.
3.8 Suspensions in a Yield Stress Fluid
Many industrial materials are composed of non-colloidal particles suspended in a

material such as an emulsion, gel, foam, or colloidal suspension, with characteristics
of concentration, interparticle interaction, etc., such that they behave as yield stress
fluids. Here we shall discuss the impact of particle inclusions on the behaviour of these
mixtures. As for the suspensions in simple liquids discussed in Sect. 3.1, we shall
only consider the case where the non-colloidal particles are large enough compared
with the elements constituting the paste to ensure that the latter can be treated as a
continuous medium on the scale of the particles.
3.8.1 Displacement of an Object Through a Yield Stress Fluid
It can be shown using scaling arguments like those in Sect. 3.2.6 that the viscous
drag force Fc on a solid object of characteristic dimension R, moving very slowly
through a yield stress fluid that is macroscopically at rest, is proportional to the yield
stress and the surface area of the object:
Fc = λ R 2 νc , (3.28)
where λ is a coefficient that depends on the shape of the object. Fc is therefore the
critical force required to set the object in motion through the fluid. In contrast to a
Newtonian fluid for which the drag force tends to zero when the object is displaced
infinitely slowly, the drag in a yield stress fluid is therefore never less than Fc . For a
sphere [8], we have λ = 14ϕ . This is a remarkable result since it shows that we cannot
estimate the critical force by assuming that the yield stress is simply attained just on
the surface of the sphere, because we would then have λ < 4ϕ , owing to the fact that
the total force is then obtained by summing the projection of the elementary forces
in the direction of motion of the sphere. The much higher value of λ shows that the
displacement of a sphere, even at very low speed, induces a significant deformation
and hence a solid–liquid transition in a volume with boundaries located at a distance
of the order of 2R from the centre of the sphere (see Fig. 3.24).
When the object is moving through the fluid at a non-negligible speed V , the
viscous drag force can be expressed as the sum of Fc and a speed-dependent term.
For example, for a fluid whose behaviour is well described by a Herschel–Bulkley
116 3 Suspensions
Fig. 3.24 Solid and liq- Solid regions

uid regions around a sphere (deformed) Liquid region
moving very slowly through a
yield stress fluid
Limit surfaces V
model, i.e., ν = νc + k Ω̇ n , numerical simulation shows that this extra term can be
written in the form λ R 2 k(V /l)n , where l is the length l = 1.35R. Therefore, the
total viscous drag force can be written in the form
n
V
F = λ R 2 νc + k ,
l
which would correspond to the force due to a simple homogeneous shear on an area
λ R 2 and over a thickness l. This observation provides a rough idea of the way in
which the fluid flows around the object. However, it should not be forgotten that
the fluid regions outside the zone that becomes liquid when the object goes by will
also undergo deformations in the solid regime. This point is important in order to
understand that the volume of paste liquefied when the object goes past is in fact
rather small. The material located outside this region simply moves out of the way
momentarily before returning to its original position once the object has moved on,
implying essentially elastic deformations.
3.8.2 Stability
The physical mechanisms coming into play when preparing this kind of material are
similar to those discussed for suspensions in Newtonian liquids. On the other hand,
the stability of the particles is quite another matter since, given that a minimum force
is needed to displace an object, particles of higher density can remain suspended in
a yield stress fluid.
Assuming as in the case of a Newtonian fluid that the pressure distribution is
hydrostatic, the force due to gravity reduced by the buoyancy from Archimedes’
principle is still given by (αS − αL )gξ. The particle thus remains in equilibrium in
the fluid as long as
3.8 Suspensions in a Yield Stress Fluid 117
∂αgξ < λ R 2 νc . (3.29)
This expression can be used to derive either a critical radius below which an object of
given density remains in equilibrium, or a critical yield stress above which an object
of given size will not move.
A suspension of particles in a yield stress fluid is therefore stable if the particles
are small enough and the yield stress high enough. However, such a suspension is not
necessarily stable under flow. Consider for example a material under simple shear
with shear rate Ω̇ in a horizontal plane. In this case, since the yield stress fluid is in its
liquid regime in the vicinity of each particle, the suspended particles will see around
them [9] a fluid with apparent viscosity
νc + k Ω̇ n
η= .
Ω̇
This viscosity is to a first approximation independent of their displacement speed V

perpendicular to the shear plane, should they be moving in that direction, provided
that V is not too great, i.e., V /R √ Ω̇ . In fact, the viscous drag is here simply
proportional to the speed, which means that particles with higher density than the
fluid will sediment out.
3.8.3 Behaviour
If we place particles in a yield stress fluid, we obtain another yield stress fluid. Indeed,
whatever the concentration of added particles, these particles will be immersed in a
continuous network of paste. To make the mixture flow, a high enough stress must
be imposed to break this network and cause it to flow. The minimum stress able to do
this will be the yield stress of the mixture. It can even be shown [10] that the resulting
yield stress fluid will behave qualitatively like the interstitial fluid, but with a higher
apparent viscosity. So a suspension of particles in a fluid that can be described by
the Herschel–Bulkley model will behave in a way that can be described by the same
kind of model with the same power law.
We can say a little more about the behaviour of such a suspension in its solid
regime. Indeed, if the interstitial fluid can be assumed to be essentially elastic, the
equations describing the evolution of the system as it deforms are formally the same
as those describing a Newtonian fluid (replacing the deformations by the shear rates).
As a consequence, the impact of the presence of particles on the elastic modulus is
analogous to the effect of a suspension on the viscosity a Newtonian fluid. We may
thus distinguish a dilute regime for which
G = G 0 (1 + 2.5β) , (3.30)
118 3 Suspensions
and a semi-dilute regime for which the elastic modulus can be described to a first
approximation by the Krieger–Dougherty formula

β −2.5βc
G = G0 1 − . (3.31)
βc
It is tempting to think that the yield stress of the mixture will depend on the concentra-
tion in the same way as the elastic modulus. In fact, the situation is more complicated
because the presence of particles has a non-negligible impact on the apparent critical
deformation of the mixture. The latter may be significantly lower than the critical
deformation Ωc of the interstitial fluid. This is easy to understand. Assume for exam-
ple that the particles are gathered together in a central layer, as in Sect. 3.4.1. The
apparent deformation can then be written in the form Ωapp = (1 − β)Ωeff , where
Ωeff is the effective deformation of the paste. In this particular case, the apparent
critical deformation is obtained when the critical deformation of the interstitial paste
is reached, i.e., at the value Ωc (β) = (1 − β)Ωc . In the general case (dispersed
particles), we should therefore expect Ωc (β)/Ωc to decrease significantly with the
concentration. If we assume that the paste is purely elastic below the yield stress,
its yield stress can be written νc = GΩc , and the yield stress of the mixture can be
estimated using the relation νc (β) = G(β)Ωc (β). We then conclude that νc (β)/νc
will increase more slowly than G(β) with the concentration, since Ωc (β) decreases
with the concentration. More detailed theoretical considerations2 lead to the formula
⎛
(1 − β)G(β)
νc (β) = νc . (3.32)
G
References
1. Richardson, J.F., Zaki, W.N.: Sedimentation and fluidisation, Part 1. Trans. Inst. Chem. Eng.
32, 35–53 (1954)
2. Batchelor, G.K.: The stress system in a suspension of force-free particles. J. Fluid Mech. 41,
545–570 (1970)
3. Einstein, A.: Eine neue bestimmung der moleküldimensionen. Ann. Physik 19, 289–306 (1906)
4. Krieger, I.M., Dougherty, T.J.: A mechanism for non-Newtonian flow in suspensions of rigid
spheres. Trans. Soc. Rheol. 3, 137–152 (1959)
5. Jeffery, G.B.: The motion of ellipsoidal particles immersed in a viscous fluid. Proc. R. Soc.
Lond. A 102, 161–179 (1922)
6. Shaqfeh, E.S.G., Fredrickson, G.H.: The hydrodynamic stress in a suspension of rods. Phys.
Fluids A 2, 7–24 (1990)
7. Wagner, N.J., Brady, J.F.: Shear thickening in colloidal dispersions. Phys. Today 62, 27–32
(2009)
8. Beris, A.N., Tsamopoulos, J.A., Armstrong, R.C., Brown, R.A.: Creeping motion of a sphere
through a Bingham plastic. J. Fluid Mech. 158, 219–244 (1985)
9. Ovarlez, G., Barral, Q., Coussot, P.: 3D jamming and flows of glassy materials. Nat. Mater. 9,
115–119 (2010)
2 See Ref. [10].

References 119
10. Chateau, X., Ovarlez, G., Luu, T.K.: Homogenization approach to the behavior of suspensions
of noncolloidal particles in yield stress fluids. J. Rheol. 52, 489–506 (2008)
Further Reading
11. Barthès-Biesel, D.: Microhydrodynamics and Complex Fluids. CRC Press, New York (2012)
12. Batchelor, G.K.: An Introduction to Fluid Dynamics. Cambridge University Press, Cambridge
(1967)
13. Guazzelli, E., Morris, J.F.: A Physical Introduction to Suspension Dynamics. Cambridge Uni-
versity Press, Cambridge (2011)
14. Guyon, E., Hulin, J.P., Petit, L.: Physical Hydrodynamics. Oxford University Press, Oxford
(2001)
15. Munson, B.R., Young, D.F., Okiishi, T.H.: Fundamentals of Fluid Mechanics, 2nd edn. Wiley,
New York (1994)
16. Ottino, J.M.: The Kinematics of Mixing: Stretching. Cambridge University Press, Cambridge,
Chaos and Transport (2004)
Chapter 4
Polymers
Abstract Here we discuss systems made of much larger molecules than simple
liquids, in fact the macromolecules known as polymers. The peculiarity of polymers
which gives them their most remarkable mechanical properties is that the conforma-
tion, that is, the geometric configuration of a given molecule, is not predetermined.
Depending on the stresses it may be subject to, such a molecule can adopt a whole
range of different conformations, from completely folded up into what is known
as a coil to a fully stretched out configuration. The size and flexibility of polymers
can lead to more complex internal structures than one finds for materials made up
of small molecules, and they have consequences for the mechanical behaviour of
mixtures that include them. As for simple materials, when the temperature rises, the
polymer chains become more mobile relative to one another. The polymer is in a
liquid state. We shall mainly be concerned here with polymers in the liquid state or
else solidified by specific interactions between the chains (crosslinked polymers or
polymer gels). Different concentration regimes can be identified depending on the
interactions between the polymer chains. We are then in a position to discuss the
mechanical behaviour of polymers in these different regimes.
4.1 Introduction
In the chapter on simple materials, we focused on matter made up of identical, rel-

atively small atoms or molecules. Owing to their small size and the electron cloud
surrounding each of their components, these elements can be treated as rigid spheres
when investigating the relationship between their structure and their mechanical
properties. In the present chapter, we shall be concerned with much bigger mole-
cules known as macromolecules. These are usually polymers, built up by joining
many copies of a certain group of atoms in a long chain. This basic unit is called a
monomer or repeat unit (see Fig. 4.1). The degree of polymerisation N ◦ is the num-
ber of repeat units in a chain. This is typically several thousand or several tens or

DOI: 10.1007/978-3-319-06148-1_4,
122 4 Polymers
Fig. 4.1 Monomers of sev- (a) CH 2

eral common polymers.
a Polyethylene. b Polyvinyl H
chloride. c Polyamide. d
Polystyrene. Each polymer (b) CH 2 C
consists of a repeated chain of
these basic building blocks
Cl
(c) CH 2 (CH2)5 CO
(d) CH 2 CH
hundreds of thousands, although it is not fixed for a given species and depends on
how the material was synthesised. Polymers may also be ramified, carrying branches
built up from different monomers. These are copolymers. In this chapter, we shall
discuss only the basic rheophysics of simple polymers.
The peculiar feature of polymers which leads to their remarkable mechanical
properties is that the configuration, i.e., the geometric form of a given molecule,
is not determinate. Depending on the circumstances, a polymer chain may assume
a range of different forms, going from completely coiled up to fully stretched out.
The size and flexibility of polymers leads to more complex internal structures than
one would find in materials made up of small molecules, and this has significant
consequences for their mechanical behaviour. The shape, also referred to as the
conformation, of each molecule can vary considerably depending on the environment
(liquid or polymer) and the flow history. In different circumstances, the chains can
occupy a very large region, become entangled together, or else coil up into globules.
An important practical consequence is that the mechanical behaviour of a liquid
can be radically altered by suspending polymer chains in it. Moreover, new kinds
of collective structures can emerge when polymers are used, including entangled
structures and micellar arrangements in which a finite number of chains is assembled
into a specific structure.
As for simple materials, when the temperature rises, the chains become more
mobile relative to each other and the polymer enters a liquid state. As the temperature
falls, the resulting reduced mobility generally brings the system into an amorphous
(glassy) state (see Sect. 4.7). For some polymers with sufficiently regular chains,
a solid crystalline state can be obtained. We shall be mainly concerned here with
polymers in the liquid state or else solidified by specific interactions between the
chains (cross-linked polymers or polymer gels) (see Sect. 4.5).
Since the flexibility of polymer chains is related to their structure, we shall first
need some tools for describing their spatial organisation and apparent size (see
Sect. 4.2). These results will serve as a useful reference, but as soon as the chain is
4.1 Introduction 123
placed in suspension in a liquid, interactions with the solvent will necessarily affect
its shape (see Sect. 4.3). New kinds of interaction (compression, entanglement) will
occur when several chains are placed in suspension in the same volume of liquid.
Different concentration regimes can be identified with reference to such interactions
(see Sect. 4.4). We will then be in a position to discuss the mechanical behaviour of
polymers in the various regimes (see Sect. 4.6).
4.2 Structure
The shapes of polymer chains condition their interactions and more generally the
rheophysics of polymer-based systems. In practice, it is found that polymer chains
can assume all kinds of shapes. Although these may look like a random walk when
going from one end of the chain to the other, we shall nevertheless be able to describe
the average characteristics of large numbers of chains with reference to the average
length, length distribution, and apparent volume. We shall then discuss the physical
origin of this random orientation more precisely.
4.2.1 Apparent Length of a Chain
We assume that a polymer chain can be built up by joining together N sections of

length b with random relative orientations. Under such conditions, the apparent length
of a chain, that is, the actual distance between its two endpoints, will differ from its
true length, because the chain follows a more or less tortuous path through space (see
Fig. 4.2). This apparent length thus varies widely depending on the tortuosity of the
path. However, under the above assumption, the statistics of the relative orientations
is given. It is therefore possible to estimate the average apparent length r of a chain
comprising N sections, defining it as the average of the apparent length of the chain
over all possible paths.
If P is the number of possible paths starting from one end of a chain comprising
N sections, then for a given configuration p, the apparent length of the chain is the
magnitude of the vector rp from one end to the other. This vector is the sum of the
(p)
elementary vectors bi joining the ends of neighbouring sections (see Fig. 4.2), viz.,

N
(p)
rp = bi . (4.1)
i=1
For practical reasons regarding the mathematics, we shall consider the mean squared
apparent length of these paths, denoted by
124 4 Polymers
Fig. 4.2 Apparent length, b6

represented by the vector rp , b5 b10 b11
of a polymer chain comprising
sections of length b and b9
b7 b4
random relative orientations, b12
represented by the vectors b8
b3 b14 b15
bi=1,...,17 b2 b16
b13
b1
b17
rp
1 2
P
r 2 = r 2 ∞ = rp . (4.2)
P
p=1
Substituting (4.1) into (4.2) and expanding out, we obtain

⎡ ⎧
⎢
(p) ⎪
P N
1 ⎢ (p)2 (p)
r2 = ⎣ bi + bj · bk ⎪ ⎨ . (4.3)
P
p=1 i,j,k=1
j=k
(p)2
In this expression, we have bi = b2 , since the sections are of constant length.
(p) (p)
In addition, as there is no correlation between two arbitrary vectors bj and bk
(p) (p)
such that j = k, the average of the scalar product bj · bk over all configurations
vanishes. The result is thus r 2 = Nb2 , whence we may write
⇒
r = r 2 ∞1/2 = Nb , (4.4)
bearing in mind, as for the diffusion of colloidal particles (see Sect. 5.2.2), that this
expression merely gives an indication of the mean apparent length, while strictly
speaking it is the square root of the mean squared length. Note in passing that the
effective length of the molecule is Nb which is significantly longer than the root
mean squared apparent length given by (4.4).
4.2.2 Distribution of Apparent Chain Lengths
Since the flow of a polymer-based material will involve deformation of the chains, it
is useful to be able to specify the distribution of possible shapes of the chains, in order
to identify their conformations, or at least the most probable lengths of the chains,
given the characteristics of the medium. Here we discuss a very simple case that will
be used as an essential reference, namely, the distribution of apparent lengths, i.e.,
4.2 Structure 125
the probability of finding a chain of some specific apparent length r, assuming that
there is no restriction on the orientations or relative positions of the polymer sections.
To determine this distribution while avoiding heavy computations, we first note
that the path followed from one section to the next within a chain is just like the
path resulting from a random walk with steps of length b in arbitrary directions. To
begin with, consider the simplest case of one dimension. In this case, the distance
between the two endpoints results from the wanderings of a walker who moves in
steps of length b either to the left or to the right along a straight line. The chain
comprises A steps to the left and B steps to the right, whence the total number of
steps is N = A + B and the chain has length x = (B − A)b. Therefore,
1 x 1 x
A= N− , B= N+ ,
2 b 2 b
and we see that the distance x fixes the values of A and B. We would now like to know
how many ways there are to obtain a given distance x. This is the number of ways
of choosing A from N, viz., Z = N!/A!B!. A good first order approximation can be
obtained from this expression using the Stirling formula N! √ (N/e)N . Assuming
also that x is in fact much smaller than Nb, which is usually the case since we have
already
⇒ seen that the average distance between the two endpoints is of the order of
N, we eventually find
x2
ln Z √ N ln 2 − ,
2Nb2
or again,
Z √ 2N exp −α 2 x 2 , (4.5)
⎩
where α = 1/2Nb2 . To find the probability that the walker attains a distance x, we
must divide Z by the total number X of possible paths. The latter is found by summing
the values of Z obtained for all distances x that can be achieved in N steps, i.e.,

Nb ∞
X= Z(x) √ 2 N
exp −α 2 x 2 dx . (4.6)
x=−Nb −∞
Since
∞ ⇒
exp −x 2 dx = ξ ,
−∞
⎩
it follows that X √ 2N ξ/α 2 and the probability that the final distance is x is
therefore α
P(x) = ⇒ exp −α 2 x 2 . (4.7)
ξ
126 4 Polymers
This is a Gaussian distribution, just as one would expect given the completely random
nature of the steps the path takes through space.
This result can be generalised to a random walk in three dimensions. We then
obtain a Gaussian once again but with a different prefactor: the probability that the
vector joining the two ends of the chain lies between r and r + dr is α(r)dr 3 , where
3
β 3
α(r) = ⇒ exp(−β r ),2 2
β= . (4.8)
ξ 2Nb2
4.2.3 Radius of Gyration
In practice, we do not measure r directly. Instead we measure the radius of gyration

RG , defined as the square root of the mean squared distance between the different
carbon atoms and the centre of gravity of the chain. This is given by
1
N

2
RG = (ri − rg )2 . (4.9)
N
i=1
Note that we have changed the notation for the average. We now use ∞ to denote the
average over all possible configurations. Using the definition of the centre of gravity,
viz.,
1
N
rg = rj ,
N
j=1
and temporarily dropping the averaging sign which should apply to each of the terms
under the summation signs, we obtain
⎡ ⎛ ⎞2 ⎧
1
N
⎢ 21
N
1
N
1
N
⎪
Rg2 = ⎣ri 1 − 2ri · rj + ⎝ rj ⎠ ⎨ .
N N N N
i=1 j=1 j=1 j=1
The last term in this expression can be rewritten in the form
1
N
ri · rj .
N2
i,j=1
We thus obtain
4.2 Structure 127
1 2 1 2 1 2
N N N

Rg2 = 2
ri −2r ·r
i j +r ·r
i j = 2
ri −r ·r
i j = 2
rj −rj ·ri .
N N N
i,j=1 i,j=1 i,j=1
Summing the last two expressions, we obtain
1 2
N

ri − 2ri · rj + rj
2
,
N2
i,j=1
and hence finally,

1
N

2
RG = (ri − rj )2 . (4.10)
2N 2
i,j=1
According
to the above calculation
of the apparent length of a chain, we know that
(ri −rj )2 = |i−j|b2 . Since N i=1 N(N +1)/2, the
i = term under the first summation
N 2
sign is (N − i)(N − i + 1) + i − i /2. Now using i=1 i = N(N + 1)(2N + 1)/6
2
and neglecting terms of second order since N ≈ 1, we obtain

N
N
N3
|i − j| √ ,
3
i=1 j=1
whence finally we have ⇒

Nb
RG = ⇒ . (4.11)
6
Since the molar mass M of the polymer is proportional to N, (4.11) shows that the
radius of gyration is proportional to the square root of the molar mass of the polymer.
4.2.4 Extension of a Chain Under Traction
As we have seen above, the apparent length of the chains of a polymer at equilibrium
depends on the physicochemical characteristics of the system and the temperature.
Extending a molecule, or more generally an ensemble of molecules, will thus require
a certain supply of energy to the system.
Imagine now that we could take hold of the two ends of a large ensemble of
molecules initially in equilibrium and pull them apart. By doing so, we would extend
all the molecules by a certain amount (see Fig. 4.3), thereby modifying the number
of possible configurations and reducing the entropy of the system. According to the
basic thermodynamics discussed in Appendix B, we know that, for such a system at
constant temperature, the change in free energy will be equal to the work done. During
128 4 Polymers
Fig. 4.3 Examples of

conformations of a poly-
mer chain before (upper) and
after (lower) an increase in
its apparent length due to the A B
action of a traction force
r
A B
this operation, the internal energy will not change since all the interactions between
the different types of molecules will be preserved to a good approximation, so the
change in free energy will have the form dF = −T dS. Moreover, the work done on a
chain of length r can be expressed in terms of a force f which stretches the molecules
through a distance dr in the direction of f, giving an increment ∂W = f · dr = f dr.
Now the entropy of chains of length r is given up to a constant by (B.9) as S = kB ln α.
Hence, using (4.8), it follows that
f = 2kB T β 2 r . (4.12)
Since this force is simply proportional to the distance r and the system returns to
equilibrium (in terms of the length distribution) when the force is removed, the
behaviour of the chain is linearly elastic. The stiffness of this chain is therefore
3kB T
k0 = . (4.13)
Nb2
It is quite understandable that this stiffness should increase when the effective length
of the chain decreases, since the number of degrees of freedom then also decreases.
Note also that the stiffness is proportional to the temperature. This happens because,
when there is more thermal agitation in the system, atoms can explore the various
configurations more quickly and the chain tends to return more easily to its equilib-
rium configuration, making it all the more difficult to stretch. However, the linear
elasticity deduced by this method is obviously just an approximation, bearing in mind
the more complex effective structure of a long chain that may be partly entangled.
Moreover, since the interactions of the molecules with the solvent can cause them to
differ from the form resulting from a random walk, the above results are only strictly
valid in the case of a good solvent or in the case of a polymer melt or a concentrated
polymer (see below). On the other hand, the qualitative conclusion that the chains
behave like elastic objects remains true whatever the solvent–polymer interactions.
4.2 Structure 129
4
3
gauche
109.5°
1 2
trans
a
gauche
Fig. 4.4 Possible relative positions of four successive carbon atoms (black dots numbered from 1
to 4) in a polyethylene chain. The angle between two successive sections is fixed, whence atoms 3
and 4 must necessarily lie on the corresponding dashed circle. Furthermore, each atom has three
positions of minimum energy, i.e., two ‘gauche’ and one ‘trans’, shown here only for atom 3
4.2.5 Persistence Length
The various conformations of the chain are made possible by the fact that the
relative orientation of two consecutive segments joining three carbon atoms is not
fixed. However, this orientation between two consecutive segments is not arbitrary
since only a finite number of relative orientations is actually possible. But after a
long enough sequence of segments, all relative orientations of the two end segments
become equiprobable. This argument justifies using statistical calculations to deter-
mine the average characteristics of the chains of a polymer, provided that the chains
contain enough segments to be able to represent them as the successive connection
of N sections of length b with random relative orientations.
As an example, consider the simplest macromolecule polyethylene, comprising
a chain of CH2 groups with a CH at each endpoint (see Fig. 4.1a). The backbone of
the chain is made up of C–C bonds, which are not extendible, maintaining a typical
length a = 1.54 Å. Furthermore, the angle between two consecutive bonds between
carbon atoms is fixed at 109.5∝ , thereby picking out a circle on which the next
carbon atom can sit, as shown in Fig. 4.4. Considering now a section of the chain
comprising a certain number of segments, the number of possible paths between
the two end segments increases very rapidly with the number of segments, and the
relative orientation of these two segments becomes random for all practical purposes
beyond a high enough number of segments. So beyond the chain length b associated
with this critical number of segments, the relative orientations of two sufficiently far
removed segments are completely uncorrelated (see Fig. 4.5).
This length, referred to as the persistence length, depends on the chemical
properties of the given polymer. Indeed, the relative orientations available to two
130 4 Polymers
θ
b
Fig. 4.5 Polymer chain made from segments of length a (thin continuous line). Sections comprising
a certain number of successive segments of total length greater than the persistence length b have
a relative orientation θ that may be treated as random. Segments equivalent to these sections are
represented by thick continuous lines
Energy
gauche gauche
ΔE Δε
trans
-120° 0° 120°
Angle
Fig. 4.6 Energy levels of carbon atoms in the polyethylene molecule for different values of the
relative orientation between two consecutive sections
neighbouring sections are determined by the interactions between each atom and its
neighbours. Hence, in the polyethylene molecule, two consecutive sections are not
free to occupy all possible conformations while maintaining a fixed angle from one
to the other (see Fig. 4.4): only 3 positions actually correspond to potential energy
minima and these are located at intervals of 120∝ from one another. Two of them,
referred to as gauche, are associated with the same energy level, while the third,
referred to as trans, is associated with a lower energy level. We denote the energy
barrier between these two energy levels by ϕE and the energy difference between
the gauche and trans positions by ϕπ (see Fig. 4.6).
The ratio of the populations of trans and gauche at equilibrium can be calculated
using standard arguments of statistical physics. If the number of gauche positions is
4.2 Structure 131
ng , the number of transitions from this position per unit time (to go into the trans
position) depends on the height of the energy barrier that has to be overcome, i.e.,
ϕE − ϕπ, and is given by

ϕE − ϕπ
ng f0 exp − ,
RT
where f0 is a characteristic frequency of the system, equal for example to 1011 s−1 in
the case of polyethylene. The number of transitions from trans positions to gauche
positions is

ϕE
2nt f0 exp − ,
RT
where nt is the number of trans positions and the factor of 2 comes from the fact
that there are two possible gauche positions. Finally, at equilibrium, the number of
transitions in the two directions must be equal, whence it follows by equating the
two expressions above that

ng ϕπ
= 2 exp − . (4.14)
nt RT
This equation shows that, above a certain temperature of the order of ϕπ/R, the ratio
x = ng /nt is of order 1, implying that each bond is equally likely to be oriented in
either of the two directions. In this case, the path followed soon becomes random.
On the other hand, at lower temperatures, ng /nt will be close to 0, implying that
the trans positions will dominate. The chain then has a well determined shape since
the relative orientations of two successive sections are more or less fixed. However,
above a large enough number of bonds, changes in direction may start to appear,
resulting in more varied chain conformations.
◦
The probability of having no gauche on a chain with n◦ segments is (1 − x)n .
◦
The probability of having at least one gauche is thus 1−(1−x)n , or to first order n◦ x,
an approximation that improves as the probability of having a gauche decreases, i.e.,
as x ∇ 1. The persistence length b is the distance beyond which there is a significant
probability, say of order 1, of having at least one gauche position among all the
relative orientations of consecutive segments. In this way, the number of gauche
positions is generally much greater than 1, making the relative orientations of the
two end segments more or less random. Finally, the length b = λ n◦ a of a section is
proportional to the length a of a segment, the number n◦ of segments it contains, and
a form factor λ which accounts for the tortuosity of the section. The total number of
elementary segments is then N ◦ = n◦ N. Using (4.1), this yields
λ ϕπ
b√ a exp . (4.15)
2 RT
132 4 Polymers
We can thus obtain an expression for the average apparent length in terms of the
elementary length of the segments using the relation b = λ n◦ a in (4.4).
We can now calculate the typical time a trans position can remain in place. To a first
approximation, this is the ratio between the number of trans positions and the number
of transitions per unit time, viz., nt /nt f0 exp(−ϕE/RT ). This so-called persistence
time is thus given by
1 ϕE
θ = exp . (4.16)
f0 RT
It is the characteristic response time of the polymer chain when it is subjected to the
action of any kind of force. The following analyses will only be valid when this time
is much shorter than the characteristic time of the given phenomenon.
4.3 Polymers in Solution
Up to now we have considered the conformation of the molecules independently of

any environmental effects. In particular, we have neglected two key aspects of the
problem:
• The fact that it is impossible for two atoms to sit at the same point in space, i.e., a
chain cannot intersect itself.
• Interactions between polymer and solvent.
The first point implies that the path described by the consecutive bonds is not totally
random. As it seeks to avoid itself, the chain tends to extend further out into space.
The second effect means that, depending on the affinity of the monomers for the
solvent, the chain may tend either to swell in such a way as to increase the number of
interactions with the elements of this solvent, or to shrink in such a way as to increase
the share of interactions between its own elements. The relative importance of these
effects will in the end determine the effective dimension of the chains in solution.
A precise description of the various effects that come into play for a polymer
in solution is very complex. However, an approximate description is nevertheless
possible, reflecting the main physical phenomena and leading to generally reasonable
results. The idea here is to estimate the total free energy F = U − TS of a chain,
taking into account the different terms associated with the phenomena discussed
above, then determine the conditions that would minimise this energy.
4.3.1 Configurational Free Energy
According to (4.8), the configurational free energy Fconf = −TS = −kB T ln α of a

chain of length r made up of N sections is given up to a constant by
4.3 Polymers in Solution 133
3kB Tr 2
Fconf = . (4.17)
2Nb2
This expression assumes that the atoms can occupy all spatial positions with the
same probability whatever the arrangement of the other atoms. The fact that it is
impossible for two atoms to sit at the same point in space induces a change in the
entropy.
To estimate this term, we focus specifically on the factor that arises from the
distribution of positions of the different elements in the expression kB ln Z for the
entropy, where Z is the number of possible configurations. Here we shall only carry
out a rough calculation, assuming that we must distribute a total of N ◦ repeat units
of volume Ω in a total volume ν, without taking into account the fact that these units
are actually bound together to form chains. To simplify the calculation, we shall also
assume that the repeat units can only occupy a finite number of positions in a space that
has been divided up into small elementary volumes Ω. The calculation proceeds by
allocating the units one by one to these volume elements. It resembles the calculation
of the entropy of a condensed system which led to (2.24) in Sect. 2.4.1, viz.,
N ◦
1 ν N◦
Z→ − ,
N ◦! Ω 2
using the reasonable approximation β = 1/2. The entropy is thus given by

◦
(ν/Ω)N ◦ N ◦Ω N

S = kB ln Z = kB ln + kB ln 1 − .
N ◦! 2ν
The first term is the configurational entropy associated with the random distribution
of N ◦ particles in a volume ν without taking into account the volume of the particles.
The second term is the modulation induced by steric effects, which is what concerns
us here. It can be written to first order as −kB N ◦2 Ω/2ν. The associated free energy
is thus
N ◦2 Ω
Fexcl = kB T . (4.18)
2ν
4.3.2 Free Energy Associated with Interactions Between Molecules
We now focus on the impact of interactions between chain elements and solvent
molecules, which gives rise to a free interaction energy. To simplify, we only take
into account interactions between nearest neighbours and we assume that the repeat
units of the polymer are distributed uniformly throughout the liquid, so that the
interaction energy depends directly on the concentration of the polymer. Under such
134 4 Polymers
conditions, the number of contacts Np between monomers can be found by first

expressing the fact that, for each unit, the probability of contact with another unit is
zcΩ, where z is the number of neighbours of each unit or each solvent molecule and
c is the concentration of monomers in the solvent. This calculation can be repeated
for each of the N units, but each contact is then counted twice, whence finally,
1 ◦
Np = N zΩc . (4.19)
2
We proceed in the same way to determine the number of contacts Nps between
monomers and solvent molecules. Each unit is in contact with z solvent molecules,
which gives a total of z N , from which we must subtract the number of contacts
prevented by the existence of contacts between monomers. Now each contact with
another monomer takes up a volume Ω, thereby excluding Ω/Ω contacts with solvent
molecules, where Ω is the volume of a solvent molecule. The total number of pre-
vented contacts is therefore equal to the product of Np and Ω/Ω , that is, (Ω/Ω )N ◦ zcΩ,
whence finally,
Ω ◦
Nps = z N ◦ − N zcΩ . (4.20)
Ω
To calculate the number of contacts Ns between solvent molecules, we assume that
adding the polymer has created pockets within the liquid which have separated the
solvent molecules, thereby reducing the number of contacts. Let Ns0 denote the num-
ber of contacts when there is no polymer. For each pocket created by adding a
monomer, the solvent–monomer interface is increased, and this reduces the number
of solvent–solvent contacts by z /2. However, a certain number of contacts will not
have disappeared due to contacts between monomers. This number is the product of
Np with the number of contacts of a monomer with solvent molecules prevented by
a contact between monomers, viz., Ω/Ω . Finally, we obtain
1 1Ω ◦
Ns = Ns0 − N ◦ z + N zcΩ . (4.21)
2 2 Ω
The total energy of the system due to interactions is U = Np wp + Nps wps + Ns ws ,
and using (4.19)–(4.21), this yields
w ws Ω Ω ws
U = N ◦ zcΩ ◦
p
+ − wps + N z wps − + Ns0 ws . (4.22)
2 2 Ω Ω 2
To a first approximation, the volume Ω of a repeat unit is a3 . For its part, the chain
is as though confined in a volume of order r 3 . Within this volume, the concentration
of monomers is c = N ◦ /r 3 . It follows that, in (4.22), only the first term depends on
r. We may then write the total energy in the form
4.3 Polymers in Solution 135
Fig. 4.7 Possible shapes of

a polymer chain in a poor (a) χ>1/2
solvent (a), in a theta solvent
(b), and in a good solvent (c)
(b) χ=1/2
(c) χ<1/2
N ◦2
U = −2kB T χ a3 + const. , (4.23)
2r 3
where
1 z Ω Ω
χ= 2wps − wp − ws .
2 kB T Ω Ω
4.3.3 Total Free Energy and Chain Size

Using (4.17), (4.18), and (4.23), the total free F = Fconf + Fexcl + Uint is given by

1 N 2 b3 3r 2
F = kB T (1 − 2χ ) ◦ 3 + + const. (4.24)
n λ 2r 3 2Nb2
We note that this energy is expressed as the sum of a term associated with the
configurational free energy of a chain without specific restriction (the second term)
and a correction term associated with steric effects and interactions between the
various atoms (the first term). Depending on the value of the parameter χ , the chain
may behave in different ways:
• When χ = 1/2 (see Fig. 4.7b), the correction term is zero. The excluded volume
effect is cancelled by the effect of interactions between the molecules of the dif-
ferent components. There are therefore no specific restrictions on the chain: its
geometric characteristics are governed entirely by the random walk process and
r → bN 1/2 . In this particular case, the interstitial liquid is called a theta solvent.
136 4 Polymers
• When χ < 1/2, the correction term is positive, but its derivative is negative, so
there is a free energy minimum for a value of r such that dF /dr = 0, i.e., when
r √ (n◦ λ 3 )−1/5 bN 3/5 √ (n◦ λ 2 )1/5 aN ◦3/5 . (4.25)
In this case, the dimension of the chain is greater than when governed by the
random walk hypothesis alone (see Fig. 4.7c). This would be considered a good
solvent for the given polymer, because the chains tend to occupy a greater apparent
volume in the liquid.
• When χ > 1/2, the attractive effect of the polymer–solvent interactions exceeds
the repulsive effect of the excluded volume. The energy minimum is obtained for a
chain of length zero, so the chain tends to coil up on itself to form a compact globule
(see Fig. 4.7a). This would be considered a poor solvent for the given polymer.
Note that, more or less by good fortune given the approximations made here, the
above analysis delivers the dependence (4.25) of r on N which is in fact observed
experimentally for a good solvent. A more accurate analysis would involve several
other corrections. Note also that the appearance of the chains in solution can be
altered by varying the temperature, thereby inducing changes in χ . The chain may
collapse below a critical temperature.
4.4 Several Chains in Solution
When there are several chains in solution in the same sample, a new type of interaction
may occur, namely, interactions between chains. Given the complex shape of each
chain, these interactions have varying degrees of complexity, depending on the
relative distance of the molecules. At long range, the molecules interact through
their apparent volumes (dilute regime), but when they come closer, they can begin
to interpenetrate and/or become entangled (semi-dilute or concentrated regimes).
The appearance and dimension of the molecules depends on the structure obtained
in these different regimes. Under these conditions, the first parameter we can actu-
ally compute among those controlling the different regimes is the effective volume
fraction occupied by the polymers in solution, which can be described through the
monomer concentration per unit volume, or monomer volume fraction,
γ = nΩN ◦ , (4.26)
where n is the number of chains per unit volume. However, we shall see that other
characteristics of the system may play a decisive role. These include the length of
the chains or their apparent dimension (which depends on the solvent).
Note that γ can be related to the mass fraction τ defined as the mass of polymer
per unit volume in the solution. The mass fraction can be expressed in the form
τ = nM/NA , where M is the molar mass, i.e., the mass of one mole of the polymer,
4.4 Several Chains in Solution 137
and NA is Avogadro’s number, i.e., the number of molecules in one mole, equal to
6 × 1023 . It follows that γ = τNA ΩN ◦ /M.
When considering the dependence of the material properties on the chain length,
one generally refers to the molar mass M of the polymer, which is simply proportional
to N ◦ for a given monomer (M = N ◦ m, where m is the molar mass of the monomer).
The general behaviour described here in terms of N ◦ is thus the same as one would
observe in terms of M for a given type of monomer.
4.4.1 Dilute Regime
When a few chains are placed in the solvent at sufficient distance from one another,
they will assume the same shape as if they were alone (see Sect. 4.4). This is the dilute
regime (see Fig. 4.8). In this regime, there is no contact between chains, i.e., their
apparent volumes do not interpenetrate. If we consider to a first approximation that
we are dealing with a suspension of coils of apparent volume RG 3 , we can calculate
the apparent volume fraction of coils in the solution:
η = nRG
3
. (4.27)
So we have the dilute regime whenever η is less than the maximum packing fraction
of these coils. For spheres, we know that this critical concentration is of the order
of 0.6 and depends on the way the particles are arranged. For polymers, given the
approximations made regarding the shape and volume of objects in suspension, it
suffices to assume that this critical concentration is of order 1. From (4.26) and
(4.27), we deduce that the upper bound on the effective concentration in the dilute
regime is γc √ ΩN ◦ /RG 3 . Note that, in this case, the average local concentration, i.e.,
◦
within a coil (ΩN /RG ), is therefore equal to the average monomer concentration in
3
the solution. This property remains valid above the critical concentration, because
the chains interpenetrate and there is no significant difference between the average
concentration and the local concentration on the scale of a few molecules.
For a good solvent, (4.25) implies that
γc √ N ◦−4/5 , (4.28)
where α = (n◦ λ 2 )−3/5 . For a typical value N ◦ ∼ 104 , assuming that α √ 1, we

have γc √ 5 × 10−4 . Critical coil packing is therefore attained for values of the
monomer volume fraction several orders of magnitude less than the usual value for
a suspension of non-colloidal particles (of the order of 0.6).
138 4 Polymers
N
N'
Φ<Φc Φc Φc< Φ <Φ+ Φ+ < Φ <1

Dilute Critical concentration Semi-dilute Concentrated
Fig. 4.8 Main regimes of geometric configuration for different values of monomer volume fraction
when polymers are dissolved in a solvent
4.4.2 Semi-dilute Regime
In contrast to suspensions, emulsions, and foams, it is relatively easy to increase

the concentration of a polymer solution beyond γc , because the chains are still far
from filling the volumes they appear to occupy. They can still interpenetrate and if
necessary compress somewhat in order to decrease their apparent volume. This is
the beginning of the semi-dilute regime.
Once there is interpenetration, the apparent volume of a chain is crossed by other
chains. One way to describe the structure is to identify volumes that are smaller
than those usually associated with whole chains in solution in the dilute regime but
containing a series of sections belonging to a single chain (see Fig. 4.8). The structure
as a whole can then be described as a compact assemblage of chains comprising
N < N sections. We shall refer to these as blobs (see Fig. 4.9). It is just as though we
had dissolved chains of this new length at the critical maximum packing concentration
of the coils. The monomer fraction then corresponds to the transition between the
dilute regime and the semi-dilute regime for fictitious chains comprising N ◦ = n◦ N
repeat units, i.e., γ = γc (N ◦ ) √ αN ◦−4/5 for a good solvent. Finally, we deduce
the length of these fictitious chains, viz., N ◦ √ (γ/α)−5/4 , noting that it decreases
as the concentration increases, as one would expect.
The characteristic size of these blobs is called the correlation length and denoted
by φ (see Fig. 4.9). When the chains are much longer than the correlation length, two
networks of the same concentration but with chains of different lengths will look the
same on the scale of the blobs. As a consequence, φ does not depend on the length
of the chains but only on the monomer concentration. We thus try to express φ as a
power law in γ. Since we also know that, when γ = γc , φ is equal to the radius of
gyration, we may deduce the general form φ = RG (γc /γ)m . Given the expressions
Fig. 4.9 Correlation length for a polymer in the semi-dilute regime
for RG and γc in terms of N ◦ in the case of a good solvent, we must have m = 3/4
for φ to be independent of N ◦ , whence
φ = aα 5/12 γ−3/4 . (4.29)
We then note that φ → N ◦3/5 , which agrees with the very rough idea obtained by
saying that the size of the blobs is given by (4.25). Furthermore, the above formulas
imply a relation between the number of monomers per blob and the size of the blobs,
namely, a3 N ◦ √ γφ 3 .
4.4.3 Concentrated Regime
The semi-dilute regime continues until the number N of sections of the fictitious
chains is of the order of 1 and the radius of gyration is of the order of the length of
a section, viz., RG √ b. The concentration is then equal to the critical concentration
γ+ which, given the above results for a good solvent, is equal to
a 3
γ+ √ n ◦ . (4.30)
b
In this case, the sections are almost in contact with one another (see Fig. 4.8). This
situation, obtained with a highly concentrated solution or a polymer melt, is the
concentrated regime.
Considering their complex structure, one might expect the chains to tend to repel
one another in this regime. In fact, this is not the case. For example, it is found
that the density of a polymer melt, made up solely of chains, i.e., without solvent,
does not depend significantly on the number of repeat units per chain and is close
to the density of the crystallised polymer. This shows that the monomers are very
close together whatever the situation. Furthermore, since deuterated chains can be
observed by neutron diffraction, it has been shown that the radius of gyration of the
140 4 Polymers
chains in a polymer melt is the same as for a coil in random conformation in a theta
solvent.
On the face of things, this result, known as Flory’s theorem, is rather surprising.
However, there is a relatively simple explanation. The key point here is that the
affinity between the monomers belonging to different chains is the same as between
monomers of the same chain. As a consequence, the interactions between repeat units
of a given chain and the other chains, in contrast to the polymer–solvent interactions,
does not produce any swelling effect or folding up of the chain. Furthermore, the
entropy effect discussed above for polymers in solution, which caused a certain
swelling of the chains, no longer operates. Since each chain is now surrounded
by other chains, any space freed up by one section will soon be occupied by a
section of some other chain. When there are no interaction effects or excluded volume
effects, calculations of the kind carried out in Sect. 4.4.1 would lead to the conclusion
that the conformation and size of the chains are the same as for chains in dilute
solution in a theta solvent. A detailed demonstration would nevertheless require
more sophisticated arguments. One can reach this conclusion even more quickly
by viewing the solution as a network of sections in contact: to create a chain, we
start with an arbitrarily chosen section and proceed by successively associating new
sections, each one in contact with the last. In this process, we note that no argument
would favour any particular choice of contact, whence the path followed by the chain
is simply the path of a random walk.
4.4.4 Entanglement
Going beyond the above considerations, we must take into account another important
but less easily quantified feature of the system, namely, the way the molecules obstruct
each other’s motions. To this end, we consider two situations:
• Each chain can move relative to the others (with the help of thermal agitation, of
course) without needing to deflect neighbouring chains from the motion imposed
on them by the macroscopic flow.
• The chains can only move relative to one another by deforming or by displacing
their neighbours. We consider then that the chains are entangled.
A certain vagueness in these definitions shows how difficult it is to specify what is
meant by entanglement from a geometrical standpoint. To simplify, in the follow-
ing we shall consider that there is entanglement at a point of contact between two
monomers belonging to two different chains, assuming implicitly that such a point
of contact is very likely to induce some kind of obstructive mechanism in the relative
motions, as described above.
Given these characteristics, one would naturally expect any entanglement to play
a significant role in the mechanical behaviour of the material. It is thus important to
specify how many points of entanglement there are. For any given repeat unit in a
chain, the probability of contact with a repeat unit of another chain is proportional
to the volume surrounding it and to the volume fraction of monomers, i.e., γ. This
Semi-dilute
N* entangled
Concentrated
entangled
Φc Semi-dilute
unentangled
Dilute
Concentrated
unentangled
Φ+ Φ
Fig. 4.10 Schematic phase diagram for the different structural regimes of a polymer and different
values of the monomer fraction and chain length. The curve of the critical concentration γc is given
by (4.28), while the curve separating the unentangled and entangled regimes is given by (4.30),
taking the constant equal to 2
probability must be multiplied by the number of monomers per unit volume to obtain
the number of points of contact per unit volume, which is therefore proportional to
γ2 . Furthermore, according to (4.26), the number of molecules per unit volume is
proportional to γ/N ◦ . It follows that the number of points of contact per molecule
is proportional to γN ◦ . For a given monomer, one would therefore expect the state
of entanglement to be similar along curves of the form
γN ◦ = const. (4.31)
An entangled structure will be one with at least some critical number of points of
contact γN ◦ per molecule (of the order of two).
Using the monomer concentration criteria for transition from one regime to
another as expressed in (4.28) and (4.29), together with the criterion for appear-
ance of an entangled structure described above, we may plot a diagram specifying
the various states of the material as a function of chain length and concentration
(see Fig. 4.10). Given the different approximations made above, not only in the cal-
culations but also in the specification of certain phenomena, this diagram must be
regarded as providing only an approximate classification.
142 4 Polymers
Fig. 4.11 Bond formation in an entangled polymer
4.5 Cross-Linked Polymers and Polymer Gels
A range of materials are formed by polymer chains bound together at a certain

number of cross-links. When the structure is solid, i.e., when it cannot be significantly
deformed without breaking, this is referred to as a polymer gel. The cross-links may
be of different kinds: covalent bonds between chain elements (epoxy resin) or via a
sulfur atom (rubber) giving chemical gels, or localised crystalline regions (gelatine)
or polystyrene blocks (thermoplastic elastomers) giving physical gels.
In all these cases, a three-dimensional network of interconnected polymers forms,
with a structure analogous to that of an entangled network in the semi-dilute or
concentrated regimes but with cross-links rather than points of contact (see Fig. 4.11).
The mechanical properties of the system depend on the number of cross-links and
the characteristics of the chains. In the following we shall refer to a part of the chain
between two successive cross-links on the chain as a section. Here we shall only be
concerned with the case where these sections are long and flexible enough to justify
applying our earlier results.
When there are no applied stresses, the network adopts a certain (equilibrium)
shape which results from the relative arrangement of the cross-links. In contrast to a
polymer in solution or a polymer melt, the chains cannot move through the network
with the help of thermal agitation, simply because they are attached to one another. On
the other hand, they can always deform between two cross-links under the effects of
thermal agitation, and there may even be a small agitation of the cross-links about their
average position. The set of all these average positions constitutes the equilibrium
configuration. This configuration corresponds to a minimum energy associated with
a maximum total entropy. When the material is deformed, this induces a deformation
of the network which displaces the cross-links relative to their equilibrium positions.
The sections will thus be stretched or contracted. A force must be applied in order
to deform the system. When this force is removed, since the system is made up of
an ensemble of essentially elastic chains, it will return to its equilibrium position of
its own accord. The material itself thus behaves as an elastic solid.
Such a system has a disordered structure made up of sections oriented in all
directions. To simplify the analysis of its properties, we assume that, when a deforma-
tion is imposed on the sample, all elements are displaced in proportion to this defor-
mation. In other words, we assume that the deformation is perfectly uniform on the
4.5 Cross-Linked Polymers and Polymer Gels 143
scale of the cross-links. Under these conditions, if the deformation induces a change
in the sample lengths in the various directions by factors (εx , εy , εz ), the initial posi-
tion (x, y, z) of a cross-link changes to (εx x, εy y, εz z) after deformation. Therefore,
considering two cross-links located at the origin and (x, y, z), respectively, the length
of the corresponding section changes from r0 given by r02 = (x 2 + y2 + z2 )1/2 to r1
given by r12 = (ε2x x 2 + ε2y y2 + ε2z z2 )1/2 . Neglecting the excluded volume effect and
referring to (4.17), the change in entropy per section is
3kB 2
ϕS = − 2
εx − 1 x 2 + ε2y − 1 y2 + ε2z − 1 z2 . (4.32)
2Nb
We now calculate the change in total entropy per unit volume assuming that there
are n sections per unit volume and that these sections are oriented uniformly in all
directions. In this case, we have ϕSvol = nϕS∞. In this calculation, we may use
x 2 ∞ + y2 ∞ + z2 ∞ = Nb2 /3, which follows naturally from (4.4). This yields
nkB 2
ϕSvol = − εx + ε2y + ε2z − 3 . (4.33)
2
Consider for example a simple elongation in the x direction, in which the sample is
deformed in such a way that every length in this direction varies by a factor εx = ε. In
the other directions, the lengths vary by the same factor ε = εy = εz . If the material
⇒
is incompressible, we must have εx εy εz = 1, whence it follows that ε = 1/ ε.
The change in free energy is then

nkB T 2
ϕFvol = −T ϕSvol = ε + −3 .
2
(4.34)
2 ε
According to (B.5), the work done by the external forces per unit volume is given
by δWvol = σ π, where σ is the normal stress and π = ε − 1 is the deformation of
the material in the principal direction. Since we also know that ϕFvol = δWvol , we
may calculate the stress from (4.34), which yields

ϕFvol 1
σ = = nkB T 1 + π − . (4.35)
π (1 + π)2
This does not correspond to linear elastic behaviour except for small deformations
π ∇ 1. In this case Young’s modulus is E = 3nkB T . Using a similar approach, it can
be shown that, under simple shear, the elastic modulus for small deformations is
G = nkB T . (4.36)
These results show that the mechanical properties of rubber depend solely on the
temperature and the density of cross-links. As long as the behaviour of the chains is
accurately described by the arguments put forward in this chapter, the chemistry of
144 4 Polymers
Fig. 4.12 Response of a .

polymer melt or concentrated γ0
polymer solution to a simple
shear. Variation of the stress
(lower) when the shear rate
changes abruptly from zero to t
some finite value, then returns τ
.
to zero (upper). To begin with, μγ0
we observe a linear elastic
regime, then a permanent flow
regime, and finally, relaxation
to zero stress when the flow
has stopped .
Gγ0
θ t
the chains and the exact nature of the cross-links play no role. The elastic modulus of
an elastomer thus decreases as the reciprocal of the average distance between cross-
links. So the greater the cross-linking, the greater the force required to produce a
deformation. Note, however, that if there are too many cross-links, this means that the
sections become too short to remain flexible and the material will then become very
rigid. Finally, elastomers are soft solids because their modulus is 1000–100 million
times lower than the modulus of a metal. They are also viscoelastic because the
stretching and relaxation of the chains are not instantaneous due to the Brownian
motion of the chains between cross-links and the possibility of viscous dissipation
in the interstitial liquid.
4.6 Mechanical Behaviour of Liquid Polymers
As for a suspension, the primary effect due to the presence in solution of objects
like polymer chains that are much bigger than the interstitial liquid molecules is
to increase the viscosity of the mixture as compared with that of the liquid alone.
Naturally, this effect becomes all the more important as the polymer concentration
increases. However, we cannot use the same arguments to predict the viscosity as a
function of the solid volume fraction, since we must take into account the distinguish-
ing features of polymer chains, such as their very high aspect ratio, adaptation of
the chain conformation, and entanglement. In this context, we may distinguish three
regimes of mechanical behaviour associated with the different geometric structures
described in Sect. 4.4. We can then predict certain aspects of the way the viscosity
will depend on the chain concentration and chain length.
4.6 Mechanical Behaviour of Liquid Polymers 145
Fig. 4.13 Representation

of a polymer solution by
an initial vertical line of
elastic chains (left) that will
tend to move relative to
one another and deform (by
γel. γ total
stretching), leading to a ‘liquid
γ liq.
deformation’ and an ‘elastic
deformation’, respectively
4.6.1.1 Viscoelasticity
One specific feature of the suspended entities (polymer chains) in the case of liquid
polymers is that they can stretch significantly and hence store elastic energy. As a
consequence, the macroscopic behaviour of the material is in part the same as that of
an elastic solid, and in part the same as that of a simple viscous liquid. For example,
if we take a polymer solution at rest and suddenly impose a simple shear at constant
shear rate, we observe that the stress first increases steadily in proportion to the
deformation (see Fig. 4.12). If the stress is released at this stage, the deformation
drops back to zero. Over short times, the material thus behaves like an elastic solid.
But if the shear is maintained, the stress levels out to a plateau corresponding to a
permanent flow regime (see Fig. 4.12). The material then behaves as a liquid. These
effects are all the more noticeable when the polymer concentration is high. The
physical origin of this type of behaviour is explained as follows. If the material is
subjected to a sudden stress, it reacts to a large extent by deformation of its constituent
chains, whence the elastic behaviour. But if the external forces are maintained, the
chains, possibly still deformed, begin to slip past one another, as in a viscous fluid.
To a first approximation, this behaviour can be understood by assuming that the
medium comprises an elastic phase of modulus G and a Newtonian viscous phase
of viscosity μ, arranged in such a way that the total deformation is the sum of the
deformations of each of these components (see Fig. 4.13). This is the simple Maxwell
model. Under simple shear, when the shear stress is Ψ , the total deformation is thus
ω = ωtot = ωel + ωliq , where ωel = Ψ/G and ω̇liq = Ψ/μ. Differentiating ωtot with
respect to time, we obtain a differential equation relating the stress to the shear rate:
1 dΨ Ψ
ω̇ = + . (4.37)
G dt μ
Consider for example a simple shear imposed at time t = 0, viz., ω̇ (t > 0) = ω̇0 , on
a sample that is initially at rest, i.e., Ψ (t < 0) = 0. The solution to (4.37) is then

Gt
Ψ = μω̇0 1 − exp − . (4.38)
μ
146 4 Polymers
This equation shows that, over very short times compared with the characteristic
time ε = μ/G, we have Ψ √ Gω̇0 ϕt = Gω , in other words, the material behaves as
an elastic solid. Over long times compared with ε, we have Ψ √ μω̇0 , in other words,
the material behaves as a Newtonian fluid. If having reached the permanent regime
we now suddenly immobilise the sample, we have Ψ (t < 0) = Ψ0 and ω̇ (t > 0) = 0,
whence the solution to (4.37) is

Gt
Ψ = Ψ0 exp − . (4.39)
μ
The stress thus decreases steadily to zero. The material relaxes with characteristic
time ε. These results agree well with the standard characteristics of viscoelastic fluids
such as those illustrated in Fig. 4.12.
Given these specific properties, the viscoelastic characteristics of polymers have
mainly been studied by imposing an oscillatory motion rather than a continuous flow
under simple shear. Suppose, for example, that we impose a sinusoidally varying
stress Ψ = Ψ0 sin ωt on a material whose behaviour is well described by the above
model. The resulting deformation follows directly from (4.37):
Ψ0 Ψ0
ω = sin ωt − cos ωt , (4.40)
G μω
which can also be written ω = ω0 sin(ωt −ϕ), where the amplitude of deformation ω0
and the phase shift ϕ satisfy Ψ0 = Gω0 cos ϕ and Ψ0 = μωω0 sin ϕ. This technique can
be used whatever the behaviour of the material. We then deduce two characteristics
of the material, namely the elastic modulus G and the viscous modulus G , defined
as follows: Ψ0 Ψ0
G = cos ϕ, G = sin ϕ . (4.41)
ω0 ω0
When the behaviour of the material is well described by the Maxwell model, this
experimental procedure ascertains the characteristics of the material directly since
G = G and G = μω. The characteristic relaxation time is then given by ε =
G /ωG .
4.6.1.2 Shear Thinning
Shear thinning effects are also observed with polymers. That is, when the shear rate
is low enough, the viscosity is constant, i.e., independent of ω̇ , but when the shear
rate is high enough, the viscosity begins to fall with increasing ω̇ . There is a rather
simple qualitative explanation for this phenomenon. Since the ability of the chains
to stretch depends on the forces applied to them, one also expects the extension to
vary with the applied stress or, what comes to the same, with the shear rate. This may
then be accompanied by an increasing tendency to align in the direction of flow. Put
another way, the equilibrium structure of the material in the permanent flow regime
will change in such a way as to offer less resistance to flow as the shear rate increases.
Such shear thinning effects are particularly sensitive in the concentrated regime.
4.6.2 Dilute Regime
4.6.2.1 Permanent Flow Behaviour
In this regime, we may treat the polymer coils dispersed in the solvent as porous
spheres of radius RG , in which the pores have average volume RG 3 /N. An order of
magnitude for the characteristic size of the pores is thus l = (RG /N)1/3 , which is also
3
approximately equal to the maximum thickness of the liquid layer in contact with
the surrounding liquid. For a given stress applied to the mixture, the resulting shear
rate in the liquid on the outskirts of the coil is therefore of the order of RG /l times
lower than would be induced in a liquid volume free of any polymer chain and of
linear dimension RG . As a consequence, the apparent viscosity μ of the liquid within
the porous coil is of the order of N 1/3 times the apparent viscosity of the suspending
liquid, i.e., usually more than 10 times greater. To a first approximation, it is as
though the coils and their interstitial liquid were in fact rigid spheres suspended in
the solvent.
When η is at most of the order of a few percent, we know that the viscosity
of a suspension of spheres agrees well with the prediction of Einstein’s model in
(3.16), i.e.,
μ
= 1 + 2.5η . (4.42)
μ0
Given the relation between γ and η, the second term of (4.42) is proportional to
N 1/2 for a theta solvent and N 4/5 for a good solvent. In practice, in the second
case, the observed coefficient is actually slightly less than 4/5. This discrepancy may
reasonably be attributed to the approximations that have been made here, namely,
sphericity of the coils and absence of hydrodynamic exchange between the inside
and outside of the coils. In any case, (4.42) provides us with a way of estimating the
chain size: if we know the monomer concentration, we can deduce N by measuring
the viscosity of the mixture.
When the concentration increases beyond this range, the viscosity increases qual-
itatively as for a suspension of solid particles (see Chap. 3). In other words, the rate of
change of the viscosity increases with the concentration until it reaches a high value.
This can only be described very roughly using the Krieger–Dougherty formula (see
Sect. 3.4.4), because for polymers the interactions between chains or between the
liquid phases assumed fixed in the above approach play an ever more important role
as one approaches the semi-dilute regime, and in addition they depend on the quality
of the solvent.
However this may be, it is found that the viscosity can be high with a low effective
monomer concentration as soon as N becomes large. It has thus become common
practice to add polymers to all kinds of products, such as shampoos and washing-
148 4 Polymers
up liquids, the aim being to significantly increase their viscosity without noticeably
changing their overall chemical composition.
The viscoelastic behaviour of dilute polymer solutions derives from the elastic
properties of the molecules suspended in the liquid (see Sect. 4.2.4). The standard
rheophysical approach here is to assume that certain characteristics of the behaviour
of the components on the local level—and in particular the relaxation modes—
reflect the observed features of the macroscopic behaviour. In this context, we may
for example estimate the characteristic time ε by representing a chain as two objects
joined together by a spring of negligible thickness compared with the dimensions of
the object. We now imagine that the two objects are moved a distance r0 from their
equilibrium position, after which the system is allowed to evolve freely in a liquid.
In the frame of one of the two objects, neglecting inertial effects and thermal agita-
tion, the position of the other object as time goes by can be found by examining the
balance of forces exerted on this object: on the one hand, there is the elastic restoring
force −k0 r given by (4.12), where r is the distance from the equilibrium position,
and on the other hand, viscous friction due to the motion of the object through the
liquid. According to (3.5), the viscous friction is −λ bμ0 ṙ, where λ depends on the
shape of the object. The balance of forces is therefore expressed by
dr
k0 r + λ bμ0 =0, (4.43)
dt
which has solution r = r0 exp(−t/ε), where ε = λ bμ0 /k0 . Considering now a

molecule comprising N links, giving a friction coefficient of the form −λ bNμ0 , and
using for example the expression (4.16) for k0 in the case of a theta solvent, we obtain
λ μ0 b3 2
ε= N . (4.44)
3kB T
In this context, ε is the characteristic time required by the chain to return to equilib-
rium when the sections remain aligned. To a first approximation, ε could thus be the
characteristic viscoelastic time of the system identified in macroscopic experiments
(see Sect. 4.6.1).
However, the reality is more complex. The model used above does not take into
account the mutual interactions between the different sections of the chain structure
and thermal agitation which tends to slow down the return of the chain to its equilib-
rium conformation. A stretched chain relaxes through an interplay of several coupled
relaxations. In 1953, Rouse suggested describing these effects by means of a series
of relaxation times ε0 , ε1 , . . . , εn associated with the vibrations of sections with
different lengths and different elastic moduli G0 , G1 , . . . , Gn , and this basic idea
underpins several theoretical approaches since developed. Given that the physical
phenomena underlying these relaxations are always the elasticity of certain parts of
the chain and viscous friction due to the displacements of the monomers through the
liquid, the above relaxation times have the form αε, where α < 1 is a dimensionless
factor depending on the given mode.
It turns out that the technique discussed above, namely observing the response
of the fluid to periodic excitation, provides direct access to the properties of the
material under the assumption that these properties correspond to this model. The
general solution to (4.37) in terms of the flow history ω̇ (t < t) is
t
Ψ= G(t − t )ω̇ dt , (4.45)
−∞
using the relaxation modulus function G(t) = G exp(−t/ε), which completely

describes the behaviour of the material in terms of its relaxation characteristics,
these being very simple in this case. For a material whose properties are described
by a set of relaxation times, the expression for the function G(t) can be generalised to

n
G(t) = Gk exp(−t/εk ) , (4.46)
k=1
whence the material behaviour is described by
t
n
t − t
Ψ= Gk exp − ω̇ (t )dt . (4.47)
εk
−∞ k=1
In this case, the elastic and viscous moduli of the material can be calculated directly:
ωk2 ε2k ωk εk
G (ω) = Gk , G (ω) = Gk . (4.48)
k
1 + ωk2 ε2k k
1 + ωk2 ε2k
Despite the many physical characteristics of the system that may potentially be
involved in its macroscopic behaviour, the latter is relatively simple in the extreme
flow regimes. So for very low oscillation frequencies ω → 0, we obtain

G √ ω2 Gk ε2k , G √ ω Gk εk ,
k k
while for high frequencies ω → ∞, we find

G √ Gk , G −→ 0 .
k
150 4 Polymers
Fig. 4.14 Typical frequency

dependence of elastic and
viscous moduli of a liquid
polymer (continuous line) and Liquid
viscous modulus of a simple polymer
liquid (dashed line), shown on
a logarithmic scale. Note that, G''
following the theoretically
predicted decrease, the vis- 1 G'
cous modulus increases once G''
2
again at high frequencies Simple liquid
ω
The theoretical basis of the behaviour of these materials is thus deduced in terms of
the frequency dependence of the elastic and viscous moduli, and the results agree
well with experimental observations (see Fig. 4.14).
4.6.3.1 Behaviour in the Permanent Flow Regime
In this regime, structural changes induced by flow play a crucial role in the behaviour
of the material. Under shear, chains can stretch, line up in the flow direction, and/or
disentangle. At low shear rates, the structure has time to return to equilibrium, but this
is not so at high shear rates. In practice, the viscosity is found to remain constant below
a critical value ω̇c of the shear rate, then fall off significantly with increasing shear
rate, following a power law η → ω̇ −n , where n lies between 0 and 1 (see Fig. 4.15).
This decrease in the apparent viscosity with increasing shear rate is explained by
the fact that the average structure changes when the shear rate reaches higher values.
Indeed, the polymers are more and more stretched out, aligned, and/or disentangled.
Since the relative motion of any given chain becomes more dependent on the motions
of the other chains when the chains are long or the concentration is high, this change
in structure occurs all the sooner as the flow rate is increased. For this reason, the
critical shear rate ω̇c indicating the change of regime decreases as the chain length
or concentration increase. The critical shear rate can be estimated by assuming that
the material recovers its equilibrium structure after a characteristic time ε for the
return to equilibrium of a single chain alone in the liquid. Under such conditions, as
soon as ε is greater than the characteristic time of the flow, i.e., 1/ω̇ , the chains no
longer have time to return to their equilibrium configuration during an elementary
deformation. This implies that ω̇c √ 1/ε. In the following, when we discuss the
impact of certain parameters like temperature, chain length, etc., on the behaviour
of the material, we shall only be concerned with the viscosity η0 associated with the
plateau obtained for low shear rates.
Fig. 4.15 Dependence of

the apparent viscosity of log(η)
a polymer solution on the
shear rate for different molar
masses, concentrations, and
temperatures. The sketches
illustrate disentanglement,
stretching, and alignment of
the chains in the flow direction N,Φ,1/T
. .
γc γ
In practice, it is found that the viscosity on the plateau η0 = η(ω̇ → 0) increases

with the chain length according to the following almost universal options:
η0 → N for N < Nc , (4.49a)

η0 → N 3.3 for N > Nc , (4.49b)
where Nc is a critical chain length separating the two regimes. In fact, to a first
approximation, Nc varies as the reciprocal of the concentration γ and the transition
thus occurs for a critical value of Nγ. With reference to the diagram of Fig. 4.10,
this suggests that the transition between the two regimes occurs beyond some critical
level of entanglement.
The trends observed for the viscosity on this plateau can be explained very
qualitatively. When the chains are short, they have enough relative mobility for
shear to occur through relative motion of neighbouring chains, possibly along some-
what tortuous paths. If a given deformation increment is directly associated with the
relative motion of two neighbouring chains, the associated friction will be propor-
tional to the chain lengths and we obtain a formula like (4.49a). When the chains
are entangled, flow requires the molecules to make their way through the network
using a much more complex process, based in particular on cooperative phenomena.
This explains why the resistance to flow increases much more quickly with the chain
length.
In the concentrated regime, the elastic nature of the material can be very striking. This
is observed, for example, when a stress is applied to a material initially at rest. The
resulting deformation increases rapidly to a plateau called the rubber plateau. In this
phase, the response of the material is thus largely that of an elastic solid, undergoing
a finite deformation for a given stress. This phase lasts for a time θ , after which the
deformation increases linearly with time, corresponding to flow at constant veloc-
ity (see Fig. 4.16). The material then flows like a viscous liquid with viscosity μ.
152 4 Polymers
Fig. 4.16 Deformation

response (lower) when a σ0
step-shaped stress is imposed
(upper) on a polymer melt or
concentrated polymer solu-
tion. Dashed line response of t
a simple Maxwell fluid under γ
the same excitation
σ0 /μ
σ0 /G
θ t
Although the simple Maxwell model or the Rouse model described above do indeed
predict an initial increment in deformation followed by flow with constant apparent
viscosity, they do not predict the retardation time associated with the rubber plateau
(see Fig. 4.16). The characteristic time θ for the end of the plateau can be very long,
giving the impression that the material will remain in a solid state. This time is
nevertheless related to the viscosity and the elastic modulus by
μ
θ= . (4.50)
G
With this kind of material, a plateau is also observed in the curve of G as a function of
frequency (see Fig. 4.17). This plateau becomes more pronounced for longer chains.
It turns out that the critical frequency at which the G plateau appears is roughly equal
to 1/θ , in agreement with the above observations. The material reacts essentially like
an elastic solid as long as the characteristic excitation time in a given direction is less
than θ . This observation underlies the empirical Cox–Merz law ⇒ according to which
the apparent viscosity at a shear rate ω̇ is equal to the modulus G 2 + G 2 /ω of the
complex viscosity at frequency ω = ω̇ .
Up to now, we have assumed that elasticity was mainly attributable to extension
of individual chains. However, in the concentrated regime, things are a little more
complicated. The random nature of the chain conformations implies that the chains
are totally entangled, looking something like a plate of spaghetti. When it comes to
spaghetti, entanglement blocks the system and it behaves like a viscoelastic solid.
Indeed, a given chain can only move relative to the others by adopting a complex
path that essentially follows its own conformation. There is thus no way to impose
relative motion on the chains in some specific direction without breaking them. With
polymers, however, there is another possibility. They may be able move relative to
one another with the help of thermal agitation, engaging in cooperative motions that
Fig. 4.17 Frequency depen-

dence at constant temperature G'
of a polymer melt or concen-
trated polymer solution for G
different molecular masses
1/θ ω
Fig. 4.18 Reptation of a chain through an entangled polymer network. The initial position of the
chain is the continuous black line and its final position is the thick dashed line. The fictitious tube
in which the chain moves is represented by thin dashed lines
involve several chains at once. This kind of motion was called reptation by Pierre-
Gilles de Gennes, because it is reminiscent of the movement of a snake: a given
meander of the chain moves gradually forward thanks to a series of deformations
occurring simultaneously in the chain and in the network that contains it. When this
meander arrives at the end, the chain will have moved as a whole through the network
(see Fig. 4.18).
Since the entangled network is not altered for small deformations, everything
happens in this regime as if the points of contact within the network were fixed. The
elastic modulus of such a system can thus be estimated using a similar approach to
the one developed for a cross-linked polymer (see Sect. 4.5), which leads to (4.36).
Introducing the average length of the sections expressed as a number Ne of repeat
units and noting that n = γ/a3 Ne , the modulus can be rewritten in the form
kB T
G= γ. (4.51)
Ne a3
In practice, typical values of Ne are of the order of a few hundred carbon bonds.
The characteristic time required to leave the rubber plateau can be estimated by
taking it to correspond to the time when the network changes configuration. In a
154 4 Polymers
sense, the initial elastic structure will have been broken, for it will no longer be able
to recover its initial configuration. This is a somewhat qualitative idea. On a local
level, the reconfiguration of the network is associated with the instant of time when
the chain has just managed to exit its initial tube through the network. In other words,
this chain will have moved a distance equal to its own length. Given that these effects
are attributed to thermal agitation, we may assume that this motion occurs in the form
of a diffusion of the chain through the network, with some diffusion coefficient D.
To a first approximation, the molecule moves along its tube through a medium with
apparent viscosity μc . The resulting viscous friction can be decomposed into an
elementary friction on each section with coefficient μ0 , so that μc = Nμ0 . The
diffusion coefficient is then D = kB T /μc b. In this case, we know (see Chap. 5) that
the diffusion time for a displacement over a distance equal to the length L = Nb of
the tube is given by
L2 μ0 b3 3
θ= = N . (4.52)
D kB T
After the time θ , we may consider that the initial entanglement has been forgotten in
favour of a new one. Note that, up to a constant, θ is N times longer than the return
to equilibrium time of a single chain alone in the liquid [see (4.44)], associated to a
first approximation with the characteristic viscoelastic time of a dilute solution.
Using (4.50) and (4.52), we can estimate the viscosity of the medium, which
turns out to be proportional to N 3 . Globally then, the reptation model predicts a
rapid increase in viscosity with the size of the chains, and this suggests that it does
indeed properly account for the main physical phenomena. However, experimental
results [see (4.49b), i.e., μ → N 3.3 ] do not agree exactly with these predictions. This
may be explained by the fact that the model does not take into account fluctuations
in the length of the tube and the motions of its walls due to thermal agitation of the
molecules in the network.
Among branched polymers, reptation is obviously much more difficult. The mole-
cule must first withdraw one of its branches from the tube containing it. De Gennes
(1979) showed in this case that the relaxation time depends exponentially on the
chain length.
Rheophysical methods are much more delicate in this intermediate regime than in
the two extreme regimes just discussed. A very rough approach is to describe the
mechanical properties of this type of material by treating it as a polymer melt in which
the sections are effectively blobs of the semi-dilute solution. With this assumption,
the elastic modulus can be estimated using (4.51) and taking into account the effective
blob concentration γ = γ/N ◦ in the mixture. Using the results of Sect. 4.4.2, this
leads to
kB T
G= . (4.53)
Ne φ 3
The full relaxation time for the structure can be estimated by extrapolating (4.52),
assuming that the sections now have length φ :
3
μ0 3 N
θ= φ . (4.54)
kB T N
The viscosity then follows from (4.50), using (4.53) and (4.54).
4.7 Effect of Temperature
The temperature dependence of the mechanical properties of polymer materials

depends on how the molecules can arrange themselves, e.g., crystalline or amor-
phous structure, etc. Here we shall mention only a few general trends concerning
polymers in the concentrated regime without chemical or physical bonds between
the molecules. As long as the temperature is high enough, these materials will be in
a liquid state of the kind we have been discussing up to now. In contrast, at a low
enough temperature, they will enter a solid or amorphous state. Now we have seen
that a liquid polymer will appear to behave like a solid at high enough excitation
frequencies. This suggests that there may be an analogy between the impacts of a
temperature reduction or an increased excitation frequency on the apparent behaviour
of a polymer.
When the temperature of a polymer is lowered (or when the excitation frequency is
raised), the disentanglement time Ψt becomes longer than the characteristic flow time,
the chains no longer have time to disentangle, and it is as though the entanglement
were a permanent feature during the flow. Under such conditions, the material appears
to behave as though it were vulcanised (see Sect. 4.5), responding essentially as an
elastic solid. The elastic modulus will then be of the order of 105 Pa.
If the temperature is lowered still further, or if the excitation frequency is further
raised, the spontaneous rotations of the C–C bonds no longer have enough time to
proceed during the characteristic flow time. The chain deformation will then no longer
be entropic but must be forced by the applied stress. The few relative motions of the
chains also involve forced rotations and therefore dissipate a great deal of energy. In
this case, the elastic modulus will be some two to three orders of magnitude higher
than in the previous state.
At even lower temperatures or even shorter excitation times, the rotations of the
C–C bonds become practically impossible. The material becomes very hard and
brittle. At a critical temperature, a discontinuity occurs in the rate of change of the
specific volume. This is the glass transition temperature. For small deformations,
the elastic modulus is typically around 4 × 109 Pa. For a semi-crystalline polymer,
the behaviour is still governed by the interstitial phase of the amorphous polymer,
so the elastic modulus is barely higher.
The analogous changes in behaviour under temperature increase or reduction
in the excitation frequency result from the fact that the apparent behaviour of the
156 4 Polymers
material is essentially governed by the ratio of its (intrinsic) relaxation time and
the characteristic flow time. Increasing the excitation frequency (reciprocal of the
characteristic flow time) or reducing the temperature (which varies as the reciprocal
of the temperature) thus induces qualitatively the same kind of change in the apparent
behaviour. These considerations can be formulated in terms of the temperature and
time dependence of the relaxation modulus:
G(t, T ) = F(aT t, T0 ) , (4.55)
where F(t, T0 ) is the relaxation modulus at a reference temperature T0 and aT is a

temperature-dependent coefficient. It has been shown that aT is given empircally by
the Williams–Landel–Ferry formula
α(T − T0 )
log aT = − . (4.56)
β + T − T0
Taking T0 for the glass transition temperature, we have α = 17.4 and β = 51.6 K.
Note that, if the behaviour of the material were governed by a single relaxation
time then, according to (4.44), (4.45), and (4.55), aT should simply be inversely
proportional to the temperature. The more complex variations actually observed [see
(4.56)] show that there must in fact be several relaxation times.
Further Reading
1. Barnes, H., Hutton, J.F., Walters, K.: An Introduction to Rheology. Rheology

Series. Elsevier, Amsterdam (1989)
2. Daoud, M., Williams, C. (eds.): Soft Matter Physics. Springer, Berlin, Heidelberg
(1999)
3. Flory, P.J.: Statistical Mechanics of Chain Molecules. Hanser Publishers, New
York (1988)
4. de Gennes, P.G.: Scaling Concepts in Polymer Physics. Cornell University Press,
Ithaca (1979)
5. Goodwin, J.W., Hughes, R.W.: Rheology for Chemists—An Introduction. RSC,
Cambridge (2000)
7. Larson, R.G.: The Structure and Rheology of Complex Fluids. Oxford University
Press, Oxford (1999)
8. Oswald, P.: Rheophysics: The Deformation of Flow and Matter. Cambridge
University Press, Cambridge (2009)
9. Tabor, D.: Gases, Liquids and Solids, 3rd edn. Cambridge University Press,
Cambridge (1991)
10. Tirrell, M.: Rheology of Polymeric Liquids. In: Macosko, C.M. (ed.) Rheol-
ogy, Principles, Measurements and Applications, pp. 475–512. Wiley, New York
(1994)
Chapter 5
Colloids
Abstract When the particles in a suspension have sizes in the range from 10 nm to
1 μm, they become liable to the effects of thermal agitation by the molecules of the
interstitial liquid, and at the same time, they can interact at distances of the same
order as their own size, through various kinds of attractive and repulsive forces.
These colloidal effects can play a major role in the mechanical behaviour of the
mixture. The sum total of all the interactions between particles may assume a more
or less complex form, but there are nevertheless two main classes of system: those for
which the balance of forces between the particles is essentially repulsive, and those
for which it is essentially attractive. Rheophysical arguments can explain the origin
of the largely non-Newtonian mechanical behaviour of these two types of system
when the effects of colloidal interactions are predominant.
5.1 Introduction
The phenomenon of sedimentation, or indeed the opposite, creaming, observed in

suspensions, foams, and emulsions (see Chaps. 3 and 6) are drastically reduced and
even become negligible if the particles or inclusions in the liquid are very small, let
us say somewhere between 10 nm and 1 μm. The reasons are twofold. First, objects
in this size range, said to be colloidal, are sensitive to the thermal agitation of the
molecules of the interstitial liquid, and secondly, they can interact at distances of
the same order of magnitude as their own size through various types of repulsive or
attractive forces. The point is that these colloidal effects can counteract the effects
of gravity. They also play a major role in the mechanical behaviour of the mixture. It
is not uncommon for just a few percent of colloidal particles in suspension in a pure
liquid to yield a fluid of apparent viscosity several orders of magnitude higher than
that of the interstitial liquid, while the effect on the viscosity of mixing in the same
fraction of non-colloidal particles would be almost negligible. Moreover, the short

DOI: 10.1007/978-3-319-06148-1_5,
158 5 Colloids
range interactions between colloidal particles give rise to markedly non-Newtonian

mechanical properties.
Various kinds of objects give rise to colloidal effects, including microdroplets in
certain emulsions, giant micelles produced by the association of large numbers of
molecules, and others. The most common case is nevertheless that of solid particles.
In the following, we shall focus on the latter, but the physical characteristics discussed
here will be generally applicable to other types of object. Indeed, the colloidal nature
of an object derives basically from its size and can be significant as soon as the particle
size in at least one direction lies in the range specified above. So there can be colloidal
effects with fibres longer than 1 μm but with significantly smaller diameter. Note,
however, that the upper limit of the colloidal range given here is merely an indication.
In practice, colloidal effects fall off gradually as the particle size increases. In specific
circumstances, there may be significant colloidal effects with particles whose linear
dimensions far exceed the micrometre.
The primary effect of the small size of colloidal particles is their sensitivity to
thermal agitation. This leads to erratic motions of the particles, known as Brownian
motion (see Sect. 5.2). It is generally assumed that the particles are in thermal equilib-
rium with the surrounding liquid, so there is then a sense in which Brownian motion
is an indicator of particle temperature. As for polymers (see Chap. 4), this plays a key
role in the spontaneous evolution of the system, e.g., in relaxation after excitation
and gradual structuring processes when at macroscopic rest (see Sect. 5.10).
Like the molecules in a liquid, these small particles tend to exert attractive forces
on one another, deriving from the van der Waals forces between their atomic con-
stituents (see Sect. 5.3). This is a very basic effect which tends to bring colloidal
particles together and hence cause them to aggregate. However, there are gener-
ally ions adsorbed onto their surface, and these give rise to electrostatic forces
which may hinder aggregation (see Sect. 5.4). Aggregation can also be prevented
by adsorbing long molecular chains onto the particle surfaces (see Sect. 5.5). Finally,
if there are polymers dissolved in the interstitial liquid, this can bring about further
attractive effects (see Sect. 5.6). The system is said to be stable whenever the interac-
tions between particles allow them to distribute themselves uniformly throughout the
liquid.
Taking into account all these interactions, the result may be rather complex, but
one can nevertheless distinguish two main classes of system: those for which the
balance of forces between the particles (see Sect. 5.7) is basically repulsive, and
those for which this balance of forces is basically attractive. Rheophysical arguments
can explain the generally markedly non-Newtonian mechanical behaviour of these
two types of system (see Sects. 5.8 and 5.9) when colloidal interactions dominate.
Repulsive systems are essentially simple yield stress fluids, while attractive systems
are essentially thixotropic yield stress fluids. When hydrodynamic interactions due
to flow of the interstitial liquid are predominant, the properties of these systems tend
toward those of simple liquids (see Sect. 5.10).
5.2 Brownian Motion 159
Fig. 5.1 Molecules of a

liquid (black) colliding with
a colloidal particle (grey)
suspended in the liquid.
Arrows show motions of
molecules leading to contact
with the particle
5.2 Brownian Motion
5.2.1 Basic Principles
5.2.1.1 General Considerations
A solid particle placed in a liquid is surrounded by a large number of molecules

in constant thermal agitation (see Sect. 2.2.2), but it is itself made up of atoms or
molecules subjected to the same thermal agitation. Given that the motions of the
elements making up the particle have much smaller amplitudes than those making
up the liquid, the situation is usually described by treating the particle as a perfectly
uniform rigid body whose surface is permanently subjected to collisions with the
liquid molecules (see Fig. 5.1).
When the system is at rest macroscopically, e.g., a suspension at rest in a
container, the molecules do not favour any particular direction as they move around.
As a consequence, the time average of the total force exerted by the molecules on the
particle due to these collisions will be zero. However, over time scales of the order of
the characteristic time of one collision, which is of the order of θ = 10−13 s, the total
force will fluctuate significantly in a way that depends on the instantaneous magni-
tudes and directions of the molecular velocities. If the particle inertia is sufficiently
low, i.e., the particle is small, the particle motion is thus governed at any given time
by a force of different magnitude and direction to the one affecting it the moment
before. The particle ends up following a quite unpredictable trajectory through the
liquid (see Fig. 5.2). This is called Brownian motion.
Although Brownian motion is completely unpredictable on a local and instanta-
neous scale, certain features of the motion can be quantified over long enough times
by making two key physical assumptions. The first is the independence or decorre-
lation of the successive changes in velocity and position of the particle. This is due
to the decorrelation of the successive values of the total instantaneous force. Sup-
pose then that we describe the particle motion by the series of positions it occupies
after successive time intervals ◦ θ . The resulting motion looks just like a random
walk (see Sect. 4.2.2) in the sense that, in each step, the particle displacement is
160 5 Colloids
Fig. 5.2 Random trajectory (continuous line) followed by a particle as time goes by under the
effects of Brownian agitation, leading to an overall displacement r shown by the dashed line. Given
the fluctuations in the total force over very short time scales, each linear section shown here is a
simplified representation of the motion on the scale of observation. On a finer scale, the trajectory
of the particle over such a section would in fact be made up of smaller segments also pointing in
various directions
totally random. In particular, it is completely independent of its current position and

the previous history of its motion. As a consequence, after a rather long time, the
average displacement is zero, because the particle is just as likely to occupy any two
positions at equal distances in opposite directions. However, the average distance
between the particle and its initial position will increase as time goes by, because the
particle explores more and more positions in the space around its initial position.
The second key hypothesis is that the particle is in thermal equilibrium in the
liquid, i.e., it has the same mechanical energy as the medium in which it is suspended.
More precisely, its energy is distributed according to the Boltzmann distribution (see
Sect. 2.3.1) for the given temperature. Now the mechanical energy of the liquid is
fully reflected in the agitation of its molecules, which have average kinetic energy
3kB T /2 (see Sect. 2.4), an expression that does not depend on the mass of the given
object. We shall thus assume that the average kinetic energy (of thermal agitation)
of the particle is also 3kB T /2.
5.2.1.2 Average Displacement
We now consider the displacement r(t), i.e., the vector from the initial position of
the particle to its position at time t (see Fig. 5.2). This vector has length equal to the
distance travelled by the particle since the initial time, which we shall write |r| =
√
r2 . We seek the average distance reached after a given time. Since the collisions
are random, if we track any given particle, this distance will vary enormously. If
we consider many identical particles under the same conditions in the liquid, we
obtain a probability distribution p(r) for the particle displacements. To quantify this
displacement of the particles under the effects of thermal agitation, we must consider
a certain average of the particle displacement that takes this distribution into account.
In the present situation, given the diffusion effects produced by Brownian motion,
it turns out to be more convenient to consider the variation of the mean squared
distance ∞r2 = p(r)r2 d3 r than the average distance.
The motion of each particle is governed by Newton’s second law. At any given
time, the particle feels a fluctuating force Ffluc due to collisions with molecules in the
liquid. At the same time, as it moves through the liquid of viscosity μ, any particle is
also subject to a viscous drag force. To a first approximation, i.e., neglecting inertial
effects, this force can be described as the force associated with a steady-state motion
at the instantaneous velocity ṙ(t) of the particle (see Sect. 3.2.6), which would be
kdμṙ, where d is a characteristic linear dimension of the particle. The equation of
motion of the particle is thus
d2 r dr
m 2
= Ffluc − kμd . (5.1)
dt dt
To simplify the discussion, we decompose the motion in space along three perpen-
dicular axes and describe the particle position in terms of its three components x, y, z
in this frame. Equation (5.1) then decomposes into three similar equations, one for
each direction. Note in passing that r2 = x 2 + y 2 + z 2 and, since no direction is
favoured, we may deduce that
1 2
∞x 2 = ∞y 2 = ∞z 2 = ∞r .
3
Consider for example the x component of (5.1). Multiplying this equation by x and
noting that
d(x 2 ) dx
= 2x ,
dt dt
we obtain
d2 x 1 d(x 2 )
mx = x Fx,fluc − kμd . (5.2)
dt 2 2 dt
Using the identity
2
d2 x d dx dx
x = x − ,
dt 2 dt dt dt
then taking the average, (5.2) can be rewritten in the form1
1 Note that
df df 3 d d∞ f
= p(r ) d r= p(r ) f d3 r = .
dt dt dt dt
162 5 Colloids

d dx dx 2 1 d
m x −m = x Fx,fluc − kμd x 2 .
dt dt dt 2 dt
The first term on each side is zero because the velocities, positions, and fluctuating
force are decorrelated, so the average of their product is equal to the product of their
averages and each of these averages is itself zero. Doing the same with the other
two components, summing, and using the assumptions of thermal equilibrium and
equipartition of energy, i.e., the average kinetic energy of each particle is 3kB T /2,
which leads here to

dx 2 1 dr 2
m = m = kB T ,
dt 3 dt
we end up with
1 d
3kB T = kμd ∞r2 ,
2 dt
which can be integrated to give
6kB T
∞r2 = t, (5.3)
kμd
assuming that the particle is at r = 0 at t = 0. This expression shows that the mean
squared distance is simply proportional to the time elapsed since the initial time and
that it increases as the temperature of the medium is raised or its viscosity reduced.
5.2.2 Diffusion
The very simple expression just deduced for the mean squared distance hides a
more complex reality with regard to individual particles, since each particle moves
spontaneously in all directions in the liquid, while the positions reached on average
over the whole particle ensemble move further and further away from their points of
departure. This is diffusive motion. The coefficient of proportionality
kB T
D= (5.4)
kμd
which arises in the expression (5.3) for the mean squared distance is called the
diffusion coefficient. From this last equation we may also deduce the time taken to
reach a distance of the order of the particle size, i.e., ∞r2 ⇒ d 2 :
kμd 3
t0 ⇒ . (5.5)
kB T
For beads of radius d = R = 0.1 μm in water at room temperature, we find a time

of the order of 4 × 10−3 s. In other words, compared with their size, small particles
diffuse extremely quickly through a liquid. This time increases very quickly with the
size of the particles, reaching 4 s for 1 μm beads, and 4 000 s for 10 μm beads. So in
practice Brownian diffusion is negligible for particle sizes greater than a few tens of
microns, since no perceptible motion would be recorded during a typical experiment.
Given the variability of the motion from one particle to another, it would only be
with very many particles in suspension that one could easily observe this diffusion
phenomenon. A group of particles placed in some specific region of a pure liquid
will thus form a kind of cloud spreading out through the liquid at a rate related to the
particle size. But one could not identify any kind of displacement front because the
probability of finding a particle at a given point decreases gradually with the distance
and, statistically speaking, it is always possible that there will be some particles very
far from the initial position even after a very short diffusion time.
As an example, let us investigate the way a particle ensemble will distribute
itself in a suspension in the simple case of one-dimensional diffusion. The amount of
particles in a volume element is described by its concentration ξ, from which we may
ascertain the number of particles in a volume dα by evaluating ξdα. To simplify,
we assume that ξ only varies in the x direction. According to (5.3), we know that,
during a time dt, the particles travel a distance dx such that dx 2 = 2Ddt. Consider a
slab of thickness dx at x. Since (5.3) only involves the squared distance, the particles
in this slab have probability 1/2 of moving a distance dx in either direction, and
the same goes for the particles located in the two neighbouring slabs at x − dx and
x + dx. It follows that the change in the number of particles in the slab at x after a
time dt is given by
1
ξ(x − dx) − 2ξ(x) + ξ(x + dx) .
2
Furthermore, using the definition of the concentration, the change in the number of
particles in this slab can also be written ξ(t + dt) − ξ(t) dα. Equating these two
expressions and dividing the right-hand side by 2Ddt and the left-hand side by dx 2 ,
then taking the limit dt √ 0, we obtain the following equation, known as Fick’s
second law:
βξ β 2ξ
=D 2 . (5.6)
βt βx
More generally, this equation describes the way an initially heterogeneous suspen-
sion becomes homogeneous as time goes by due to the steady redistribution of the
concentration brought about by particle diffusion. To get some idea of how this works,
we consider the special case of a concentration peak located initially at the origin,
i.e., ξ(x, 0) = ∂(x), where ∂(x) is the Dirac distribution2 and observe the diffusion

2 ∂ is a function defined by the properties ∂(x = 1) = 0 and ∂(x)dx = 1.
164 5 Colloids
Fig. 5.3 1D diffusion profiles at different times for particles starting out from zero. The profiles are
shown as a function of the dimensionless position and time variables X = x/R 2 and T = Dt/R 2 =
t/t0 . Equation (5.7) which governs this diffusion can then be written in the form ξ(X, T ) =
(4θ T )−1/2 exp(−X 2 /4T ). The initial profile is a Dirac delta function at X = 0 and the other
profiles correspond, from the next sharpest to the most spread out, to times 0.05, 0.25, 1, 2, 5
of the particles from this peak. This provides a good model for what happens when a
droplet of a liquid containing a high particle concentration is placed in a large volume
of the same pure liquid. It is easy to check that in this case (5.6) has solution
exp(−x 2 /4Dt)
ξ(x, t) = . (5.7)
(4θ Dt)1/2
The concentration profile described by this equation is shown in Fig. 5.3. As time
goes by, the central peak gradually drops. In fact its height goes as (4θ Dt)−1/2 and
the particle distribution steadily becomes uniform.
5.2.3 Rotational Diffusion
The directions of the forces due to collisions of the liquid molecules with a given
particle do not necessarily pass through the centre of mass of the particle. This
causes the particle to rotate about its centre of mass. Formally, the equation for these
rotations is analogous to (5.1) up to a length factor, replacing the left-hand side by
the angular momentum of the particle, the first term on the right-hand side by the
fluctuating torque, and the second term on the right-hand side by the viscous drag
torque. The mathematical treatment of this phenomenon thus follows the same lines
as for translational Brownian motion and leads to a diffusion equation for the angle of
rotation θ in a given plane, viz., ∞θ 2 = Drot t, where the rotational diffusion constant
is given by
kB T
Drot = ϕ , (5.8)
μL 3
with L the length of the axis whose rotation is observed and ϕ a factor depending on
the shape of the object.
5.2.4 Osmotic Pressure
Since colloidal objects suspended in a liquid diffuse under the effect of thermal
agitation, they tend to maximise their occupation of the available space in the liquid.
The ceaseless motions of these elements look very like the motions of molecules in
a gas. Their kinetic energy is of thermal origin and there is no favoured direction
of motion. Carrying out similar calculations to those in Sect. 2.3.4 for the pressure
exerted by a perfect gas on a solid wall, we can thus find the pressure exerted specif-
ically by the colloidal particles, which yields

2
1 dr
p = nm ,
3 dt
where n = ξ/v is the number of particles per unit volume, with v the average volume
of the particles. According to the assumption of thermal equilibrium, we have

2
dr 3kB T
= ,
dt m
from which we may deduce the value of the specific pressure exerted by the colloidal
particles. This is called the osmotic pressure:
Posm = nkB T . (5.9)
This pressure can be measured in a suspension by preventing the particles from

reaching a certain region while leaving the way open for the liquid molecules, typi-
cally using a filter with much finer mesh than the particle size but porous to the liquid
molecules. The liquid phase containing the particles exerts a higher pressure on the
filter than the pure liquid phase, precisely because it is increased by the osmotic
pressure.
5.2.5 Sedimentation and Brownian Diffusion
Diffusion can counteract sedimentation or creaming to some extent. Under the action
of gravity, the particles tend to accumulate in a given region, while Brownian diffusion
tends to redistribute them homogeneously throughout the liquid. When both effects
occur at the same time, a certain equilibrium profile is obtained, associated with the
166 5 Colloids
Fig. 5.4 Suspension of

Brownian particles that are
more dense than the host
liquid. The concentration in
the container, represented by
grey level here, falls off with
increasing height x
x0
balance between these two trends (see Fig. 5.4). In a vertical cylindrical container,
the equilibrium profile is given by a function ξ(x) expressing the concentration as
a function of the height. Consider a cross-section of thickness dx at height x. The
change in the number of particles in this cross-section over a time interval dt as a
result of sedimentation is equal to the difference between the amount of particles
arriving from the upper layer and the amount leaving this layer, i.e.,
βξ
πsedimentation = Vfall ξ(x + dx) − ξ(x) dt = Vfall dt .
βx
In this expression, we shall simplify by using the speed of fall of a particle in the
steady-state regime, given by balancing the force of gravity gvλΩ and the drag
force kdμVfall (see Sect. 3.2.5), which yields Vfall = gvλΩ/kdμ. The change in the
number of particles in this layer due to diffusion can be found directly from (5.6):
β 2ξ
πdiffusion = D dt .
βx2
The equilibrium profile is obtained when the particle concentration in the layer is not
changing, i.e., when πdiffusion + πsedimentation = 0. This is easily solved, yielding

gvλΩx
ξ(x) = ξ0 exp − . (5.10)
kB T
At first glance it is surprising to see that this profile does not depend on the friction
between the liquid and the moving particles, i.e., μ does not appear in (5.10). How-
ever, this is easily explained by the fact that viscous effects will have the same effect
on both types of motion. The result seems even more natural when we note that this
profile has the form of a Boltzmann distribution (see Appendix B.5) where the role
of the potential energy is played by the mechanical potential energy associated with
gravity, i.e., gvλΩ. It is just as though, at equilibrium, the particles came to occupy
different potential energy levels associated with different heights above the bottom
of the container while continuing to maintain exchanges between different levels by
diffusion.
Let us look a little more closely at this concentration profile. Over a thickness
of the order of x0 = kB T /gvλΩ, the concentration is of the order of the maximum
concentration. Slightly beyond this distance, the concentration is close to zero. This
thickness increases as the particles get smaller, and varies very quickly with the par-
ticle size, going as d −3 . For spheres of diameter 0.1 μm and density twice the density
of water, the critical thickness is around 100 μm, so in this situation sedimentation
will be significant under typical experimental conditions (for a container of height
10 cm). However, for similar spheres of diameter 0.01 μm, this thickness increases
to 10 cm, whence sedimentation will be negligible.
5.3 Van der Waals Forces
As discussed in Sect. 2.2.4, atoms and molecules are subjected to van der Waals
attractive forces which fall off very rapidly with increasing distance. A colloidal
particle is made up of a very large number of atoms which can interact at a distance
with all the atoms constituting a neighbouring particle. In order to calculate the
resulting interaction, we may to a first approximation sum the mutual interactions
between pairs of atoms of each of the particles. To this end, we assume that the
interaction potential between atoms has the form −C/r 6 (see Sect. 2.2.3). We first
consider the case of an atom located at the point O and a particle occupying all the
space from a plane situated at distance d from this atom (see Fig. 5.5) and made up of
atoms with number density Ω. The interaction potential between the single atom and
a ring of radius x, thickness dz, and width dx with centre at distance z > d from the
atom is given by −CΩ2θ xdxdz/(x 2 + z 2 )3 . Integrating over all the rings contained
in the half-plane x ≈ [0, ∝) and z ≈ [d, ∝), we find the total potential
θΩC
U =− . (5.11)
6d 3
If instead of the single atom we have a particle of surface area A and infinite thickness
(see Fig. 5.5), the potential is found by integrating the expression (5.11) over d ≈
[h, ∝), which yields
KA
U =− , (5.12)
12θ h 2
where K = θ 2 Ω 2 C is the Hamaker constant, which has dimensions of energy and

is in fact a property of the material. The above expression can be used to estimate
the interaction potential between two particles with parallel plane surfaces of area
168 5 Colloids
Fig. 5.5 Calculating the van

der Waals interaction between
a molecule O and a material
occupying all space from
the plane situated at distance
d from this molecule. We
sum the interaction energy dx
between this molecule and all
the molecules contained in dz r
disks of radius x and located
z O
at distance z > d
x
A situated a distance h apart, where h is much smaller than their dimensions in the
z direction. Indeed, in this case, the elements situated at the ends in the directions
perpendicular to z and those situated at much greater distances than h will make a
negligible contribution to the total potential, so most of the particle elements will
‘see’ an infinite solid surface.
For two spherical particles with distance r = 2R + h between their centres, the
calculation is slightly more involved and leads to

K 2R 2 2R 2 r 2 − 4R 2
νa = − + + ln ,
6 r 2 − 4R 2 r2 r2
which for small distances h ∇ R simplifies to
KR
νa ⇒ − . (5.13)
12h
In fact the underlying assumptions for these calculations are not fully valid in the
general case. In particular, the assumption that one can simply add together inter-
actions between dipoles is not justified since these dipoles all influence one another
mutually and even within each particle. Furthermore, at large enough distances, the
retardation effect mentioned in Sect. 2.2.4 plays a non-negligible role, implying that
the interaction potential no longer goes as h −6 , but rather as h −7 . In the end, a full and
accurate analysis, e.g., based on the Lifshitz theory, requires a detailed knowledge
of the electrical properties of the materials. However, a good approximation can be
5.3 Van der Waals Forces 169
obtained by neglecting retardation and using simple expressions like (5.11) or (5.13)
with values for the Hamaker constant derived from the Lifshitz theory.
The value of the Hamaker constant K is crucial in this analysis. It can be estimated
using the value of C determined in Chap. 2. Taking into account the fact that the
polarisability of an atom (or a spherical molecule) of radius R is ϕ = 4θ χ0 R 3 , we
have
3
K = vR 6 θ 2 Ω 2 .
4
For atoms in a cubic arrangement, we have Ω = (2R)−3 and in addition v is of the

order of the first ionisation potential, viz., 5 × 10−10 eV. For most materials, we then
find that K lies between 0.4 and 4 × 10−19 J, with the highest values occurring in
metals.
Under the effect of the attractive van der Waals interaction, two particles tend
to move together and would eventually come into contact. In practice, however,
they cannot come closer than a few times the interatomic distance, of the order
of h = 10−10 m, owing to the Born repulsion which arises due to the fact that
their electron clouds cannot interpenetrate. The minimum interaction potential (of
maximum magnitude) is thus attained at this distance. For two spheres of diameter
100 nm in contact with K = 10−20 J, we then find νa = 3 × 10−19 J, which is
about 70kB T . It follows that, when two particles like this touch one another, they can
no longer move apart solely under the action of thermal agitation, since the energy
attracting them together is significantly greater than their thermal kinetic energy.
However, they can be prevented from entering into contact and hence acquiring
such a high interaction energy, simply by keeping them a certain distance apart. To
achieve this, repulsive forces must be introduced. There are two main ways. One is
to add ions in solution which then create electrostatic forces between the particles
(see Sect. 5.4), and the other is to adsorb polymers onto the particle surfaces leading
to steric hindrance (see Sect. 5.5).
Note that, if the above calculation of the maximum energy is attempted for very
large particles, i.e., such that A is of the order of a few square millimetres, we
obtain very high values, in particular for plane particles in contact along their largest
surface, because the ratio A/ h 2 is then very high. However, in practice, such attractive
forces are never obtained between two macroscopic solid bodies because their surface
always has some degree of roughness, so the effective points of contact only ever
occupy a negligible area compared with the total surface area (see Sect. 7.2.2). On
the other hand, with deformable solids like cling films which better conform to the
surface of the objects they come into contact with, and which adhere well to a whole
range of solid surfaces, a large effective contact area is then possible, leading to
strong adhesion.
170 5 Colloids
Fig. 5.6 Upper Spatial dis-

tribution of positive and
negative ions in the dou-
ble layer formed on the
solid surface (thick black
line). Lower Concentration of
co-ions and counter-ions and
total concentration as a func-
tion of the distance from the
surface
Ion
concentration
Counter-ions
Co-ions
Distance/Debye length
5.4 Electrostatic Forces
For most natural colloidal systems, the repulsive inter particle forces are due to the
presence of ionised species on the particle surfaces. These surface charges can have
various origins. They may arise due to irregularities in the internal structure leading
to a charge deficit which is then balanced by a surplus of the opposite charge at the
surface. This is what happens in clays. There may also be ionisable groups which
can only occur at the particle surface, such as Si–OH groups on silica surfaces.
Another possibility is that ions in solution may adsorb onto the particle surfaces.
These different sources indicate how difficult it is to control the surface charge,
which may be highly sensitive to impurities whose origin and composition are poorly
ascertained.
Despite the presence of these surface charges, colloidal suspensions remain elec-
trically neutral. However, particles with charged surfaces can be expected to interact
with each other through Coulombic electrostatic interactions. In fact, in polar solu-
tions, i.e., aqueous solutions, ions in solution play the role of counter-ions which are
attracted to the surface, while ions of opposite sign are repelled by the surface (see
Fig. 5.6). This so-called double-layer effect tends to neutralise the solid surfaces in a
somewhat diffuse way within the solution. Far from the surface, ion concentrations
always balance. In the end, the interactions between particles depend on the way the
charges distribute themselves throughout space.
5.4 Electrostatic Forces 171
Let us investigate how the charges will be distributed through the liquid in a simple
case: an infinite uniformly charged plane solid surface. For this configuration, the
variables can only depend on the distance x from the plane surface. The spatial
charge distribution leads to a potential γ in the solution which satisfies the laws
of electrostatics. Furthermore, the charge distribution is governed by the laws of
thermodynamics. The value of γ can be derived by considering the combination of
these two phenomena.
We assume that the ions in the solvent are point charges, in fact, a number z of
charges ±e. We consider the electrostatic field E = −∇γ derived from the potential
γ. The force on a given ion due to this electrostatic field is F = zeE, whence
zeγ is the electrostatic potential of this ion in space. Inserting the above relation
between E and γ into Gauss’ theorem, which expresses the relation between the
electrostatic field and the spatial distribution of charges, i.e., ∇ · E = Ω/χ, where χ is
the permittivity of the medium, we obtain Poisson’s law λγ = −Ω/χ, which relates
the electrostatic potential to the spatial charge distribution. In the present case, by
symmetry, γ is a function of x alone, whence
β 2γ Ω
=− . (5.14)
βx2 χ
We now apply the laws of statistical thermodynamics to the ions, assuming that they
are distributed in the solvent according to the Boltzmann distribution (see Appen-
dix B.5) as a function of the local potential energy. When the potential is γ, the
probability of finding one ion is therefore proportional to

±zeγ
exp − .
kB T
If the ion concentration in the solution is n 0 , we deduce that the positive or negative
ion concentration at some specific point is

±zeγ
n ± = n 0 exp − . (5.15)
kB T
Note that this is consistent with the boundary conditions, since far from the solid
surface the potential is zero and the positive and negative ions should have the same
concentration n 0 . The charge density is related to the ion concentrations according to

zeγ zeγ
Ω = ze(n + − n − ) = zen 0 exp − − exp . (5.16)
kB T kB T
Putting together (5.14) and (5.16), we obtain the equation for the spatial distribution
of the potential γ, viz.,
β 2γ 2zen 0 zeγ
= sinh . (5.17)
βx 2 χ kB T
172 5 Colloids
This must be solved with the boundary conditions that both γ and βγ/β x tend to
zero as x tends to infinity. For small potentials zeγ/kB T ∇ 1, the solution is
γ = γ0 exp(−τ x), (5.18)
where γ0 is the potential at the particle surface and

1/2
2e2 z 2 n 0
τ −1 = (5.19)
χkB T
is the Debye length. The variation of the charge density and shape of the potential
predicted by this model are shown in Fig. 5.6. Note that the variation of the potential
is slight beyond a distance τ −1 , implying that this length is like a screening length
for the electric field associated with a particle’s surface charges.
The surface potential γ0 can be related to the surface charge density by equating
the surface
∝ charge with the total charge per unit area within the double layer, i.e.,
η0 = − 0 Ωdx. Introducing the value of γ into the expression (5.16) for the charge,
we then obtain
zeγ0
η0 = (8n 0 χkB T )1/2 sinh , (5.20)
2kB T
which reduces to η0 = χτγ0 for weak potentials. We thus see that the surface
potential depends on the charge density η0 , but also the ionic composition of the
medium through τ.
When two charged surfaces come close enough to one another within a polar
liquid, their double layers can no longer maintain the configuration described above.
In a certain sense, they interact, but this interaction is not direct. The ions must dis-
tribute themselves in some other way, taking into account the proximity of the solid
surfaces. There thus occurs an effect rather like osmosis due to this confinement,
leading to a repulsive force between the solid surfaces. The corresponding calcula-
tions go beyond the scope of this book, but we nevertheless mention the resulting
formulas. In the weak potential limit and for large distances between particles, i.e.,
exp(−τh) ∇ 1, the interaction potential per unit area is
νe = 2τχγ02 exp(−τh) , (5.21)
for two plane surfaces, and
νe = 2θ Rχγ02 exp(−τh) , (5.22)
for two spheres. We thus find the same type of variation for the electrostatic interaction
potential as for the electrostatic potential around a solid surface, viz., a decaying
exponential with characteristic decay length equal to the Debye length τ −1 .
5.5 Effects Due to Adsorbed Polymers 173
Fig. 5.7 Reduction in the

number of possible positions
of a rod fixed at one of its
ends onto a solid surface (left)
s
when another solid surface is
brought within range (right) l
θ
x
5.5 Effects Due to Adsorbed Polymers
One technique commonly used to stabilise colloids is to coat the particle surfaces
with polymers. The chains adsorb onto the surface by one end, while the other end
is partly free to move around in the solution. The presence of these polymer chains
adsorbed onto the particles makes it much harder for them to approach one another
at short range. The corresponding repulsive force comes from the energy needed to
reduce the configurational entropy (see Sect. 4.3) of the chains that find themselves
between the two solid surfaces. By bringing the particles close together, we reduce the
space available to the polymers and thus also the number of accessible geometrical
configurations.
This effect can be understood through the following highly simplified description.
Consider a rigid molecule like a cylindrical rod of length l and cross-sectional area s
attached to a plane surface as shown in Fig. 5.7. This molecule can occupy different
positions in space, characterised by the position of its endpoint on the hemisphere
of radius l. To simplify, we divide this hemisphere into elements of area s and
assume that the end of the molecule can only occupy positions corresponding to
these elements. There are thus 2θl 2 /s possible positions for the molecule. If a solid
surface is now brought to within a distance x < l of the first surface as shown in
Fig. 5.7, the number of available positions is reduced. The area of the hemispherical
surface still available to the end of the rod is calculated by integrating the surface
elements located between y and y + dy, for y ≈ [0, x]. These elements, associated
with an angle θ relative to the solid
xsurface, have circumference 2θl cos θ and width
dy/ cos θ . The total area is thus 0 2θldy = 2θlx. It follows that the number of
possible positions varies from Z 0 = 2θl 2 /s to Z 1 = 2θlx/s, implying that the
configurational entropy is given up to a constant by
Z1 x
S = kB ln = kB ln .
Z0 l
174 5 Colloids
Fig. 5.8 Appearance of Rg L

chains adsorbed onto a particle
surface. When the number of
chains per unit area φ is low
enough (left), the chains do
not get in each other’s way and
typically have lengths close to
their radius of gyration. But
when this number reaches a
certain level (right), the chains
tend to stretch out to increase
the entropy of the system
Γ<1 Γ>1
The free energy associated with this change in configurational entropy is F = −T S.

The work done in bringing the two surfaces together is given by δW = f → dx, where
f → is the external force on the two plates. This work is equal to the change in free
energy dF of the system at constant temperature (see Appendix B.4), whence the
repulsive force f = − f → between the two plates is
dF kB T
f =− = .
dx x
We thus obtain an estimate of the repulsive force between two solid planes with a
polymer chain adsorbed onto one of them. This calculation can be extended to the
case of several chains adsorbed onto each surface, to estimate the repulsive force
between the two surfaces, but we must then take into account several other effects.
For one polymer chain, considering the possible rotations of the bonds between
successive elements (see Sect. 4.2), the number of possible configurations does not
depend solely on the distance between the two solid surfaces as assumed above.
Furthermore, when a high enough concentration of polymers is adsorbed onto one
of the surfaces, the chains begin to get in each other’s way across this surface. If they
were to keep the same form as in a dilute solution, they would be highly entangled.
In order to maximise their entropy, the chains tend to occupy a larger region of space
above the solid surface, whence they begin to extend outward in the direction normal
to the solid surface, producing a kind of ‘brush’ formation (see Fig. 5.8). As discussed
above, when two such polymer brushes on two solid surfaces are brought together,
the configurational entropy of the system is reduced, so energy must be supplied.
The chains have less space to stretch out and must in the end work their way in
amongst one another.
Note that, in order for two polymer-coated particles to repel one another, the
liquid has to be a good solvent for the polymer. If this is not the case, i.e., if
there is a certain attraction between the polymer segments that would otherwise
lead to a phase separation if the polymers were alone in the solution, the polymer
induces an attractive interaction between the particles. Finally, and quite generally, the
5.5 Effects Due to Adsorbed Polymers 175
interaction between two polymer-coated particles is strongly dependent on the

distance to which the polymers extend out into the liquid. This distance is primarily
controlled by the length of the chains, but also to some extent by the density of
adsorbed chains and by the interactions between polymer and liquid.
The importance of the density and the affinity of the polymer for the liquid can be
ascertained through the following simplified analysis. When the chains are far enough
apart, they do not disturb one another and the chain length is in practice close to the
radius of gyration Rg in a dilute solution, where we have Rg ∝ N 3/5 in a good
solvent (see Sect. 4.3.3). When the chain concentration is high enough, they begin to
get in each other’s way and hence stretch out to a length L, the brush thickness, to
occupy a little more space. We may thus take up the calculations of Sect. 4.3. The
energy associated with stretching the chain to length L is E = 3kB T L 2 /2N b2 . The
volume available per chain is Lb2 /n, where n/b2 = φ is the number of chains per
unit area, whence the first energy term in (4.24) is
nN2
E = kB T b3 (1 − 2ε ) ,
2Lb2
and the total energy is
nN2 3kB T L 2
E = kB T b3 (1 − 2ε ) + + const. (5.23)
2Lb2 2N b2
This is minimal for 1/3
n(1 − 2ε )
L⇒ bN . (5.24)
6
The chains thus extend more rapidly than the usual growth as N 3/5 for a chain in a
dilute suspension.
More sophisticated calculations (see Israelachvili 1991) can estimate the potential
√ surfaces for a thin coating of polymers (φ < 1, i.e., the
per unit area between solid
interchain distance b/ n is greater than the radius of gyration Rg ), whence each
chain interacts freely with the other solid surface:

h
νs ⇒ 36φ kB T exp − , 2Rg < h < 8Rg . (5.25)
Rg
√
For a dense coating, i.e., when a/ n < Rg , we have

100L 3/2 θh
νs ⇒ φ kB T exp − , 0.4L < h < 1.8L . (5.26)
θ L
The potential subsequently decreases very quickly when h > 2L, i.e., when the
chains have little chance of coming into contact.
176 5 Colloids
Fig. 5.9 Depletion, i.e., lower concentration than the average in the liquid as a whole, of certain
entities (small disks) in the region between two neighbouring particles
5.6 Depletion Interactions
This type of interaction occurs when, apart from the main particles, the suspension
contains some species with entities of intermediate size, let’s say d, somewhere
between these particles and the solvent molecules. The most common case is a
polymer in solution which does not adsorb onto the particle surface. We may then
consider that, to a first approximation, d is equal to the radius of gyration Rg of the
molecules (see Sect. 4.2.3). The centres of these entities cannot lie at distances less
than d/2 from the particle surfaces. In this region, their concentration is therefore
depleted. When the particles are far removed from one another, this phenomenon has
no particular consequence. On the other hand, when two particles are rather close
to one another, there is a region of low concentration in these entities between the
particles, or even a total absence, when h is less than d, and such regions thus differ
from the rest of the suspension (see Fig. 5.9). The difference in osmotic pressure
between the two regions tends to equalise the concentration by bringing the colloidal
particles together, thereby inducing an attractive force.
The interaction energy can be estimated by taking it as the work needed to separate
the two particles to infinity against osmotic pressure. We know that, when the particles
are far enough apart, they do not interact at all through this mechanism, because the
concentration of the added species will then be uniform throughout the space between
them. To a first approximation, we assume that the energy is significant only when
the depletion layers overlap, i.e., when h lies between 0 and d. The energy is thus
assumed negligible when h > d. For entities not interacting in the liquid, the osmotic
pressure in the rest of the solvent is given by (5.9). The entities in suspension can
be treated as the molecules of a gas in a situation where we reduce the available
volume by separating the two colloidal particles over this range of distance, as long
as the entities cannot enter the intermediate region. The interaction energy is the
5.6 Depletion Interactions 177
work required to go from distance h to distance d, which can be estimated from the
work done by the pressure in a gas, viz., − pdα. In the present case, this gives
ν = −Posm dα(h), (5.27)
where dα(h) is the displaced volume, which is also the volume of the intersection of
the two depletion layers, when the distance is h. For particles and spherical entities,
dα(h) is twice the volume of a spherical cap (see Fig. 5.9). This √ can be calculated
by summing the volumes of disks of thickness dx and radius R 2 − x 2 between
x = R − d/2 − h/2 and R. When d, h ∇ R, this yields
θ
dα(h) ⇒ R(d − h)2 ,
2
and we obtain finally
θ
νd = − nkB T R(d − h)2 . (5.28)
2
This interaction is never very strong, but it may be of the order of kB T if the concen-
tration of the entities is high enough, whence an increasing tendency to aggregate.
5.7 Balance of Interactions
Whatever the characteristics of the given system, colloidal particles will be subjected
to attractive van der Waals forces which tend to bring them to within very short range
of one another. In addition, there may be a depletion effect due to the presence of
some other species, often a polymer, in suspension in the liquid, and this will also
contribute to the attraction between particles. To counteract these effects, there are
basically two possibilities. One is to get ions onto the particle surfaces, thereby
inducing an electrostatic repulsion, and the other is to adsorb polymers onto their
surfaces to induce steric repulsion. When several of the above effects occur together,
the total interaction potential between particles is simply the sum of the potentials
resulting from each of the interactions:
ν = ν a + νe + νs + νd . (5.29)
Figure 5.10 illustrates this idea with just two types of interaction, i.e., van der Waals
and electrostatic. The potential due to the van der Waals interactions is usually attrac-
tive. When graphed as a function of the interparticle distance, the energy increases
with distance, so the interaction force f = −dν/dh deriving from this potential is
negative. The attractive force tends to infinity when the distance is very small and
it is then much greater than the electrostatic repulsion force. However, as discussed
above, the closest approach between the particles is limited by the extent of the elec-
tron clouds, i.e., the Born repulsion. On the other hand, the electrostatic potential is
178 5 Colloids
Fig. 5.10 Typical profile

of the two main interaction
potentials νa and νe and
also the total interaction
potential as a function of the
distance between two colloidal
particles in suspension in a
liquid. The curves plotted here
were obtained from (5.11)
and (5.20) with arbitrary
coefficients
repulsive, but at short range it tends to a finite value according to (5.21) and (5.22).
Finally, the total potential is dominated by the van der Waals attraction at short and
long range, but the repulsion may be dominant in the middle range. In the example of
Fig. 5.10, there is thus a potential barrier at h 0 , and this tends to prevent the particles
from coming too close together.
In practice, the total potential can assume different forms depending on the values
of the attractive and repulsive potentials. When the attractive potential dominates the
repulsive potential over the whole range, the total potential will obviously remain
attractive (case A in Fig. 5.11). The particles come within very close range and ther-
mal agitation or ordinary mechanical action will be unable to separate them. They then
aggregate, occupying a much smaller volume within the liquid, and the suspension is
therefore unstable. In this case, we speak of flocculation. When the repulsive poten-
tial is higher, there will be an energy barrier (cases B and C in Fig. 5.11). If this
barrier is of the order of a few times kB T , the particles could only occasionally get
past it. They will therefore remain a certain distance apart and the suspension will
be stable.
In order to destabilise a suspension, a salt can be added to an electrostatically
stabilised suspension. This reduces the Debye length, which in turn decreases the
magnitude of the electrostatic forces in relation to the van der Waals forces. In the case
of a polymer-stabilised suspension, a poor solvent can be added. Then interactions
between the polymers will induce an attractive force between the particles. Another
possibility is to eliminate polymers adsorbed onto the colloidal particle surfaces
using some physical or chemical means. Finally, one can add a polymer that does
not adsorb onto the particles, since this would increase depletion interactions.
In the following, we shall only be interested in stable suspensions for which the
potential barrier is high enough to prevent irreversible aggregation effects. In this
case, either there is repulsion up to very long range (case C) or there is a secondary
minimum at relatively short range (case B). In a system in which case C prevails,
the particles tend to move very far apart, at least insofar as this is possible given
the presence of the other particles. They will thus disperse themselves uniformly
throughout the liquid. In the following, this kind of suspension will be referred
5.7 Balance of Interactions 179
Fig. 5.11 Different possible

profiles of the total interac-
tion potential between two
particles, sum of an attractive
van der Waals potential of the
form (5.13) and a repulsive
electrostatic potential of the
form (5.22), using arbitrary
parameter values
to as a repulsive system. In case B, the particles will place themselves at points of

secondary minimum, which are local potential wells. They will thus be held a certain
distance apart and will be able to extract themselves from these positions rather easily
with the help of thermal agitation or ordinary mechanical action. This is called weak
aggregation. In the following, we shall refer to this kind of suspension as an attractive
system.
One should bear in mind that the above description is only valid over certain
limited time scales. Indeed, even if their average energy is less than the level of
the potential barrier, it is always possible that at some point a particle will have
sufficient energy to get past. Considering the depth of the attractive potential wells,
the probability of a particle then getting back past the barrier in the other direction is
generally much lower. As a consequence, any colloidal suspension should in principle
be unstable over a long enough time scale. In practice, the stability of a suspension
must thus be related to a particular observation time.
Finally, note that, in the general case, one is dealing with particles assuming a
whole range of sizes and shapes, in simultaneous interaction with several particles,
and this may give rise for instance to attractive interactions with one end of a particle
and repulsive ones with the centre. Under such conditions, describing the system by
means of a simple interaction potential between two particles, and one which depends
on a single scalar parameter, is clearly a much more delicate matter. The main types of
interaction identified above are nevertheless sufficient to give a qualitative description
of the main kinds of average net balance of forces between neighbouring particles in a
complex homogeneous system. Thus the two ideal systems (repulsive and attractive)
studied below will illustrate the main situations that may be encountered.
180 5 Colloids
Fig. 5.12 Two typical forms

of essentially repulsive inter-
action potential. (a) Rapid
decay at a certain distance. (b)
Gradual decay with distance
5.8 Behaviour of Repulsive Systems
Here we consider those systems in which two particles alone in the liquid would
tend to repel one another to large distances. This corresponds to the case where the
repulsive forces are high enough compared with the attractive forces. In the general
case, the force and energy decay steadily with distance [curve (b) in Fig. 5.12]. We
then speak of soft repulsive suspensions. In some cases, the force drops off very
quickly at a certain distance [curve (a) in Fig. 5.12] so that it is difficult for two
particles to come within any shorter range, while they interact negligibly further out.
We then speak of hard repulsive suspensions. In reality, of course, one could not
draw the line between these two types of system. The transition from one to the other
is continuous and somewhat arbitrary, but this classification is nevertheless useful
for understanding the physical origin of certain kinds of behaviour.
5.8.1 Hard Repulsive Suspensions
This situation is mainly encountered (i) when the particles interact through a rather
high electrostatic potential or (ii) when they are coated with a dense polymer brush.
The so-called hard sphere model then represents the interaction potential in the form
of a step function of infinite height below a certain distance χ R (see Fig. 5.12) and
zero height beyond, rather like the potential in (2.3) (see Sect. 2.2.6). Under such
conditions, the particles cannot come within a distance 2χ R of one another.
In case (i), it is possible for liquid to flow within a layer of thickness χ R around a
particle, but the presence of ions interacting with the solid surface will hinder motions
tending to carry them out of this layer. To a first approximation, it is just as though
this layer had a much higher viscosity than the viscosity of the suspending liquid.
The magnitude of this effect, known as the primary electroviscous effect, depends
5.8 Behaviour of Repulsive Systems 181
in particular on the ionic force, the ion concentration, and the particle size. In case
(ii), the adsorbed polymers also significantly hinder the flow of liquid through the
layer of thickness χ R around a particle. In both cases, the hard sphere approximation
can thus be extended, considering that the liquid contained in the layer of thickness
χ R forms a region that is rigidly bound to the particle. This amounts to treating the
particles as having an effective radius r = R(1 + χ).
Under these conditions, the behaviour of suspensions of hard repulsive spheres
can be described with the help of the ideas used for non-colloidal particle suspensions
(see Chap. 3). These suspensions are Newtonian over a broad range of concentrations,
with a viscosity that increases slowly at first for increasing concentration, as predicted
by the Einstein model of (3.16), then more quickly in the semi-dilute regime, for
example, as predicted by the Krieger–Dougherty model of (3.19). However, the
solid fraction that must be taken into account in this context is not the concentration
ξ of solid particles, but the apparent concentration ξ ∼ of impenetrable hard spheres
of radius r . For a given volume α of suspension, the volume of particles present in
the solution is αξ. The volume of hard spheres is thus αξ(r/R)3 , which can also
be written αξ ∼ . It follows that ξ ∼ = ξ(1 + χ)3 .
5.8.1.2 Effect of Brownian Motion
The above analysis does not take into account one of the key features of colloidal
particles, namely, the fact that they can diffuse through the liquid under the effects of
thermal agitation in the motion known as Brownian motion. Relative to the macro-
scopic flow imposed by external conditions such as a simple shear, this agitation
induces further motions of the interstitial liquid and hence further dissipation of
energy which increases the apparent viscosity of the mixture. One would expect this
phenomenon to become important for low shear rates because then, over a given
lapse of time, the particles would have time to diffuse over much greater distances
than those induced by the imposed macroscopic flow. On the other hand, for high
shear rates, the effects of such diffusion would soon become negligible.
Qualitatively, Brownian motion has a similar effect to the agitation of molecules
in a gas (see Chap. 2). Here it is as though we had a ‘gas’ of particles immersed in a
liquid. The viscosity of a system like this results from momentum exchanges when the
particles migrate from one layer to another. On the other hand, in the present case, the
process leading to momentum transfer within a layer is no longer collisions between
molecules as it would be in a standard gas, but viscous dissipation due to the motions
of particles through the liquid. Consequently, it is not directly thermal agitation, and
hence the value of the temperature through (2.1), that determines this dissipation,
but rather the magnitude of the diffusion effect compared with the imposed flow.
To describe this phenomenon, we therefore attribute an apparent viscosity σB which
decreases for increasing values of the ratio Pe, known as the Péclet number, between
the characteristic diffusion time and the characteristic flow time.
Let us consider a simple shear. Under the effect of this flow alone, relative to a
neighbouring particle at distance b, a particle moves at a speed of the order of bΨ̇ , so
182 5 Colloids
Fig. 5.13 Dependence of

the apparent viscosity of a lnμ
colloidal suspension on the
Péclet number Pe and the
solid fraction
φ
Pe
the characteristic time required to cover a distance of the order of b is thus Th = 1/Ψ̇ .
Under the effects of thermal agitation alone, a particle travels a distance of the order
of b in a time estimated as Tb ⇒ kμb2 /kB T by (5.3). We thus obtain
Tb kμΨ̇ b2
Pe = = . (5.30)
Th kB T
Whatever the shear rate, viscous dissipation is associated with the macroscopic
motion imposed on the fluid. To a first approximation, we may assume that this
dissipation simply adds to dissipation produced by Brownian motion and leads to a
total viscosity for the system equal to the sum of the viscosities associated with each
phenomenon:
σ = μ(ξ ∼ ) + σB (Pe) , (5.31)
where μ(ξ ∼ ) is the viscosity of a suspension of non-colloidal particles at

concentration ξ ∼ . In this expression, the extra viscosity σB due to particle diffu-
sion is a function that decreases as Pe increases and tends to zero when Pe √ ∝
(negligible Brownian diffusion). This result is shown in Fig. 5.13. When Pe ◦ 1,
particle diffusion is negligible compared with the motion induced by the macro-
scopic flow of the suspension, so the viscosity is just the viscosity of the equivalent
non-Brownian suspension. But when Pe ∇ 1, the particles diffuse to much greater
distances than the relative motions induced by the macroscopic flow, so the apparent
viscosity is higher than for the equivalent non-Brownian suspension.
When the solid fraction increases, the number of particles involved in diffusion
also increases and this tends to boost the number of momentum exchanges. However,
diffusion is slowed down by the presence of the surrounding particles which tend to
get in the way of the diffusing particle. These two effects thus tend to balance out to
some extent, but the second effect, which shows up through the increase in μ as a
function of ξ in (5.30), begins to dominate when the concentration goes above a few
percent. Finally, the Péclet number associated with a given shear rate remains constant
or increases with the particle concentration, while σB remains constant or decreases.
Fig. 5.14 Suspension of 2.2 μm fluorescent PMMA beads observed by confocal microscope. Left
In an ordered state after a sufficiently long rest time. Right In a disordered state after shaking the
system slightly. Reproduced with kind permission, from Habdas and Weeks [1]. Copyright 2002
Elsevier
At the same time, the viscous dissipation induced by the macroscopic flow increases.
The level μ(ξ ∼ ) of the viscosity plateau at high values of the Péclet number thus rises
(see Fig. 5.13). Finally, in practice, the rise in the viscosity associated directly with
the Brownian motion disappears when the concentration increases.
5.8.1.3 Behaviour of Concentrated Suspensions of Spheres of Uniform Size
With this type of suspension, when the apparent concentration is high enough, a
specific consequence of the colloidal nature of the particles is observed. For concen-
trations ξ ∼ above 55 %, a crystalline phase is obtained, adopting one of the maximal
packing configurations, i.e., hexagonal close packed (hcp) or face-centered cubic
(fcc). This is surprising since this is almost the minimal concentration for a disor-
dered packing of non-Brownian spheres, and disordered packings can be obtained
up to concentrations of around 64 % (see Chap. 3). In fact, these systems have even
more novel properties. When the concentration is gradually increased until ξ ∼ reaches
49.4 %, a crystal of concentration 54.5 % suddenly forms (see Fig. 5.14). This is a
genuine liquid–solid phase transition. If a suspension is prepared at an intermediate
concentration, it separates into two coexisting phases. Note that this crystal generally
adopts an hcp or fcc configuration when χ < 1.7 and a body-centered cubic (bcc)
configuration when χ > 1.7.
This phenomenon has an entropic origin. When the particles arrange themselves
into the crystal structure, they lose entropy associated with their previous long-range
collective disorder, but they gain entropy associated with local disorder because
they have much more room to move around their equilibrium position. Indeed, in a
disordered packing, the particles form a network of contacts which severely restricts
184 5 Colloids
the possibilities for relative motions of the particles. Brownian motion thus plays
a fundamental role in this transition, because it allows the particles to explore the
various configurations and end up choosing the one with highest entropy. Such a
transition would be impossible within a system of non-Brownian particles at rest.
If the material is allowed enough time to organise itself during its preparation,
a crystal structure is obtained from a concentration of 49.4 % and upto 74 %. How-
ever, a disordered structure can also be obtained up to a concentration of 64 % by
quickly mixing the particles and the liquid. The resulting ‘glassy’ structure is stable
because the particles obstruct one another sufficiently to prevent any spontaneous
rearrangement into a crystal structure.
5.8.2 Soft Repulsive Suspensions
Here the total interaction potential falls off gradually with the distance between the
particles. In contrast to the hard sphere case, it is now a much more delicate matter
to represent the particles by larger volumes encompassing a region in which liquid
motions are slowed down by the presence of ions or polymers. Indeed, the transition
between an apparently very viscous liquid domain that is more or less ‘frozen’ around
the particle and an independent liquid domain becomes much more gradual.
As long as the particles remain at a distance where colloidal interactions can be
neglected, the suspension can be described approximately as in the hard sphere case
by using an apparent concentration ξ ∼ of solid particles, defined for example from the
distance χ R such that the potential is equal to half its maximum value. In this case,
at low solid fractions, there is a dilute regime, then a semi-dilute ‘hydrodynamic’
regime (see Fig. 5.15) in which the behaviour of the suspension is Newtonian and the
apparent viscosity increases with the concentration as it would for a non-colloidal
suspension. Furthermore, as in the hard sphere case, Brownian motion can play an
important role at low shear rates (see Fig. 5.13).
Above a critical concentration ξ0 , the average distance between the particles is
such that the colloidal repulsion forces begin to play a significant role. Each particle is
subjected to repulsive forces from each of its neighbours (see Fig. 5.15). We then enter
the concentrated regime. If thermal agitation alone is still able to change the nearest
neighbours of each particle in the suspension at rest, despite colloidal interactions,
the particles are not jammed into any particular configuration. For this to be the case,
the average interaction potential ν(b) must not be more than something of the order
of kB T . The suspension is then a relatively simple fluid without a yield stress. Given
that all flow involves viscous dissipation in the interstitial liquid and relatively high
colloidal interactions, this kind of suspension will have a high viscosity. For this
reason, in the classification of Fig. 5.15, this regime has been qualified as glassy. In
the present state of our understanding, it is difficult to go further in the description
of this type of system.
Dilute Semi-dilute Concentrated Concentrated

φ'<<1 0.01<<φ'<φ 0 glassy pasty
φ0<φ<φc φc<φ<φm
Fig. 5.15 Rheophysical regimes of a soft repulsive suspension for different values of the solid
fraction (see text)
On the other hand, when thermal agitation is unable to extract the particles from
their environment in the suspension at rest, they remain jammed within the network
of colloidal interactions which cannot be broken simply under the effects of thermal
agitation. All things being equal, this situation occurs when the particles are close
enough for the colloidal interactions to be strong enough, in other words, above
a critical concentration ξc . This corresponds to a situation in which the average
interaction potential ν(b) is of the order of kB T . The material thus obtained is a
yield stress fluid, i.e., a high enough stress must be imposed in order to break the
network of interactions. For this reason, in the classification of Fig. 5.15, this regime
has been labelled the pasty regime.
5.8.2.2 Behaviour in the Concentrated Pasty Regime
In this regime, each particle tends to repel its neighbours, but the latter are unable
to move as far away as necessary because they are already being repelled by other
particles. Each particle thus occupies an equilibrium position in the liquid, determined
by the full set of forces exerted on it by its neighbours. In a sense it is ‘trapped’ in a
potential well from which it can only escape if it is supplied with more energy than
some critical value. This type of jamming also exists in the glassy regime, but only
temporarily, because the particles can spontaneously escape from their potential wells
with the help of thermal agitation. So in the concentrated pasty regime, the system
is jammed. This material cannot be significantly deformed without applying a stress
greater than a critical value which is then sufficient to lift some of the particles out
of their potential wells.
To get a better understanding of what happens in such a suspension when we try to
deform it, consider the following 2D model. We assume that the particles are arranged
in parallel planes and displaced by an angle θ/4 (see Fig. 5.16). Furthermore, we
neglect the interactions between particles of two different observation planes (located
at different depths along an axis perpendicular
√ to the plane of the page in Fig. 5.16).
The distance between two layers is b/ 2. An infinitely slow deformation Ψ induces
186 5 Colloids
b
Φ
A D
C
E
B
y A F
b r- γ
r+ B τ
τc
γc γ
Fig. 5.16 Changing energy and stress (right) within a system made up of spheres arranged in
parallel layers during a shear deformation inducing a parallel relative motion of the given layers
(left)
√
a displacement dy = Ψ b/ 2 in the principal direction. The distance between two
neighbouring atoms as they move apart changes from r = b to

b2 b2 (1 ± Ψ )2
r± = + ,
2 2
where the plus or minus signs correspond to atoms moving apart or moving together,
respectively (see Fig. 5.16). For small deformations Ψ ∇ 1, we have

Ψ2 Ψ
r± ⇒ b 1 + ± + O(Ψ 3 ) .
8 2
We can calculate the stress ω that must be applied in order to induce this deformation.
To do so, we take the force f applied to each particle in a layer and divide it by the
area associated with each particle in a layer to obtain ω = f /2b2 . We also know
that f = −E ∼ (y), where E is the energy associated with the interaction between
a particle and the particles in the neighbouring layer. Taking into account only the
interactions with nearest neighbours, we have E(y) = ν(r+ ) + ν(r− ), where ν is
the interaction potential between two particles. Carrying out a second order Taylor
expansion and introducing the above expression for r± , we then find

1 ν∼ (b)
E = 2ν(b) + kΨ 2 + O(Ψ 3 ) , k= + ν∼∼ (b) .
4 b
In general, k is positive. For example, for a power law repulsive potential of the form
ν = ϕ/r n , where ϕ > 0, we have
ϕn 2
ν∼ (r ) + r ν∼∼ (r ) = .
r n+1
Since we also know from the above equation that E ∼ (y = 0) = 0, we note in passing
that the chosen initial position (labelled A in Fig. 5.16) does indeed correspond to a
relative minimum of E, and hence to an equilibrium position √ of the system. Finally,
in the small deformation limit, we obtain ω = GΨ , with G = 2k/b. In this regime
Ψ ∇ 1, the suspension thus has linear elastic behaviour.
If we consider latex beads of diameter 0.1 μm, interacting together electrosta-
tically due to z ⇒ 1000 electrons on their surface, the Coulomb energy between
two beads a distance r apart will be ν = z 2 e2 /4θ χr = ν0 b/r . Taking a distance
b ⇒ 1 μm, we find ν0 ⇒ 1 eV, which is of the same order of magnitude as the inter-
action energy between two metal atoms. According to the above formulas, we then
find G ⇒ 0.25 Pa. This is about ten orders of magnitude smaller than in ordinary
(atomic) crystals. But since the interaction energy between the elements is of the
same order, this difference is essentially due to the fact that colloidal particles are
much bigger objects than atoms, and typical values of b are therefore much greater
than the distance between two atoms. This shows that colloids are necessarily ‘soft’
solids from the point of view of their mechanical properties in the solid regime,
essentially due to the characteristic length scales of their structure.
5.8.2.3 Solid–Liquid Transition
In the above model, the stress required to produce a certain deformation first increases
with the deformation. The system is then a solid, i.e., it deforms finitely and when
the force is removed it returns to its initial configuration. Beyond a critical point C,
corresponding to the point of inflection of the energy curve and associated with a
deformation Ψc , the necessary stress decreases and becomes negative. This means
that, if we have imposed a greater stress than the maximum value ωc reached at point
C and if this stress is maintained or if it is removed after reaching Ψc , the system will
tend to deform of its own accord up to position D, then on to positions E and F. In this
case, the initial configuration is broken, the particles have changed neighbours, and
the system will no longer spontaneously return to its initial position. We have thus
left the solid regime, but it should be noted that, since the particles are in fact identical,
the new configuration looks just like the initial one. If we continue to impose a stress
greater than ωc , the material will deform to the next point of inflection associated
with the same stress level, and this process will repeat at the next step, and so on.
The deformation thus proceeds in the same way with the layers slipping parallel to
one another. So finally the material flows and we have reached the liquid regime.
188 5 Colloids
5.8.2.4 Liquid Regime
Up until now we have considered only infinitesimal deformation rates, i.e., quasi-
static deformations. If the particle trajectories do not depend on the flow rate, the
stress terms associated with colloidal interactions will not depend on the shear rate.
One must therefore continually apply the minimum stress ωc able to keep the material
flowing (see last section). The main effect of the shear rate is to generate additional
viscous dissipation due to motions of the interstitial liquid, and this increases the
shear stress one needs to apply. In the framework of this simple model, the addi-
tional stress term μB Ψ̇ is simply proportional to the shear rate, since the particle
distribution remains constant on average. The resulting behaviour of the material
in the liquid regime is then described by ω = ωc + μB Ψ̇ , known as the Bingham
model. We shall see in the following that this model cannot reproduce the behaviour
actually observed in these materials, probably because the particle trajectories are
not completely independent of the shear rate.
5.8.2.5 Real Repulsive Suspensions
In practice, the particles in colloidal suspensions are rarely spherical and of uniform
size. This has analogous consequences for the dilute and semi-dilute regimes to
those described for non-colloidal suspensions (see Chap. 3). Whatever the physical
characteristics of the particles, there is always a critical concentration beyond which
the particles jam up owing to their mutual interactions (concentrated pasty regime). In
this case, one must apply a stress greater than a certain critical value in order to unjam
them. Up to a critical deformation, we thus have an elastic solid regime, because when
the stress is removed, the particles return spontaneously to their initial configuration.
More precisely, since there are also some viscous effects, due in particular to the
motions of the interstitial liquid, the material has viscoelastic behaviour in the solid
regime. Beyond the critical deformation, the particles are irreversibly extracted from
their local potential wells. Maintaining the stress, the same mechanism is reproduced
in the new configuration, breaking it once again, and this situation repeats itself until
the stress is removed. While the system remains too disordered to be able to describe
the various characteristics of the behaviour in any greater detail, with reference to
some simple scheme of organisation, the main physical mechanisms operating at the
local level are conceptually the same as in the ordered case, i.e., under simple shear,
a particle climbs up the potential well it happens to lie in and, above a certain level,
actually escapes from it, only to fall back into a similar one nearby (see Fig. 5.17).
Given the disorder which holds sway in such systems, the particles find themselves
in potential wells of different depths within the material and the stress distributes
in a non-uniform manner on the local level within the particle network. It follows
that, when a macroscopic stress is applied, it will not be distributed uniformly as
in the ideal model described above, wherein each particle could move in the same
way relative to its neighbours. For a given macroscopic stress, this can lead to a
rather broad distribution of local deformations, which means that the liquid regime
ΔE
Jamming Solid regime Liquid regime
Fig. 5.17 Conceptual relation between local evolution of particles interacting with their neighbours
and the macroscopic behaviour of the material
only begins when a single particle, at the point of the network where the induced
deformation happens to be greatest, manages to leave its potential well. A moment
afterwards, the network will have reconfigured, whereupon another particle will leave
its potential well. The qualitative implications of this scheme are nevertheless the
same as those of the ideal model, i.e., there is still a solid regime below a certain
yield stress which is just the stress that must be applied to the network as a whole
to deform it to the point where it gives way. Current theories attempt to describe the
mechanisms underlying this fracture. It seems feasible to consider that a series of
more or less diffuse ‘plastic events’ located randomly through the sample underlies
the macroscopic flow of the material. However, it remains very hard to predict the
values of the rheological parameters from the physical characteristics of the system.
Figure 5.18 shows a typical flow curve (liquid regime) for this kind of material.
The stress plateau at low speed corresponds to the yield stress. When the stress is
ramped up, then down again, the apparent behaviour of the material does not directly
follow this curve. Indeed, in general, when the test begins, the material is at rest, hence
in its solid regime. As long as the imposed stress is below the yield stress, which will
be the case during a significant portion of the increasing ramp, the material remains
in the solid regime. The observed curve (see Fig. 5.18) over this stress range does not
correspond to permanent flow, but a transient flow leading to the finite deformation
associated with each stress level below the yield stress. The values obtained for this
part of the curve thus depend among other things on the measurement time at each
stress level. When the material leaves the solid regime, the apparent flow curve does
indeed correspond to homogeneous flow in the liquid regime. During the decreasing
stress ramp, the resulting measurements remain on the flow curve right down to the
stress plateau associated with the yield stress (see Fig. 5.18).
The flow curve of this kind of material is generally very well described, over
4 or 5 orders of magnitude of the shear rate, by a Herschel–Bulkley model (see
Sect. A.10.7). Under simple shear, this has the form
ω = ωc + k Ψ̇ n , (5.32)
190 5 Colloids
Log(τ)
Liquid regime
Yield stress
Solid
regime
.
Log(γ)
Fig. 5.18 Yield stress flow curve of a soft repulsive suspension when the stress is ramped up then
down, starting from a state of rest. The material begins in the solid regime (dashed lines), for two
different measurement periods, then goes into the liquid regime (continuous line). When the stress
is lowered, the material follows the flow curve in the liquid regime down to the stress plateau
associated with the yield stress
where k and n are two parameters associated with the material. In general, n lies
between 1/4 and 1/2. So the observed behaviour differs from the behaviour of the
model material described above. In other words, the flow curve can no longer be
represented by a Bingham model n = 1. This tells us that the hypothesis that energy
dissipation due to colloidal interactions is independent of the shear rate is not in fact
justified. In any real system, as the shear rate is increased, the relative motions of the
elements within the interaction network probably become more and more complex
and the relative trajectories ever more tortuous. This very likely explains why, despite
its practical importance, there is still no simple general theory to explain the physical
basis of the Herschel–Bulkley model and to predict the values of its parameters from
the physicochemical characteristics of the material.
5.9 Attractive Systems
5.9.1 Structure
We now consider systems for which there is a secondary minimum in the curve
representing the total interaction potential. In this case, when two particles move
toward one another, they will tend to position themselves at the distance corresponding
to this minimum. In the following, we shall refer to this as aggregation. In a suspen-
sion involving many particles, these particles will continually explore new positions
by the effects of Brownian motion. Clusters involving more than two particles may
form rather quickly by successive aggregation of particles, and those particles that
5.9 Attractive Systems 191
(a)
(b)
Fig. 5.19 Structures formed by gradual aggregation of particles on an initial nucleus or seed for
different strengths of attraction as compared with the heat energy. a Weak attraction. The particles
rearrange themselves gradually to form a ‘closed’ structure. b Strong attraction. The particles remain
fixed in their initial contact position and form an ‘open’ structure
are still free will continue to explore the surrounding space until they themselves
encounter particles in some existing cluster to which they may attach themselves.
The density of such clusters depends on the attractive force associated with the
potential well. Indeed, when one particle sticks to another, if the energy barrier is
not very high, thermal agitation will easily be able to lift the particle out of the
energy well. This particle is then likely to aggregate with other particles in the same
cluster. It also allows other particles to come and join up with the initial particle.
A weak attraction thus allows the particles to explore various cluster configurations. A
phenomenon of this kind might eventually lead to a permanent reconfiguration of the
particle ensemble. However, this process is limited by the fact that any given particle,
surrounded by other particles in its cluster, has a much lower probability of getting
free again. Therefore, particles can easily explore the various rather low density
configurations until they find themselves jammed into a dense enough structure from
which they are no longer able to escape (see Fig. 5.19a). On the other hand, if the
potential well due to interparticle attraction is deep, two aggregated particles will
remain jammed into this random position, and likewise for the other particles forming
the cluster. The system will be unable to explore any other configuration and will
hold the particles in their randomly attained relative positions, with the final result
being a very open structure (see Fig. 5.19b).
These two kinds of configuration are just like those obtained by the dynamic
aggregation processes described in two now classic models, viz., diffusion limited
aggregation (DLA) and reaction limited aggregation (RLA), where particles in an
192 5 Colloids
a priori unstable colloidal suspension are assumed to explore space by Brownian

motion. These models can describe the geometrical characteristics of the resulting
structures. In the DLA model, the probability of aggregation whenever one particle
encounters another is rather high and an open structure is obtained. In the RLA
model, the probability of aggregation is low, so a particle will not necessarily stick
to the first particles it encounters, which are generally located on the periphery of an
existing cluster, but can go on exploring the surrounding space with a good chance
of attaching itself to a particle inside the cluster. The resulting cluster is thus more
‘closed’ than one obtained by DLA.
The notion of open and closed structures can be made precise. Suppose we were to
juxtapose a number N of particles in a suspension, aggregating them into a periodic
geometrical arrangement in such a way that their apparent volume was spherical of
radius ξ . The solid fraction within the cluster is then constant, whatever its volume,
and we simply obtain N ∝ (ξ/R)3 . In this result, the exponent 3 follows from
the assumption about the proportionality of the number of particles and the volume,
which itself follows from the periodicity of the arrangement. The structures resulting
from the aggregation processes described above have specific geometrical features
which do not obey such a law. This arises because, as the structure gets bigger, those
particles that are still free ‘see’ a different geometric structure, so their probability of
aggregating is different and there cannot be spatial periodicity in the arrangement. In
fact, the aggregation process leads to fractal structures, i.e., the number of particles
in a given volume is not simply proportional to this volume. In fact, we have N ∝
(ξ/R) D , where D is the fractal dimension of the structure (lying between 1 and 3).
The lower the value of D, the more ‘open’ the structure is. The value obtained from
the DLA model and corroborated by observation is of the order of D = 1.75.
If aggregation occurred by the process described above, starting from a single
‘seed’ particle, then for a given volume of liquid, there would be only one possible
structure involving a specific number of particles given by the above law and asso-
ciated with a particular concentration ξc which can be calculated from the above
relations to be ξc ∝ (R/ξ )3−D . This is not what happens in reality. Starting with a
given liquid and particles, we can prepare aggregated suspensions with a broad range
of solid fractions, some of which are well above ξc . To describe the structure of a
concentrated attractive suspension, one must take into account not only the random
aggregation processes described above, but also the way the particles get in each
other’s way when they are confined within a given liquid volume.
In general, the particles and liquid are mixed by stirring vigorously and then left
to aggregate. In this case, if there is no sedimentation, clusters of roughly uniform
size will form (see Fig. 5.20) and gradually swell up to some critical size such that
they begin to come into contact with one another. They then form an infinite cluster,
i.e., extending from one side of the sample to the other. The structure is thus formed
by aggregation of fractal clusters each of which has reached the critical size and
which we shall refer to here as flocs. To a first approximation, we may consider that
the suspension comprises a very high concentration of spherical flocs with similar
fractal structures, arranged in a disordered way (see Fig. 5.21). Under these condi-
tions, the particle concentration can be calculated in the following manner from the
Fig. 5.20 Monolayer of 3.1 mm polystyrene beads on a water–oil interface. Initially, the particles
repel one another at short range and tend to arrange themselves as shown in the left-hand image
of Fig. 5.14. Salt is then added, and this reduces the electrostatic repulsive forces, allowing the
particles to aggregate. Left State of disorganisation induced after a few moments. The particles have
started to aggregate in pairs and triples. Right The same system after 1400 min. The particles have
by now aggregated into clusters or flocs extending widely through the liquid. Reproduced with kind
permission, from Park et al. [2]. Copyright 2008 American Chemical Society
ξγ
Δθ ξ
Fig. 5.21 Structure of a colloidal suspension with dominant attractions. Flocs (right) with fractal
structure and roughly spherical external appearance are piled up (left) at the maximal packing
concentration
characteristics of the system. If n is the number of flocs per unit volume, the maxi-
mum packing concentration that has been reached is ξm = θ nξ 3 /6 and the particle
concentration is ξ = θ n N R 3 /6, whence it follows that
1/(D−3)
ξ ξ
= . (5.33)
R ξm
As one might expect, the size of the flocs tends to the size of the particles when the
concentration increases. In other words, the flocs only now contain one particle each
when the particle concentration tends to the maximal packing concentration. Quite
194 5 Colloids
Dilute Semi-dilute Infinite aggregate Concentrated

φ<<1 0.01<<φ<φ c φc φc<φ<φm
Fig. 5.22 Rheophysical regimes of an attractive suspension as a function of the solid fraction
generally, the lower the concentration or the lower the fractal dimension, the larger
the flocs.
5.9.2 Behaviour of Attractive Suspensions
We assume for the time being that the suspension has a fixed structure in the form of
independent particles or clusters of some given size, in other words, that it is inde-
pendent of the flow history. Under such conditions, a dilute or semi-dilute suspension
(see Fig. 5.22) of particles or clusters behaves in a similar way to a suspension of
colloidal particles, with viscosity depending on the volume fraction of the clusters.
Above some critical concentration ξc , the particles will form an infinite cluster (see
Fig. 5.22). The suspension then has a yield stress and we enter the concentrated pasty
regime (see Fig. 5.22).
5.9.2.2 Behaviour in the Concentrated Pasty Regime
Solid Regime. When at rest, the particles can be assumed to have reached an
equilibrium position with respect to their interactions with all their neighbours.
They thus form a given structure. When a macroscopic deformation is imposed
on this structure, this will of course induce relative motions of the particles, causing
them to shift from their equilibrium positions and hence to climb up their potential
wells. As long as these movements are not very big, the energy stored will be essen-
tially elastic and the structure will resume its initial form when the applied force is
removed. This is in practice the solid regime of the material.
For structures of this kind, two different methods have been put forward to predict
how the elastic modulus will depend on the solid fraction. When ξ is slightly greater
than ξc , i.e., when one is close to the percolation threshold of the system, it can be
shown that certain physical properties of these materials can be expressed as power
laws in the deviation from the critical concentration. De Gennes suggested that this
would be applicable to the elastic modulus of gels, resulting in a law of the form
G ∝ (ξ − ξc ) p . (5.34)
Various experimental results seem to support this result for certain colloidal
suspensions at concentrations significantly higher than ξc .
Another method has been put forward for systems at concentrations much higher
than the critical concentration. This is based on the assumption that the macroscopic
deformation of the system results essentially from deformations of flocs, which
themselves result from changes in the angles between particles in contact. This
means that two neighbouring particles are likely to slip over one another and at the
same time to move slightly further apart (see Fig. 5.21). Indeed, this mechanism
seems likely to be the least costly in terms of energy, but it would only work for very
limited deformations. A local elastic modulus k0 is associated with this phenomenon,
whence the elementary force inducing a change λθ in the angle would be given by
f = k0 Rλθ . Furthermore, it is assumed that the deformation of the floc occurs for the
main part in its ‘backbone’, this being made up from a chain of Nch = ξ dch particles
in contact, where dch is another fractal dimension. If we consider a deformation
inducing a maximum displacement of πl, this corresponds to a total change of
Nch πθ in the angle, where πl/ξ ⇒ Nch ξ πθ/ξ = Nch πθ . The energy stored in the
flocs is then
k0 (Rπθ )2 k0 R 2 Ψ 2
ν = Nch = ,
2 2Nch
and, according to (2.39), the shear modulus of the floc is given by
1 d2 ν k0 R 2
G= = ,
ξ 3 dΨ 2 Nch ξ 3
whence finally,
k0 (3+dch )/(3−df )
G= ξ . (5.35)
R
Liquid Regime. Up to now we have assumed that, in the solid regime, no link
was broken in the particle network. Since the structure of the aggregated network
is disordered, the stress distribution will be heterogeneous on the local scale of the
links between particles. As a consequence, when the stress is steadily increased, we
eventually reach a value capable of breaking a first link within the structure. This will
be the link for which the ratio of the local stress to the attractive force was highest.
The situation then differs from what happens in repulsive suspensions, because in
that case, almost as soon as an element leaves its potential well, it will fall back
into another one. In the present situation, there is no particular reason why separated
particles should immediately reconstitute links with other particles. Reconnecting
196 5 Colloids
Η ε
Fig. 5.23 Fracture and ‘liquefaction’ in a specific region of the material leading to a localisation
of the deformation within a colloidal suspension with predominantly attractive interactions. The
initial state is shown on the left. The structure constituted by the network of particles in contact is
then deformed until fracture occurs in a specific region (particles represented by white disks). To a
first approximation, particles in the fracture zone can be considered to be dispersed independently
through the liquid
with a network requires the particles to explore the space around them and encounter
other particles under the right conditions. This process will therefore take a certain
time, and this time will be longer if the surrounding particles are further away, if
the interstitial liquid is more viscous, if the temperature is lower, or if the attractive
interactions are weak. In the current state of our knowledge, there is no general
way to describe the various characteristics of this process. In the following, we shall
simply assume that a single characteristic time θ can be associated with it, defined
as some average time required for a particle that has just separated from a particle to
reconnect with another one.
Around the first broken link, the force required to shear the material is lower
than it was previously because the resistance to motion is that of a suspension of
non-aggregated particles. If the macroscopic stress is maintained at the same level,
the forces applied to the other points of contact situated around this broken link
will immediately become greater. Other links among them will thus rapidly break,
and this will further contribute to increasing the forces applied in the vicinity. The
result is a chain reaction which ends in a localised fracture in some specific region.
For a simple shear, one would naturally expect the broken links to be localised in
a region of rather limited thickness χ parallel to the shear plane. The material in
the fracture zone can then be considered as a suspension of independent particles or
small dispersed (non-aggregated) clusters, while the rest of the material is deformed
but remains in its solid regime (see Fig. 5.23). The material thus appears to flow and
the yield stress ωc is the smallest value of the stress able to produce this flow scenario.
Under the action of the shear stress ω > ωc , which is assumed to be macroscopi-
cally uniform, the material can thus be treated as rigid over a thickness H − χ, where
H is the sample thickness, and liquid over a thickness χ. Since a good few links are
broken in this flow region, the material behaves as a semi-dilute suspension of par-
ticles or small clusters, i.e., Newtonian with viscosity μ. The shear rate is therefore
ω/μ in the sheared region. The apparent shear rate in the sample is then (ω/μ)χ/H ,
and the apparent viscosity of the system is
Log(τ)
Liquid regime
Apparent
yield stress
Localization - Instability
Solid regime
.
Log(γ)
Fig. 5.24 Apparent flow curve of an attractive suspension when the stress is first ramped up, then
down again, starting with a material at rest. The material begins in the solid regime (dotted line).
Then a sudden transition is observed (upper dashed line) at a certain level of stress (apparent stress
of the material), taking the material into its liquid regime (continuous line). When the stress is
reduced, the material remains in its liquid regime to begin with, then at a certain stress level, comes
to a rather sudden halt (lower dashed line). The values of the apparent yield stresses depend on the
flow history
H
σ=μ . (5.36)
χ
If the sheared thickness χ and the viscosity μ of the liquid region were constant
whatever the flow history, σ would be constant and the colloidal paste would appear
to have Newtonian behaviour. In reality, several phenomena complicate this picture,
leading to the rather typical properties of such systems.
According to (5.36), the shear rate should increase in proportion to the stress.
In fact, when the stress is increased, further links are broken in the solid zones. The
thickness χ of the liquid zone then increases with ω . Furthermore, as the stress goes
up, further links may be broken among the clusters in suspension in the liquid zone,
and this helps to reduce the viscosity μ. We thus observe that the two phenomena
described here tend to produce an apparent viscosity ω/Ψ̇ = μH/χ which decreases
as ω increases, whence the material is in fact shear thinning.
Suppose now that, starting from a permanent flow regime at a stress level greater
than the yield stress as above, we subsequently lower the stress to some given level.
If the characteristics of the flow do not change, (5.36) predicts that the material can
still appear to flow like a liquid. In practice, this is indeed observed down to a certain
stress level ω0 , below which the material stops flowing (see Fig. 5.24). This arises
due to a factor that we have so far neglected, namely, the reestablishment of links
between particles. When the characteristic flow time 1/Ψ̇ is short, the particles do not
have time to explore their environment and encounter other particles with which they
could form new links. To a first approximation, we may assume that links broken in
a fast flow will remain so. In contrast, when the characteristic flow time 1/Ψ̇ is long
enough compared with the characteristic time θ required to reestablish a link, the
198 5 Colloids
particles have easily enough time to explore the space and encounter particles with
which they could join up, because their environment does not change significantly
during a lapse of time longer than θ . When the stress is lowered, the material flows
less quickly and links have time to reestablish. This process goes on as the speed
decreases, until an infinite cluster is able to form once again. Since the applied stress
is at this point below the yield stress, it cannot then make the material flow and it
comes to a halt.
This description constitutes the physical origin of the experimental observations
described in Fig. 5.24. In particular, we can understand why the yield stress ωc asso-
ciated with the beginning of the flow is greater than the stress ω0 associated with
the end of flow. It is because, by breaking the links at the beginning, the structure is
weakened, and it requires a certain time to reconstitute itself, a time during which we
can continue to make the material flow at a lower stress level. An increasing stress
ramp followed by a decreasing ramp thus leads to a hysteresis loop (see Fig. 5.24).
This loop comes about due to the temporal evolution of the flow characteristics,
and in particular, due to the change in the number of links within the material. This
material is thus thixotropic.
The reality of the above phenomena is a little more complex. Aggregation
processes are gradual, implying that the material will come to a gradual halt at a
stress that will then depend also on the rate of decrease of the imposed stress. Fur-
thermore, these systems generally never reach an equilibrium configuration. They
are made up of a multitude of colloidal particles interacting in a complex way and
subjected to thermal agitation. The particles constantly explore different energy con-
figurations, and the positions of lowest energy discovered as time goes by tend to
hold the particles for longer, so that the average interaction potential tends to increase
over time. The yield stress ωc thus increases logarithmically with the rest time (see
Fig. 5.24).
5.10 Pasty–Hydrodynamic Transition
Up until now we have described the behaviour of concentrated colloidal suspen-

sions (repulsive or attractive suspensions above the critical concentration allowing
a jammed or aggregated structure to persist within the liquid in the absence of any
externally applied stress) by considering implicitly that the colloidal interactions
between particles play a dominant role. It is true that, by definition, in this regime
colloidal interactions dominate the effects of thermal agitation, but the latter can nev-
ertheless play a determining part in the temporal evolution of mechanical behaviour.
Moreover, at low shear rates, one would of course expect viscous dissipation associ-
ated with the flow of the interstitial liquid to be negligible compared with colloidal
interaction energies.
This is no longer true when the shear rate is high enough. One can then enter a
regime for which the viscous dissipation associated with flow of the interstitial liquid
is considerably greater than the colloidal interaction energies. A quantifiable criterion
5.10 Pasty–Hydrodynamic Transition 199
for this regime can be obtained, for example, by comparing the energy dissipation
involved in a displacement through distance b relative to its neighbours of a particle
subjected either to colloidal interactions alone or to hydrodynamic interactions alone.
Here we assume that the average volume of the suspension associated with each
particle is b3 . The first of these can be computed by considering that, at infinitely
small speeds and hence with negligible hydrodynamic interactions, owing to the
colloidal interactions, an average stress ωc must be applied to the system, and this
will involve an energy dissipation of ωc b3 Ψ̇ . The critical point in this argument is that
we assume that this dissipation energy will not depend significantly on the shear rate.
The second case can be estimated by considering the viscous dissipation μΨ̇ b3 within
a suspension of solid particles of the same form but without colloidal interactions.
The viscosity μ of the suspension of identical non-colloidal particles depends on the
concentration and size and shape distributions of the particles. The ratio Bi = ωc /μΨ̇
of the two terms so obtained is called the Bingham number. When Bi ◦ 1, colloidal
interactions dominate and the material is in the pasty regime, which we have been
discussing in the previous sections. When Bi ∇ 1, hydrodynamic dissipation in
the interstitial liquid dominates and the relative motions of the particles are mainly
determined by hydrodynamic interactions. In this case, the suspension behaves like
the equivalent suspension of non-colloidal particles. It is thus Newtonian with a
viscosity μ.
References
1. Habdas, P., Weeks, E.R.: Curr. Opin. Colloid Interface Sci. 7, 196 (2002)
2. Park, B.J. et al.: Langmuir 24, 1686 (2008)
Further Reading
3. Coussot, P.: Rheometry of Pastes. Suspensions and Granular Materials. Wiley, New York (2005)
4. Daoud, M., Williams, C. (eds.): Soft Matter Physics. Springer, Berlin (1999)
5. Everett, D.H.: Basic Principles of Colloid Science. Royal Society of Chemistry Paperbacks,
London (1988)
6. Israelachvili, J.: Intermolecular and Surface Forces, 2nd edn. Academic Press, London (1991)
8. Mewis, J., Wagner, N.J.: Colloidal Suspension Rheology. Cambridge University Press, Cam-
bridge (2011)
9. Oswald, P.: Rheophysics: The Deformation of Flow and Matter. Cambridge University Press,
Cambridge (2009)
10. Russel, W.B., Saville, D.A., Schowalter, W.R.: Colloidal Dispersions. Cambridge University
Press, Cambridge (1989)
11. Shaw, D.J.: Introduction to Colloid and Surface Chemistry, 3rd edn. Butterworths, London
(1986)
12. Tabor, D.: Gases, Liquids and Solids, 3rd edn. Cambridge University Press, Cambridge (1991)
Chapter 6
Emulsions and Foams
Abstract Here we investigate systems containing fluid inclusions, either gaseous

or liquid, dispersed through a liquid. When they have been stabilised, that is, when
some way has been found to prevent the coalescence of the inclusions, these systems
are similar in many ways, from the rheophysical point of view, to the suspensions
presented in Chap. 3. In particular, one identifies the same concentration regimes
(dilute, semi-dilute, and concentrated), associated with different types of interaction
between the elements and different kinds of mechanical behaviour. However, there
is a fundamental difference in that, in a given regime, the behaviour of such systems
can still vary considerably with the ratio of the viscosities between the two phases
or the relative magnitude of capillary and viscous effects. Another key difference
comes from the deformability of the inclusions, which means that a fluid behaviour
can be obtained beyond the maximum packing fraction of the non-deformed objects
(of the order of 60 %), up to inclusion concentrations close to 100 %.
6.1 Introduction
In this chapter we shall discuss systems made from fluid (gas or liquid) inclusions
dispersed in a liquid. These inclusions are assumed to contain a large number of
molecules of the same fluid so that they can be considered, on their own scale,
as continuous media. A distinction is made between emulsions, made by dispers-
ing inclusions of a liquid A (the dispersed phase) in a liquid B (the continuous
phase), and foams, comprising gaseous inclusions (the dispersed phase) in a liquid
(the continuous phase). Throughout the chapter, we shall assume that the fluids in
the two phases are Newtonian of viscosity μ for the dispersed phase and μ0 for the
continuous one. The main difference between emulsions and foams comes from the
viscosity of the inclusions. Indeed, viscous effects relating to the flow of the inclusion
fluid are negligible for foams, whereas they can play a significant role for emulsions.
However, these systems can be treated in the same framework thanks to specific

DOI: 10.1007/978-3-319-06148-1_6,
202 6 Emulsions and Foams
physical characteristics conducive to a similar rheophysical approach. In particular,

they are prepared by creating interfaces between two phases, they are not stable
unless these interfaces are treated in some special way, and the deformability and
fluid nature of the inclusions play a crucial role in the resulting rheological behaviour.
In contrast to a homogeneous liquid, these systems are characterised by the pres-
ence of a very large number of liquid–liquid or gas–liquid interfaces. They thus have
one component of potential energy, associated specifically with these interfaces,
which is proportional to their total surface area. To create these interfaces from two
homogeneous volumes of each phase, a certain amount of energy must be supplied.
Various techniques for chemical or mechanical preparation have been devised (see
Sect. 6.3).
As they move around randomly within the fluid, two inclusions that happen to
come together will tend to join up into a single compact inclusion, in such a way as to
reduce their total area and hence lower their total potential energy. They are then said
to coalesce. If no particular steps are taken, the system will not therefore be stable. The
above process continues and the two phases will eventually separate out, each forming
a macroscopic region with minimal interface with the other phase. In order to hinder
this phenomenon and thereby stabilise the emulsion or foam, a stabilising agent
is added. This attaches itself to the interfaces, considerably slowing down contacts
between neighbouring inclusions of the dispersed phase. However, in the long term,
the molecules of the dispersed phase which dissolve in the continuous phase will
migrate from the smaller to the larger inclusions. This is the phenomenon known as
ripening. When bubbles or droplets come into contact, they can still coalesce because
the barrier set up by the stabilising agent is never perfectly reliable. All the issues
relating to stability will be discussed in Sect. 6.4.
Systems made up of elements dispersed in a liquid phase are analogous in
many respects with suspensions, at least from a rheological point of view (see
Chap. 3). In particular, one can distinguish the same concentration regimes, viz.,
dilute, semi-dilute, and concentrated, associating them with different types of inter-
actions between elements and different types of mechanical behaviour. However,
there is one fundamental difference which lies in the fact that, in a given regime, the
behaviour of such systems can still depend significantly on the ratio of the viscosities
of the two phases or the relative importance of capillary and viscous effects. Another
key difference comes from the deformability of the inclusions, which means that it
is possible to have a regime of fluid behaviour beyond the maximal packing con-
centration of the undeformed uniform spherical objects (around 60 %), and in fact
up to inclusion concentrations close to 100 % (see Sect. 6.5). Before discussing the
preparation, stability, and behaviour of these systems, we must begin by understand-
ing their physical properties, and in particular, energy, pressure, and deformation
properties, on the scale of a single inclusion. This will be the subject of Sect. 6.2.
6.2 Physical Properties on the Scale of the Inclusions 203
6.2 Physical Properties on the Scale of the Inclusions
The systems considered here consist of inclusions of one phase within another. The
interfacial energy associated with these inclusions will largely determine the prepa-
ration and stability of the mixture. The shape of the inclusions results for its part from
an equilibrium between the viscous stresses exerted between them due to flow within
the system and resistance to deformation due to the induced changes in interfacial
energy. Here we review the physical mechanisms underlying these phenomena.
6.2.1 Energy
Consider two fluids initially in contact over an interface of area A0 . By some mixing
operation, this interfacial area is increased to A. During this operation, most of the
molecules of each of these two fluids remain immersed in the middle of molecules
of the same fluid, so their potential energy of interaction will not change. On the
other hand, a certain fraction of the elements of A will end up on the interface with
B. In order to compute the change in energy associated with this operation, we shall
imagine it to be carried out in two stages. First of all, the elements are brought to
an interface with the air, whereupon the potential energy per unit area (the cohesion
energy) of these elements changes from wA to wA /2 (see Sect. 2.4.6). Likewise,
bringing the elements of B to an interface with the air reduces their energy by wB /2.
In the second stage, elements of A are placed in contact with elements of B, thereby
increasing their potential energy per unit area by wAB (adhesion energy). The total
energy that must be supplied to induce a change A = A − A0 in the surface area
is thus
W = ξAB A , (6.1)
where ξAB = wA /2 + wB /2−wAB is the interfacial tension between the two phases.
Consider two spherical inclusions of radius R. Their total interfacial energy is
8α ξAB R 2 . If they are joined into a single spherical inclusion, this will have radius
R ◦ such that the total volume is conserved, whence 8α R 3 /3 = 4α R ◦3 /3 and R ◦ =
21/3 R. The interfacial energy is then only 6.35α ξAB R 2 , less than the total energy
of the two separate inclusions. The interfacial potential energy of a mixture of two
phases is therefore higher the more finely dispersed it is, and this explains why energy
must be supplied in order to split the inclusions.
6.2.2 Pressure Difference Across an Interface
The pressure in two fluids with an interface is continuous across that interface if it is
planar, while a jump in pressure is observed if it is curved. To understand the origin of
this effect, we proceed as follows. Suppose we take a spherical inclusion of fluid 1 of

radius R and in which the pressure is p1 , immersed in a fluid 2 in which the pressure is
p2 , and change its volume in such a way that its radius increases by d R. The changes
in volume of each phase are dβ1 = −dβ2 = 4α R 2 d R and the change in interfacial
area is d A = 8α Rd R. While changing the volume of each phase, we must do work
against the pressure of the fluid in the corresponding volumes. The work that must be
done on the system is therefore − p1 dβ1 − p2 dβ2 . This operation is carried out at
constant temperature so the work done corresponds to the change in free energy of the
system (see Sect. B.4). The changes in entropy associated with the different spatial
configurations of the elements are very small because the phases are not mixed.
Finally, the main change in free energy comes from the changes in internal energy
associated with the modification of the interfacial area. The number of elements of
one phase in contact with the other is modified. The resulting change in interfacial
energy is thus ξAB d A and it follows that 4α R 2 ( p1 − p2 )d R = 8α RξAB d R, whence
2ξAB
p = p1 − p2 = . (6.2)
R
For similar reasons, there is a discontinuity in the pressure across the interface
between two fluid volumes of arbitrary shape. The value of the discontinuity now
depends on the two extreme radii of curvature R and R ◦ over this interface:
⎡
1 1
p = ξAB + ◦ . (6.3)
R R
It follows that the pressure is higher in smaller inclusions. The pressure in two
inclusions of different sizes is therefore different, even if the pressure is uniform
throughout the continuous phase. This phenomenon induces an imbalance between
the fluid phases contained in inclusions of different sizes, and this plays a fundamental
role in the phenomenon known as ripening (see Sect. 6.4.2).
6.2.3 Deformation of a Fluid Inclusion at Zero Speed

and Constant Volume
Deformations of inclusions play a key role during preparation, i.e., when mixing
the two phases, but also in the mechanical behaviour of the system, because these
deformations are associated with storage or release of potential energy. Consider
for instance a fluid inclusion (gas or liquid) immersed in a liquid. When the system
as a whole is in equilibrium, this inclusion will be spherical because this is the
shape that minimises the interfacial area between the two phases. If we deform this
inclusion at constant volume, we increase its area and hence also its interfacial energy.
To bring about such a deformation, a force must be exerted on the inclusion. The
b R
a a'
I
(i) (ii)
a
a
b b
I I
(iii) (iv)
Fig. 6.1 Deformation of objects with different shapes: (i) Square (continuous line) stretched to
a rectangle (dashed line), (ii) sphere stretched to an oblate spheroid (iii) or squashed to a prolate
spheroid (iv)
relation between this force and the resulting deformation is relevant to the mechanical
properties of the system. Let us now quantify this phenomenon in some simple cases.
Consider first a square of side a which is deformed by stretching it in one direction
and contracting it in the perpendicular direction to obtain a rectangle with sides
a ◦ = a(1 + ∂) and b (see Fig. 6.1). To ensure the conservation of area during this
transformation, we must have a 2 = a ◦ b, so b = a/(1 + ∂). The deformation of the
square can be expressed in the form ∂ = (a ◦ − a)/a. For a small deformation ∂ 1,
we deduce to first order that b = a(1 − ∂), and it follows that the perimeter of the
rectangle remains the same under this transformation, i.e., 2(a ◦ + b) = 4a. If we
now consider the second order terms, we find b = a(1 − ∂ + ∂2 ), so the perimeter
will have changed by 2a∂2 .
This result can be extended to cover an inclusion of arbitrary geometrical shape
in three dimensions, with a length scale of a. If the volume of the inclusion remains
constant during the transformation, the area remains constant to first order, but to
second order, for a small deformation, it is proportional to a 2 ∂2 , so the inclusion will
have gained a surface energy equal to
E = θa 2 ∂2 ξAB , (6.4)
where θ is a coefficient that depends on the shape. For example, for an initially
spherical inclusion of radius R, deformed into a spheroid of radii a = R(1 + ∂) and
b, where ∂ is the deformation, positive for a prolate spheroid and negative for an
oblate spheroid (see Fig. 6.1), we have θ = 32α/5.
The force required to deform an inclusion is found by considering the work needed
to bring about the deformation, viz., f dx = f a∂, which is equal to the change in
the free energy, whence
f = θaξAB ∂ . (6.5)
Note that the force is simply proportional to the imposed deformation. If this force is
removed, the inclusion resumes its initial shape, in the absence of other stresses. The
behaviour of an inclusion, associated strictly with surface phenomena, is therefore
linearly elastic in the limit of small deformations.
6.2.4 Displacement of an Inclusion in a Liquid at Rest
Here we consider an inclusion of a pure fluid moving at constant velocity through a

pure liquid at rest in a container whose walls are a long way from the inclusion as
compared with its own dimensions. To simplify, we shall assume that the inclusion
maintains its spherical shape. As it moves through the liquid, the fluids of the two
phases are in overall relative motion. The relative speed of the two phases along the
interface is no longer zero, as it was in the case of a solid inclusion (see Sect. 3.2.5),
owing to the fact that the fluid in the dispersed phase can flow within the spherical
volume it occupies. We thus expect the fluid motions within the inclusion to account
in part for the motion of the inclusion relative to the container. Hence, the liquid
in the continuous phase that happens to be close to the inclusion will generally be
dragged along by the motion of this inclusion at a lower speed than if the inclusion
were solid, so it will be less strongly sheared than for a solid inclusion. The viscous
drag force exerted on the inclusion due to its motion through the liquid is therefore
less than would be exerted on a solid object of the same shape subjected to the same
motion.
We may deduce the form of this drag force using a dimensional argument similar
to the one used for solid suspensions (see Sect. 3.2.5), except that there is an extra
difficulty here, namely, we must take into account the stress and velocity fields in both
the continuous and the dispersed phase. The effect of a change in the displacement
velocity or the radius of the inclusion is the same, but the impact of a change in
viscosity of one of the two phases is more subtle. We find that the force is simply
proportional to the viscosity of the continuous phase, provided that the ratio of the
viscosities of the two phases remains constant. It follows that the drag force has the
same type of functional dependence as for a solid particle, but with an extra factor
that depends on the ratio ϕ = μ/μ0 of the viscosity of the inclusions to the viscosity
of the suspending liquid:
3ϕ + 2
FD = 2α μ0 RV . (6.6)
ϕ+1
This is the most general possible formula for the displacement of an inclusion in
a fluid. In particular, when ϕ ∞ ∞, the viscosity of the inclusions is extremely
high compared with that of the continuous phase and we approach the situation for
rigid inclusions. We thus recover the expression (3.5) for the displacement of a solid
particle in a fluid. Conversely, when ϕ ∞ 0, we find FD = 4α μ0 RV , which is the
minimal drag possible. Note that the surfactants added to stabilise inclusions (see
Sect. 6.4), placed as they are across the interface, tend to hinder the motion of the
inclusion, so it is just as though it were surrounded by a layer of fluid of higher
viscosity, and this would tend to increase the actual value of the force compared with
(6.6), which neglects this aspect of the situation.
6.2.5 Sedimentation or Creaming
If the density πi of the inclusions is not the same as the density πc of the continuous
phase, the inclusions tend to move under the effects of gravity, for the very same
reasons as discussed in the case of solid suspensions (see Sect. 3.2.7). From the
equilibrium between the forces of gravity and viscous drag, the latter being given in
(6.6), we can obtain a general expression for the speed of sedimentation of a single
isolated inclusion when π = πi − πc > 0, or the rate of creaming when π < 0 :
2π R 2 g 3ϕ + 2
V = . (6.7)
3μ ϕ+1
In practice, this expression gives a rather rough indication, strictly valid only at very
low concentrations. As for a suspension (see Sect. 3.2.7), when the concentration of
inclusions increases, the speed of sedimentation is reduced, and perturbing effects
connected in particular with boundary conditions may begin to play a significant role
for moderate or high concentrations.
6.3 Preparation
The dispersion of a fluid phase (liquid or gas) in a liquid phase in the form of small
volumes is not a natural process. If two such phases are simply placed in contact with
one another, in the same container for example, we may observe either a mixture of
Mixing Abrupt temperature change

A B
B
A +B
(i) (ii) (iii)

Fig. 6.2 Establishing contact between a fluid phase (liquid or gas) A and a liquid phase B
i separated phases, ii dispersion of phase A (immiscible) in phase B, and iii homogeneous mix-
ture of two miscible phases
the two phases on the molecular level if the two phases are miscible (see Fig. 6.2i),
or a separation of the two phases in the form of a single large volume of one phase
sitting above a large volume of the other phase if the two phases are immiscible (see
Fig. 6.2iii). The result actually obtained depends on the relative concentrations of the
two phases and the temperature. The boundary between the two situations lies along
the miscibility curve. In the case of two liquids, this gives the critical temperature
above or below which, depending on the given system, the two phases are miscible.
Some systems such as mercury and water or oil and water remain immiscible at all
temperatures.
For emulsions, a first solution for dispersing a liquid phase in the form of droplets
(see Fig. 6.2ii) is to suddenly lower (or raise, depending on the system) the temper-
ature of the two liquids below (above) the miscibility curve. For at least a certain
time, after which the system must be very quickly stabilised, the resulting phase sep-
aration will take different forms depending on the respective concentrations of the
phases. Under certain conditions, one can thus form droplets of one phase dispersed
in the other, in what is referred to as spontaneous emulsification. However, the size
distribution of the droplets obtained in this way will not generally be uniform. In
the case of foams, the same operation can be carried out starting with a mixture
(which must be homogeneous because the two gaseous phases will be miscible) of
liquids, then abruptly increasing the temperature or reducing the pressure. This will
produce a dispersion of small bubbles of uniform size, but it is only possible with
certain fluids. In the end it is thus preferable to use ‘mechanical’ methods. Whatever
the fluid, these produce the interfaces between the two phases directly by deforming
them in an appropriate manner.
One technique that can be used to obtain inclusions of controlled and uniform
size is to inject one phase into the other in a series of small volumes. This is of
course costly in terms of time and energy, but it is useful when we need to produce
systems with well controlled characteristics, using so-called microfluidics techniques
6.3 Preparation 209
Elongation Flow stoppage
t=0 8
3 9
7 10
Fig. 6.3 Left Elongation induced by rotation of four parallel rollers immersed in a fluid containing
an inclusion (shaded). Centre Effect of this flow on the shape of an initially spherical inclusion
placed midway between the rollers, as viewed after different lapses of time. Right Changing shape
of the inclusion after the flow has ceased
to create microscale flows. The other, much more traditional technique is to mix the
two phases together vigorously, the underlying idea being that a simple flow such
as elongation or shear, if it is maintained for long enough, will gradually stretch the
interfaces between the two phases to the point where they separate into small, more
stable inclusions (see Sect. 6.3.2). Under these conditions, the most effective flow is
in principle the one which induces the greatest deformations. The way the mixing is
actually done with more complex flows, in which the trajectories of the fluid elements
are more likely to intersect, may also play an important role, but this effect is much
harder to quantify and it is taken into account only empirically through the different
mixer configurations available in practice.
6.3.2 Forming Inclusions by Deformation
Here we examine the way an element of a given phase separates into smaller elements
during an elongational or simple shear flow. When it is immersed in a fluid in a
simple elongational flow, an initially spherical inclusion will gradually stretch in the
principal flow direction. It thus assumes a cigar shape, and then a neck appears in
the central region, and finally the inclusion separates into two parts (see Fig. 6.3). It
is interesting to note that, if the flow is stopped at the moment when the inclusion
has the pinched cigar shape, it will continue to deform and eventually split into two
in roughly the same way. This shape is therefore unstable in itself, regardless of the
flow of the surrounding liquid.
Note that, in a simple shear, we find the same kind of stretching effect followed
by the inclusion breaking apart as shown in Fig. 6.3, but this time the long axis of
the inclusion swings steadily around an axis perpendicular to the plane and in the
direction of the flow. This happens because a simple shear can be decomposed into
an elongation and a rotation, with an angular velocity equal to half the shear rate of
the simple shear (see Sect. A.10.5).
Since it turns out that, under certain conditions, an inclusion can spontaneously
split into smaller inclusions, it is interesting at this point to examine the stability
of inclusions of different shapes. Quite generally, a cylindrical inclusion is unstable
beyond a certain critical length, at which point it will break up into smaller inclusions.
This happens because the surface area of several spheres may be less than the surface
area of the cylindrical volume from which they originated. For example, consider a
section of length l = xa of a long cylinder of radius a. If this section transforms into
a sphere, the radius of this sphere will be R = ya with x = 4y 3 /3 (conservation of
volume). The external surface area of the segment is greater than the surface area of
the sphere whenever x > 2y 2 , i.e., whenever x > xc = 9/2. In order to minimise
its interfacial potential energy, a cylindrical inclusion with high enough aspect ratio
should therefore break up into smaller spherical inclusions. However, this analysis
does not tell us the true critical length of the cylinder at which it becomes unstable,
because the sequence of events leading to the separation involves more complex
forms of interface. An exact calculation must take into account the evolution of a
small perturbation to the shape of the cylinder and investigate the conditions under
which it is more favourable for the cylinder to go on deforming to the point where it
breaks up. We then obtain the critical value xc ⇒ 6.3.
Let us return now to the dynamics. The deformation of the inclusion results from
(viscous) stresses exerted by the flow of the continuous phase. However, the inclusion
will resist this deformation, partly due to elastic phenomena associated with the
increase in the area of its interface, and partly due to the (viscous) drag opposing the
flow of the included fluid. (In the case of a bubble, the latter is of course negligible
in most cases.) Globally speaking, viscous dissipation within the inclusion tends
to slow down the deformation. After a sufficient lapse of time, the deformation will
reach a value determined by the balance between the elastic resistance to deformation
and viscous friction with the continuous phase. When the elastic resistance is much
greater than the viscous effects, the inclusion will deform only very slightly. In the
opposite situation, it will deform significantly and will end up splitting into parts as
described above.
In order to carry out a precise calculation of the resulting deformation, we must
give a detailed description of the stress field, and hence also of the velocity field, both
inside and outside the inclusion. However, an approximate calculation is possible,
for example under simple shear, if we assume that the inclusion (with characteristic
dimension R and surface area λ R 2 ) is subject, over the whole of its surface, to the
viscous stress μΩ̇ associated with a simple shear of the continuous phase, giving
a force Fv = λ R 2 μΩ̇ , and to an elastic resistance of the kind given by (6.5), i.e.,
Fe = 2θ RξAB Ω , which we extrapolate here to deformations of arbitrary type and
magnitude. When a balance is reached, we have Fv = Fe and the deformation is given
by Ω ⇒ λμR Ω̇ /2θξAB . A similar calculation for elongation leads to ∂ ⇒ λμ˙∂ /θξAB .
These expressions imply that, under a simple shear, the deformation will reach the
following value, up to a factor:
6.3 Preparation 211
Fig. 6.4 Behaviour of an inclusion in an elongational or simple shear flow. The inclusion splits
into smaller inclusions for a capillary number greater than the value associated with the viscosity
ratio ϕ on the corresponding continuous curve
μΩ̇ R
Ca = . (6.8)
ξAB
This is known as the capillary number. The capillary number can also be defined in
the same way for an elongation, replacing Ω̇ by ∂˙ in (6.8). Under these conditions,
as originally suggested by Taylor, the splitting of the inclusion into smaller elements
will be conditioned by the value of this capillary number. So a capillary number
much greater than unity implies large deformations of the inclusions which will lead
to their breaking up, while a capillary number much smaller than unity means that
the inclusions are unlikely to break up during the flow.
The exact conditions for the break-up of an initially spherical inclusion in an
elongational flow or simple shear have been determined experimentally [1]. As a
matter of fact, breakage does not depend solely on the capillary number, but also on
the ratio of the viscosities of the two phases (see Fig. 6.4). The first thing to note is
that the critical curve for elongation lies well below the critical curve for simple shear.
This arises largely because elongation is more effective with regard to extending the
inclusion than is simple shear, which includes in addition a rotational component
with no effect on the shape of the inclusion. Under elongation, the influence of the
viscosity ratio is rather slight and the critical capillary number varies between 0.2
and 1 for the relevant range of ϕ. Under simple shear, the viscosity ratio can play
a critical role. It is all the more difficult to split the inclusion as the viscosities of
the two phases become very different. There is even a critical viscosity ratio (of the
order of 4) for which this breakage is no longer possible.
These calculations can also help us to estimate the final size of the inclusions.
Indeed, for a given viscosity ratio, provided that the inclusion has a size for which the
capillary number is greater than the critical value defined by the curves in Fig. 6.4,
the flow will end up breaking it into smaller elements. When its size has reached a
value such that the capillary number is less than or equal to the critical value, it will
cease to split up. In practice, under simple shear, since the critical capillary number is
equal to about 1 in the viscosity range from 0.1 to 1, we may deduce an approximate
value for the average size of the inclusions once equilibrium has been achieved for
the given shear rate:
ξAB
Rf ⇒ . (6.9)
μΩ̇
6.4 Stability
When two initially separated phases are mixed together, inclusions of one of the
phases (the dispersed phase) are created within the other phase (the continuous
phase). This mixture will be stable1 if the inclusions of the dispersed phase can
remain in this form, and unstable if they tend to disappear in favour of larger inclu-
sions, leading in the long run to a complete separation of the two phases. The latter
may occur by direct combination of inclusions (coalescence) or by a more subtle
and much slower phenomenon in which the molecules of certain inclusions diffuse
toward others in the process known as ripening. It is thus essential to prevent coa-
lescence, which tends to very quickly destabilise the system. One commonly used
technique consists in coating the surface of the inclusions with a product known as a
surfactant, which in a certain sense isolates the dispersed phase from the continuous
phase.
6.4.1 Coalescence and Stabilisation
When two inclusions are nearest neighbours, they will nevertheless be separated by
a certain volume of the continuous phase (see Fig. 6.5a). During the macroscopic
motions of the system, and in particular while it is being prepared, these inclusions
can be made to approach one another. Furthermore, when the concentration of the
dispersed phase is high, the inclusions will naturally be pushed against one another.
When two inclusions move toward one another, the thickness of liquid between them
is reduced, but the normal viscous force (the hydrodynamic force) due to expulsion
of the intervening liquid will cause the inclusions to deform. A film then forms. This
is defined as a layer of liquid with thickness much smaller than its diameter (see
Fig. 6.5b). When there are no forces specifically able to counteract the approach and
1 To be precise, this kind of system is actually metastable, because the minimal energy configuration
is still the one corresponding to complete separation of the two phases. The configurations we refer
to for simplicity as stable correspond to local energy minima, which explains why in some cases
the system will slowly evolve toward some other state (ripening).
6.4 Stability 213
(a) (b)
R (c)
Fig. 6.5 Different stages in the encounter between two inclusions in a liquid. a Dispersion.
b Approach and formation of a film. c Coalescence
thereby stabilise the mixture, the two inclusions end up by merging and they are
said to coalesce (see Fig. 6.5c). In order to understand this phenomenon and ways
to counteract it, we must investigate how the interaction forces between the two
inclusions vary during the approach.
6.4.1.1 Hydrodynamic Force
The flow of the continuous phase out of the space between the two approaching
inclusions is analogous to the flow when two solid surfaces come together. In this
case, we know that the force that must be exerted on the two lateral volumes increases
with the speed and with the reciprocal of the distance (see Sect. 7.2.1). However, the
exact expression for this force is not easy to compute for fluid inclusions because
we must now take into account the flow of the fluid within the inclusions themselves
and also the deformation of the interfaces.
It is nevertheless straightforward to estimate this force in the limiting case where
the viscosity of the fluid in the inclusions is very low (as for a foam) and assuming that
the interfaces have the (fixed) form of two parallel disks. Since the drag on the flow
of the continuous phase is zero over these interfaces, the relative displacement of the
interfaces tends to squash the film as in a simple elongational flow (see Sect. 8.5.1).
In this situation, the normal viscous stress has the form μ˙∂, so the normal force is
μ˙∂ α R 2 , where ∂˙ = V / h is the elongation rate. It follows that the force goes as 1/ h,
i.e., much less quickly than for solid inclusions, for which the force goes as 1/ h 3 .
When the viscosity of the fluid in the dispersed phase is not negligible, the force on
two approaching inclusions will be somewhere between these two limiting situations.
In any case, we observe that it is ‘easier’ to bring together two bubbles in a foam
or two droplets in an emulsion than to bring together two solid particles immersed
in a liquid. However, when no other effects come into play, the force that must be
applied to bring together two inclusions at constant speed until they come into contact
(h ∞ 0) tends theoretically to infinity. In reality, when the two interfaces are very
close, van der Waals forces come on the scene.
6.4.1.2 Van der Waals Forces
When the thickness of the liquid film between two inclusions is small enough, van der
Waals interactions between the constituent elements of the inclusions may become
significant. The interaction energy between two such inclusions separated by a liquid
film can be estimated (see [9]) using (5.12) with
⎢ 1/2 1/2 ⎣2
K = K d0 − K c0 , (6.10)
where K d0 (resp. K c0 ) is the Hamaker constant for the interaction between elements of
the dispersed phase (resp. the continuous phase) in vacuum. Under these conditions,
the resulting force, derived from the interaction potential, goes as 1/ h 3 . Note that this
expression is valid in any kind of medium and thus applies in particular to a liquid
film separating two gaseous inclusions. For a given approach speed and for small
enough distances, the van der Waals force begins to dominate over hydrodynamic
repulsion and tends to push the two interfaces toward one another, thereby bringing
about coalescence of the inclusions.
6.4.1.3 Stabilisation
To prevent the disappearance of the films and hence stabilise the system, one must
introduce entities producing repulsive forces between the interfaces that are stronger
than the forces of hydrodynamic repulsion. There are no such forces in pure liquids,
which cannot therefore form stable films. However, repulsive forces can be intro-
duced by adsorbing surfactant molecules onto the interfaces. These are amphiphilic
molecules which have affinities for two incompatible media. They comprise several
chemical groups, some of which mix spontaneously with the molecules of one of
the media but not the other, while the other groups of the same molecule have the
opposite preference. In the commonest case, these molecules have a hydrophilic
‘head’ and a hydrophobic ‘tail’. These molecules thus tend to arrange themselves
over the interfaces between two liquids, or indeed at the free surface of a liquid. Note
that, in order to stabilise emulsions, one can also use polymers or particles which
adsorb over the interfaces. But the presence of surfactants in solution has another
advantage. Indeed, they lower the interfacial energy of the system which, according
to (6.1), favours the dispersion of one phase in the other and thereby facilitates the
preparation of a mixture.
The forces due to the presence of surfactants on the interfaces may be of electro-
static origin. When an ionic surfactant is adsorbed on the surface of the inclusions, a
double layer forms within the liquid (see Sect. 5.4), and the double layers on either
side will repel one another (see Fig. 6.6i), giving rise to a repulsive interaction poten-
tial that grows exponentially as the distance is reduced. There may also be a phenom-
enon of steric repulsion or steric hindrance (see Sect. 5.5) when the polymer chain
of the surfactant is hydrophilic and attached to a hydrophobic head (see Fig. 6.6ii).
In this case, the energy of repulsion increases very quickly from the moment when
6.4 Stability 215
A h A
h h
A B A A B A
(i) (ii)
Fig. 6.6 Different effects able to stabilise two nearby inclusions (distance h apart) with the help
of surfactant molecules (top) i electrostatic repulsion of the double layers, ii steric repulsion of the
molecules
the chains on the two interfaces begin to overlap. Finally, the total potential energy
associated with all the interactions taken together, and in particular the van der Waals
attraction plus the repulsion due to surfactants, will assume similar forms to those
described for colloidal particles (see Fig. 5.11), and there may be an equilibrium dis-
tance (the equilibrium thickness of the film) associated as always with a minimum
of this total potential energy.
6.4.1.4 Producing an Emulsion
When two phases are mixed together in the presence of a surfactant, the kind of
mixture obtained, i.e., the ‘choice’ of which will be the dispersed phase and which
the continuous phase, depends on the affinity of the amphiphilic molecules with the
two main phases. The amphiphilicity is quantified by the parameter known as the
hydrophilic–lipophilic balance (HLB). This is a partly qualitative assessment because
there is no underlying theoretical framework for this definition and there are several
versions, referring to the difference or the ratio of the affinities of the lipophilic and
hydrophilic parts of the molecule. The ‘neutral’ value is 10 or 7 depending on which
definition is used. A value close to zero indicates that the molecule is more soluble
in oil than in water, while a value above the equilibrium value indicates the opposite.
We obtain a water-in-oil emulsion (water being the dispersed phase and oil the
continuous phase) by gradually adding an aqueous phase to oil in which we have
previously mixed amphiphilic molecules that are soluble in oil and have a low enough
HLB. On the other hand, dispersing oil in an aqueous solution in which we have pre-
viously mixed amphiphilic molecules with a high enough HLB, we obtain an oil-in-
water emulsion. Finally, quite generally, when we mix water, oil, and an amphiphilic
molecule, or emulsifier, to obtain an emulsion, the continuous phase is the one in
which the surfactant mixes more easily. This is Bancroft’s rule.
6.4.1.5 Coalescence
Inclusions can coalesce even if they have been stabilised by amphiphilic molecules.
Indeed, these molecules are subject to thermal agitation, and this causes them to
move around the interface and from time to time to enter the droplet. This means that
the coating of molecules on the surface of the inclusion does not remain uniform and
constant. Two nearby inclusions may find that their liquid comes into contact due
to the formation of a short-lived opening in the amphiphilic coating resulting from
these fluctuations, i.e., a tiny region of the interface that is free of surfactant. Full
coalescence can then occur very rapidly. This situation will be less common if the
interface is coated with a higher concentration of surfactants.
Coalescence is thus largely a surface phenomenon. The volume which coalesces
per unit time is in fact simply proportional to the area of one of the inclusions (for
a dispersion of inclusions of uniform size). Indeed, up to a multiplicative factor,
there are β/R 3 inclusions per unit volume, for each of which the probability of
coalescing per unit time is proportional to the surface area it makes available, that is,
proportional to R 2 . The total volume to coalesce per unit time is therefore proportional
to (β/R 3 ) × R 2 × R 3 , i.e., proportional to R 2 . This implies that coalescence can be
slowed down by reducing the size of the inclusions. In practice, coalescence is a very
slow phenomenon on our own scale of observation for inclusions of diameter less
than 100 nm, but very fast for inclusions of diameter greater than 10 µm. Another
possibility for hindering coalescence is to lower the temperature of the emulsion.
Indeed, it is due to thermal agitation of the surfactant molecules that the amphiphilic
coating will sometimes break open in certain places. This is therefore an activation
phenomenon like the displacement of molecules in a liquid (see Sect. 2.4.2), with
frequency obeying a law of the form f 0 exp(−W/kB T ).
6.4.2 Ostwald Ripening
The interfaces are never actually fully impermeable. Since the dispersed phase is
always to some extent soluble in the continuous phase, the concentration of molecules
of the dispersed phase contained within inclusions is of course much higher than in
the continuous phase, so they tend to diffuse into the continuous phase. At the same
time, molecules will also tend to return to the inclusion by diffusion and in the end
an equilibrium will be set up. At this point, each inclusion will find itself surrounded
6.4 Stability 217
A (molecules)
A A
Fig. 6.7 Diffusion of molecules from A through B, and transfer of a small inclusion to a larger one
by a cloud of dissolved molecules. However, due to the pressure difference across

the interface from the inclusion to the continuous phase (see Sect. 6.2.2), the energy
of the molecules in small inclusions will be greater than the energy of molecules in
large inclusions. More molecules will thus diffuse into the continuous phase from
small inclusions and, in a region close to two inclusions of different sizes, this leads
to an imbalance in the concentration of molecules which will give rise to a transfer
of molecules from the smaller inclusion toward the larger one. In the end, the largest
inclusions tend to swell up while the smallest just disappear. This is the phenomenon
known as ripening.
Consider an inclusion of fluid A of radius R, immersed in a liquid in which the
dissolved concentration of A is initially c0 . Near the inclusion, at a distance r √ R,
the concentration c(r ) of A is greater than c0 owing to the diffusion of molecules
from the inclusion (see Fig. 6.7). In particular, in the liquid very close to the interface
with the fluid A, energy considerations show that the concentration of A is
2ξAB βm
c R = c exp , (6.11)
8.3RT
where βm is the molar volume of the fluid A and c is the concentration of A close
to a very large volume of A (with a locally plane interface such that R ∞ ∞). We
thus find that the concentration of the fluid A around an inclusion increases with
increasing interfacial tension and decreasing inclusion radius, two factors that are
key to the ripening effect.
Consider now the kinetics of this phenomenon. Diffusion occurs through the
mechanisms described in Sect. 5.2.2, which lead to Fick’s second law (5.6). In spher-
ical coordinates, this equation has the form
⎡
νc ν 2 c 2 νc
=D + .
νt νr 2 r νr
To simplify, we shall solve this in the steady-state regime. This will obviously be a
very rough approximation when we examine the way the inclusions change size over
time. This amounts to assuming that there is an inexhaustible source of molecules

in the inclusion, whence the concentration distribution remains constant, but there
is a nonzero flow of molecules. In this case νc/νt = 0 and the above equation
has solution of the form c(r ) = A/r + B. Given that c(r ∞ ∞) = c0 , since the
concentration is not expected to change far from the inclusion, we find
R
c(r ) = c0 + (c R − c0 ) . (6.12)
r
According to Fick’s first law, which holds for the phenomenon of diffusion but which
we shall not prove here, the flow of molecules crossing the surface of a sphere of
radius r is Φ = −4αr 2 D∇c. So at a distance R, we thus find χ = 4α D R 2 (c R −c0 ).
We can then estimate the rate at which the droplet will empty itself, since this flow
can also be written d(4α R 3 /3)/dt, and the result is
dR D
= (c R − c0 ) . (6.13)
dt R
We now consider a mixture containing two populations of inclusions of different
sizes R0 and R0 /10. The rate of change of the size of the small inclusions is much
higher than that of the large ones, owing to the term c R in (6.13) which increases
very quickly as the size of the inclusion decreases, according to (6.11). The small
inclusions therefore empty to the benefit of the large ones.
In practice, ripening often occurs very quickly for droplets of diameter less than
100 nm and very slowly for droplets with diameters greater than 10 µm. The phe-
nomenon can be slowed down by using a dispersed phase that is poorly soluble in the
continuous phase, so that c is small. One can also dissolve solutes in the dispersed
phase which cannot be transferred to the continuous phase. Ripening is then ham-
pered by the fact that it would lead to an increase in the concentration of this solute
within the smaller inclusions.
6.5 Behaviour
In the following we shall assume that the inclusions produced when the material
is prepared do not coalesce or break up during the flow, so that the material does
not undergo any irreversible changes in structure. From the rheophysical standpoint,
emulsions and foams are in many ways analogous to suspensions, consisting as they
do of solid inclusions dispersed in a liquid. The main differences stem partly from
the deformability of the inclusions in emulsions and foams and partly from the fluid
properties of the material making up these inclusions. At low inclusion concentra-
tions, these two phenomena can limit or even reverse the ‘usual’ effect on the apparent
6.5 Behaviour 219
viscosity of the system due to the presence of inclusions when they are solid. More-
over, given the deformability of the inclusions, fluid mixtures can be obtained for
inclusion concentrations close to 100 %. In this context, agents dispersed in the mix-
ture to stabilise the interfaces can play a significant role in reducing the interfacial
energy, thereby facilitating deformation of the inclusions, but also in increasing the
viscosity of the films separating the dispersed and continuous phases. However, the
key features of the behaviour of emulsions and foams can be reviewed by consid-
ering the ideal case of pure fluid inclusions in a pure continuous liquid phase, with
a given interfacial tension, i.e., neglecting the existence of surfactant films ensuring
the stability of the system.
As for suspensions, even in very simple cases such as very low concentrations
and spherical inclusions, the calculations are too complex to allow a detailed analy-
sis of the physical origin of the induced effects on the basis of simple arguments.
However, it is easy to understand qualitatively the consequences of the fluid nature
and deformability of the inclusions for the behaviour of the mixture by going back
over the calculation for a suspension in Sect. 3.4.1, but this time with deformable
fluid inclusions. We assume that the behaviour of the mixture remains unchanged
if all the inclusions are brought together into a layer of thickness h parallel to the
planes and centered on a median plane at y = H/2, as shown in Fig. 6.8. We then
simplify the problem by treating the central layer as a homogeneous fluid of viscosity
μ. This assumption is important because it means that for the time being we ignore
any particular role of the interfaces. The lower and upper layers are sheared with a
shear rate Ω̇0 , while the central layer has shear rate Ω̇ . We then obtain the velocity
field

 H −h

 Ω̇0 y, 0<y< ,

 ⎡ 2
⎧ H −h H −h H −h H +h
vx = Ω̇0 + Ω̇ y − , <y< ,

 2 2 2 2

 H +h

⎪ Ω̇ h + Ω̇0 (y − h), <y<H.
2
Conservation of momentum implies that the shear stress is constant across the whole
fluid thickness, whence γapp = μ0 Ω̇0 = μΩ̇ . The apparent shear rate is obtained by
dividing the maximum speed by the total fluid thickness to obtain
1⎨ ⎩
Ω̇app = Ω̇ h + Ω̇0 (H − h) .
H
We thus deduce the apparent viscosity μapp = γapp /Ω̇app , which, using the maximum
packing fraction, can be expressed as
⎡
μ0 τ −1
μapp = μ0 1 − 1 − . (6.14)
μ τm
Fig. 6.8 Effect of the pres-

ence of fluid inclusions on
the behaviour of the mixture φ
(upper), making the approxi- H y
mation that the inclusions are
gathered together in a layer
of thickness h with packing 0
fraction τm (centre). The
velocity profile (lower) is not
linear (dashed line), but is
piecewise linear (continuous
line), owing to the fact that the
different layers do not have h φm
the same viscosity
μ0
μ
μ0
The difference with the expression (3.13) found for a solid suspension is that the
factor multiplying −τ/τm , i.e., 1 − μ0 /μ, is less than unity and can be positive
or negative depending on whether the ratio μ0 /μ is less than or greater than 1.
As a consequence, the effect of the inclusions is very similar to the effect of the
solid particles in a suspension when the inclusions are much more viscous than the
continuous phase. On the other hand, this situation is reversed when the inclusions are
much less viscous, for then the viscosity of the mixture decreases when the inclusion
concentration is increased. It is important to bear in mind that the basic assumption
leading to these results is the unconstrained deformability of the inclusions. A more
complete investigation of the different concentration regimes will allow us to identify
the relative impact of the various phenomena (in particular the presence of interfaces)
on the behaviour.
6.5.2 Concentration Regimes
If we gradually introduce fluid inclusions of the same size into a liquid, we go

successively through four concentration regimes associated with the different types
of interaction between the inclusions.
6.5 Behaviour 221
(a) (b) (c) (d)
Dilute Semi-dilute Concentrated Compact

φ<<1 0.01<<φ<φc φc<φ<1 φ−>1
Fig. 6.9 Schematic views of the internal structure of an emulsion or a foam in the main concentration
regimes. a Dilute regime. b Semi-dilute regime. c Concentrated regime. d Compact regime
6.5.2.1 Dilute Regime
When the concentration of inclusions is low enough, i.e., τ at most a few percent,
the inclusions will be far enough apart (see Fig. 6.9a) to ensure that the perturbations
induced by the presence of one of them on the velocity field around its nearest neigh-
bours will remain negligible. Put another way, in terms of their effect on the velocity
field, it is just as though each of them were alone in the liquid. This is analogous
to the definition of the dilute regime for suspensions, except that the perturbation
distance is generally less than for a solid inclusion, owing to the deformability and
fluid nature of the inclusion. Indeed, these features allow the inclusion to put up less
resistance to the motions of the continuous phase.
6.5.2.2 Semi-dilute Regime
When the inclusion concentration is not so low, i.e., τ 1 %, the distance over
which the flow of the continuous phase is perturbed by the presence of an inclusion
is greater than the distance between two neighbouring inclusions (see Fig. 6.9b).
Everything happens as though the inclusions were interacting ‘hydrodynamically’.
The velocity field can no longer be determined by a calculation assuming that each
particle is isolated. Rather, the whole ensemble of inclusions must be taken into
account. The semi-dilute regime ranges from a concentration of a few percent up
to a critical concentration τc , for which there is a continuous network of contacts
between the inclusions. At this point we reach the concentrated regime.
6.5.2.3 Concentrated Regime
In this regime, the concentration is such that the inclusions form a continuous network
of contacts right across the sample when the elements are in a disordered configu-
ration. The inclusions are then jammed into this network and in order to break it,
the inclusions must be deformed. In this regime, there is viscous dissipation asso-
ciated with hydrodynamic interactions of the kind encountered in the semi-dilute
regime, but also associated with unjamming of the contact network, which involves
deformation of the interfaces. The concentrated regime extends from the critical per-
colation concentration τc associated with the formation of a first contact network
(see Fig. 6.9c), up to the compact regime (τ ∞ 1) in which the inclusions are highly
deformed and separated only by thin films of interstitial liquid (see Fig. 6.9d).
6.5.2.4 Compact Regime
This regime is encountered primarily in the so-called dry foams. In this case, it
has been known since [13] that the structure has rather specific characteristics. In
particular, the bubbles have polyhedral shapes such that two consecutive faces form
an angle of 120≈ and the angle between two lines of intersection of such faces is
109.6≈ . The simplest packing structure here is the one named after Kelvin, made up of
identical octahedra, but more complex structures with non-uniform size distributions
are also possible (for a given inclusion concentration).
6.5.2.5 Behaviour in Different Regimes
In the dilute and semi-dilute regimes, it can be shown by a similar method to the one
used for suspensions of solid particles in Sect. 3.3 that a suspension of fluid inclusions
is Newtonian if the droplet distribution is isotropic and constant. This assumes that
the distribution of any inclusion deformations and orientations that may occur due
to the flow of the system is stationary on average over a large enough volume of
the mixture. In reality, the transient nature of these deformations introduces specific
effects on the behaviour, e.g., a reduction in the viscosity or viscoelastic effects. In
the following, we shall investigate these various phenomena separately.
In the concentrated regime, deformation of the interfaces can play a key role on the
behaviour of the mixture. It follows that the mechanical behaviour is fundamentally
different from what is observed in the other regimes.
6.5.3 Dilute Regime
Here we consider the dilute case under the same conditions as for a solid particle
suspension. We thus aim to solve for the flow of a large volume of fluid of viscosity
μ0 containing an inclusion of another fluid of viscosity μ. The whole difficulty in
the present case comes from the fact that this inclusion can deform during the flow.
However, this deformation is limited by the effects of the interfacial tension and an
equilibrium will eventually be set up between the stresses associated with the flow
of the continuous phase and those associated with the stretching of the interface.
6.5 Behaviour 223
For low capillary numbers, the inclusions will deform only slightly, while for higher
capillary numbers, they will be able to deform to a greater extent. Here we distinguish
two asymptotic cases which have been given a detailed theoretical analysis [2]:
• Ca 1. The inclusions can deform freely. To begin with, while the droplets are
still spherical, the viscosity is given by
μ 5μ − 5μ0
=1+τ . (6.15)
μ0 3μ0 + 2μ
The general trends predicted by this formula and shown in Fig. 6.10 are consistent
with what we have found so far in the extreme cases:
– The viscosity of the mixture is the same as that of the continuous phase when
the two liquids happen to be the same μ0 = μ.
– The viscosity becomes the same as that of a solid particle suspension, that is,
μ/μ0 = 1 + 2.5τ, when μ/μ0 ∞ ∞, in other words, when the fluid in the
inclusions is much more viscous than the fluid in the continuous phase.
The main prediction of (6.15) is consistent with the results of the qualitative analy-
sis in Sect. 6.3.1, i.e., the viscosity of the mixture is greater than that of the con-
tinuous phase if the inclusions are more viscous, i.e., if μ/μ0 > 1, and less than
that of the continuous phase if the inclusions are less viscous, i.e., if μ/μ0 < 1.
In particular, for a foam, where μ = 0, we have μ/μ0 = 1 − 5τ/3.
• Ca 1. The inclusions can no longer deform:
μ μ0 + 2.5μ
=1+τ . (6.16)
μ0 μ0 + μ
Once again, this expression, the general trends of which are shown in Fig. 6.10,
does indeed predict that the mixture will have the same viscosity as a solid particle
suspension, viz., μ/μ0 = 1 + 2.5τ, when μ/μ0 ∞ ∞, i.e., when the inclusions
contain a much more viscous fluid than the continuous phase. However, the predic-
tions of this expression are somewhat more surprising in the other limiting cases
considered above. So when the two liquids are identical and μ = μ0 , the viscosity
of the mixture is higher than that of the pure liquid of the same viscosity, and in
fact we have μ/μ0 = 1 + 1.75τ. In the special case of a foam, where μ = 0, the
viscosity of the mixture is μ0 (1 + τ), which is also higher than the viscosity of the
continuous phase. These results can be explained by the fact that the presence of
non-deformable inclusions induces perturbations in the velocity field which lead
to supplementary viscous dissipation in the continuous phase. Note, however, that
this non-deformability is not the only explanation for the viscosity of a solid par-
ticle suspension, since whatever the value of μ/μ0 , the viscosity of the resulting
mixture remains less than 1 + 2.5τ. This comes from the fluid nature of the inclu-
sion. Indeed, in contrast to the assumption made when discussing suspensions in
Chap. 3, the speed of the continuous phase over the interfaces between the two
Fig. 6.10 Viscosity of a dilute dispersion of inclusions in a liquid for different values of the ratio of
viscosities of the two phases and a range of values of the capillary number. Upper curve Very small
capillary number, i.e., inclusions deform only slightly. Lower curve Very high capillary number,
i.e., inclusions deform freely. The viscosity takes intermediate values for intermediate values of the
capillary number
phases is not generally zero. In this situation, within a non-deformable inclusion,

the fluid exhibits a complex flow in such a way as to adapt to a position-dependent
velocity over its interface with the continuous phase.
Finally, it is important to note that the assumptions made in the above calculation, i.e.,
spherical inclusion, beginning of flow, do not necessarily correspond to conditions
implying Newtonian behaviour. A more precise calculation should take into account
the average equilibrium deformation state for the inclusions in the mixture. The
above formulas are merely approximations giving the general trends in the apparent
viscosity as a function of the various parameters. Note also that these systems are
always viscoelastic to some extent, since the inclusions will never reach their equi-
librium deformation instantaneously, but we shall disregard this point which is not
essential if we only wish to describe the main flow characteristics of these materials.
Beyond the dilute regime, hydrodynamic interactions become more complex and
it is very difficult to calculate the velocity field in a material made from two well
mixed fluid phases. As for suspensions, however, we can still obtain a rough idea
of the dependence of the apparent viscosity on the concentration and other system
6.5 Behaviour 225
parameters by making suitable approximations. A first step makes use of our prelim-
inary calculation in Sect. 6.3.1. In the case where the interfaces play no specific role,
this gives an idea of the dependence of the viscosity on the concentration (whatever
its value) for different values of the viscosity ratio. Another idea is to extrapolate
the trends observed on the basis of precise calculations and in specific situations for
the viscosity in the dilute regime. We then obtain the dependence of the viscosity,
not only on the concentration and the viscosity ratio, but also on capillary phenom-
ena, which gives us a more complete view of the situation. And finally, this line of
thought can be extended by applying the method used for semi-dilute suspensions
in Sect. 3.4.4, based on a differential analysis.
Consider for example a scheme in which the mixture is built up by successively
adding very small concentrations of inclusions whose effect on the viscosity of the
mixture will each time be the same as in a dilute regime. We can then carry out a
similar calculation to the one in Sect. 3.4.4, but using (6.15) or (6.16) as the basic
equations for the dilute regime and taking into account the fact that the viscosity
ratio is variable. Indeed, it will be the ratio of the viscosity of the inclusions and the
viscosity of the mixture in the previous step. Taking into account the effect of the
interactions of the inclusions with the inclusions already present, as for the Krieger–
Dougherty formula (3.19), we end up with [3]
⎡ ⎡
τ −2.5τm μ μ/μ0 + 2.5ϕ 1.5
Ca 1 =∝ 1− = , (6.17)
τm μ0 1 + 2.5ϕ
⎡ ⎡
τ −2.5τm μ ϕ − μ/μ0 −2.5
Ca 1 =∝ 1− = . (6.18)
τm μ0 ϕ−1
These expressions refer to a maximum packing fraction defined here as the maximum
packing concentration of undeformed inclusions in a disordered configuration. In
general, τm is slightly greater than τc (see Sect. 3.4). One might be tempted to use
the maximum packing fraction for the deformed inclusions, but this would not be
justified in the above context since we know that, in that particular case, we would
be dealing with a different regime, the concentrated regime, for which a new type of
interaction becomes important.
For this regime, there is a network of inclusions in contact with one another in the
system at rest (see Fig. 6.9c). Note that what we refer to as contact here corresponds
to inclusions that are very close to one another, but nevertheless separated by a film
of interstitial liquid. The network of contacts is ‘jammed’ in the sense that, as long as
the imposed deformations are weak enough, it will not break, but rather will deform
in a reversible way, and when the external force is removed, it will return to its initial
state, with the inclusions reverting to their original shape and position. The material
γ
A A
A
F B F F
B
B
E
E C E
C
D D D C
(i) (ii) (iii)

γ
γa
a
(iv) (v) (vi)

Fig. 6.11 Configuration and shapes of inclusions in the concentrated regime during a simple shear,
for an arbitrary initial distribution (upper). In the initial state (i), the inclusions are in contact with one
another and only slightly deformed. The inclusion B with the dashed outline is not in contact with the
central inclusion. During the imposed macroscopic deformation, the inclusions undergo different
deformations, shown here by grey level (with darker grey for greater deformation), depending on
their position within the network (ii). After a critical deformation, the local configuration is changed
irreversibly. The central inclusion no longer has exactly the same neighbours (iii), e.g., the inclusion
E is no longer in contact with it. The same sequence is shown at the bottom for a compact system
and cubic inclusions: initial state (iv), deformation of the local network (v), reconfiguration (see
text) (vi)
thus appears to be solid. Beyond a certain critical deformation, the network will
break, and although it may eventually form again, if we keep on applying the same
stresses, it will end up by breaking once again. The material then behaves rather as
a liquid, deforming indefinitely, or put another way, flowing. These transformations
are reversible. The material returns to the same solid state when it is left to rest after
such a flow. So what we have here is in fact a yield stress fluid.
The interactions that need to be taken into account are thus the stresses relating
to the interfaces, which tend to make the inclusions adopt shapes minimising the
interfacial energy, and viscous dissipation relating to the flow of the fluids within the
inclusions and in the continuous phase. In the solid regime, at low speeds, viscous
dissipation is negligible and the material behaves in an essentially elastic way, i.e.,
deformations of the interfaces provide a way of storing or releasing energy. If a force
is applied to the system, it deforms, but after removing the force, it goes back to its
original configuration. In the liquid regime, the two types of energy, viz., interfacial
and viscous, can both play a role.
6.5 Behaviour 227
6.5.5.1 Solid Regime
When a deformation is imposed on the system, the network of contacts between

inclusions deforms, and in doing so, deforms the inclusions themselves. Since the dis-
tribution of contacts is generally disordered, the distribution of deformations induced
in the inclusions will tend to be non-uniform, as will the stress distribution. An anal-
ogous phenomenon occurs in a disordered pile of grains (see Chap. 7). The stresses
distribute themselves according to tortuously shaped chains of force of different mag-
nitudes, subjecting the grains to a broad range of different forces depending on their
position within the network. Figure 6.11 illustrates this phenomenon for an emulsion
or a foam. Depending on the precise relative positions of the inclusions within the
initial network (at rest) (see Fig. 6.11i), the inclusions can slip by or be squashed up
against one another, giving rise to deformations of many different magnitudes (see
Fig. 6.11ii).
As the inclusion concentration increases, the problem becomes slightly simpler.
The network is tighter and the inclusions more firmly jammed together, which means
that, for a given macroscopic deformation, the fraction of deformed inclusions will
increase. In the extreme case of the compact regime, all the inclusions undergo a
deformation roughly equal to the imposed macroscopic deformation. This situation
is shown for an ideal configuration of a regular distribution of cubic inclusions in
Fig. 6.11iv. Such a distribution is not realistic, because it would not be stable, for
the angles formed by the films should evolve very quickly to the Plateau values.
However, it can be used to carry out a direct and accurate calculation which gives a
good idea of the rheophysics of these high concentration systems.
When a cube undergoes a simple shear deformation (see Fig. 6.11v), the length
of the perpendicular sides in the plane of the deformation and in the direction of
motion becomes a 1 + Ω 2 ⇒ a(1 + Ω 2 /2). The area of the corresponding faces,
which are the only ones to be deformed, thus becomes a 2 (1 + Ω 2 /2). The surface
energy of the initially cubic inclusion can then be written η = ξAB a 2 (1+Ω 2 /2) and,
assuming that no other forms of energy are involved here, the force associated with
this deformation is F = ∇η = aξAB Ω . The stress that must be applied to the cube
is equal to the force divided by the area a 2 , which yields γ = ξAB Ω /a. Since this
must hold for each cube, this stress is also the average stress in the system, which is a
function of the macroscopic deformation Ω . We can now deduce the elastic modulus
of the system to be G = ξAB /a. This kind of approach was originally developed by
Princen [4].
For a real configuration, the surface energy stored during the deformation is dis-
tributed non-uniformly in the inclusions, but on average one would naturally expect
this shear modulus to be of the same order as, and simply proportional to, the value
obtained in the above ideal case:
ξAB
G=k , (6.19)
a
up to a correcting coefficient k which increases with the concentration and depends

on the size and shape distribution, but also on the spatial distribution of the inclusions.
The present state of knowledge concerning the value of k and the way it depends on
these features remains rather limited.
Note that, in all the above arguments, we have neglected the viscous effects asso-
ciated with the flow of the two fluid phases. This gave the impression that the material
would be purely elastic in the solid regime. Although this is not quite accurate, it
nevertheless provides a good approximation.
6.5.5.2 Yield Stress
When the imposed deformation is large enough, the network is expected to break.
In this case, even if the stress is removed immediately afterwards, thus leaving the
material at macroscopic rest, the inclusions will not come back exactly to their initial
configuration. On the local level, this phenomenon happens in the way illustrated
in Fig. 6.11. Following the deformation and repositioning of each of its original
neighbours (Fig. 6.11ii), the central inclusion is once again surrounded by neigh-
bours (Fig. 6.11iii), but they are no longer exactly the same neighbours as in the
initial configuration (Fig. 6.11i). Bearing in mind what was said above about the
non-uniformity of the distribution of deformations in such a system, the deformation
able to induced a local change of configuration may vary significantly from one point
of the sample to another. However, we may assume that, in a system comprising a
very large number of inclusions, the configuration changes without modifying the
(statistical) distribution of the possible arrangements. For this reason, after the net-
work has broken locally at a few points where it was at its weakest, if the stress is
maintained, it will be possible to break the network once again a moment later at a
few other points, and so the process goes on. In this way, the material begins to flow.
The minimum macroscopic stress able to maintain this process is the yield stress.
Above this stress, the material is in its liquid regime.
Given the non-uniformity of the stress distribution in the system, we must turn once
again to the cubic inclusion model in order to find an expression for the yield stress.
In this case, the critical deformation is associated with a sudden reconfiguration of
the films (see Fig. 6.11vi), which we shall describe more precisely below for a more
realistic configuration. For our model system, this critical local deformation is also
the critical macroscopic deformation Ωc , which is of order 1. In this case, following
the same procedure as above, we deduce the critical stress reached for Ω = Ωc = 1
from → the effective extension a 1 + Ω 2 of certain sides of the cubes, with the result
ξAB / 2a.
By the same argument as for the shear modulus, we deduce for an arbitrary
configuration that the yield stress has the form
ξAB
γc = k ◦ , (6.20)
a
6.5 Behaviour 229
γ
b b b b
a c a c a c' d
d d d a
e e e e
(i) (ii) (iii) (iv)

Fig. 6.12 Successive stages in the deformation of inclusions in the compact regime as a result of
a simple shear. i Initial configuration. ii Stretching of films a, b, e, and d and shrinking of c. iii
Disappearance of film c, an unstable configuration. iv Instantaneous reconfiguration with formation
of a new film c◦
where k ◦ is a coefficient depending on the exact characteristics of the structure of

this system.
The reconfiguration of the network after breakage is relatively abrupt because the
inclusions have stored up elastic energy which is in part released by this breakage. In
the case of a real configuration in the compact regime, this sudden reconfiguration
occurs in a rather particular way. Due to the imposed deformations, it may be that
one of two contiguous films is stretched while the other is shrunk (see Fig. 6.12). This
situation can lead to the disappearance of the shortened film, so that for an instant
there is a junction of four films, an unstable situation. Immediately afterwards, during
the deformation, the films rearrange themselves to reform a stable 3-film junction.
The system thus naturally reconfigures itself into a configuration that is geometrically
identical to the initial configuration, but at the same time, this process has allowed
the material to deform. This is known as a T1 event.
6.5.5.3 Liquid Regime
It is found experimentally that, under simple shear, the flow curve of an emulsion or
a foam in the concentrated regime is well represented by a Herschel–Bulkley model
(see Sect. A.10.7):
γ > γc =∝ γ = γc + k Ω̇ n . (6.21)
We have already described the process that leads to the liquid regime, i.e., for a stress
greater than the yield stress, the network breaks, then reforms almost instantaneously,
then breaks again, and so on. This picture allowed us to find the expression (6.20)
for the yield stress, based on the fact that, for small shear rates, it is the deformations
of the interfaces that play the dominant part in the mechanical behaviour of the
material. While this description is well suited to a very slow or quasi-static flow, it
is difficult to apply to a fast flow because the network may find itself simultaneously
in the process of breaking at certain points and reforming at others. Furthermore,
the viscous dissipation due to the flow of the two fluid phases increases with the
shear rate and may possibly play a significant role at some point. At intermediate
shear rates, the effects of deformation of the interfaces and viscous effects both play
an important role and are probably coupled. Finally, at the present time, there is no
‘simple’ physical explanation for the Herschel–Bulkley form of the constitutive law
of these systems. However, it is useful to note that the rheophysics of these systems
is in many ways analogous to the rheophysics of repulsive colloidal suspensions (see
Sect. 5.8), and this gives reason to hope that some general theoretical framework will
one day be devised for such systems.
References
1. Grace, H.P.: Dispersion phenomena in high-viscosity immiscible fluid systems and applications
of static mixers as dispersion devices in such systems. Chem. Eng. Comm. 14, 225–277 (1982)
2. Taylor, G.I.: The viscosity of a fluid containing small drops of another fluid. Proc. Royal Soc.
Lond. Ser. A 138(834), 41–48 (1932)
3. Pal, R.: Viscous behavior of concentrated emulsions of two immiscible Newtonian fluids with
interfacial tension. J. Colloid Interface Sci. 263, 296–305 (2003)
4. Princen, H.M.: Rheology of foams and highly concentrated emulsions. J. Colloid Interface Sci.
91, 160–175 (1983)
Further Reading
5. Barthès-Biesel, D.: Microhydrodynamics and Complex Fluids. CRC Press, New York (2012)
6. Cabane, B., Hénon, S.: Liquides - Solutions, dispersions, émulsions, gels. Belin, Paris (2003)
7. Cantat, I., Cohen-Addad, S., Elias, F., Graner, F., Höhler, R., Pitois, O., Rouyer, F., Saint-Jalmes,
A.: Foams: Structure and Dynamics. Oxford University Press, Oxford (2013)
8. de Gennes, P.G., Brochard-Wyart, F., Quéré, D.: Gouttes, bulles, perles et ondes. Berlin, Paris
(2005)
9. Everett, D.H.: Basic Principles of Colloid Science. Royal Society of Chemistry Paperbacks,
London (1988)
10. Larson, R.G.: The Structure and Rheology of Complex Fluids. Oxford University Press, Oxford
(1999)
11. Oswald, P.: Rheophysics: The Deformation of Flow and Matter. Cambridge University Press,
Cambridge (2009)
12. Plateau, J.: Statique expérimentale et théorique des liquides soumis aux seules forces molécu-
laires. Gauthier-Villars, Paris (1873)
13. Weaire, D., Hutzler, S.: The Physics of Foams. Oxford University Press, Oxford (2000)
Chapter 7
Granular Materials
Abstract The relevant interactions in the systems discussed so far are mainly ‘soft’,
in the sense that the basic elements interact at distances of the same order or much
greater than their own size. This situation changes when the elements are rigid and
come in very close proximity compared with their own dimensions. We consider such
systems in this chapter. They have a sufficient concentration of solid non-Brownian
particles to ensure that hard interactions, involving direct contact between particles
(although possibly with a thin film of liquid between them), play an essential role in
the behaviour of the system. The rheophysical approach to systems governed by hard
interactions is somewhat delicate. The reason is that the resulting highly nonlinear
form of interaction depends strongly, even critically, on the spatial arrangement of the
particles, which itself may depend on the flow history of the material. We begin by
examining the characteristics of the main kinds of interaction within such systems.
We then describe the characteristics of three kinds of behaviour associated with each
of the main types of interaction when it happens to dominate: the friction, lubrication,
and collision regimes. We also consider the transitions between these regimes.
7.1 Introduction
Most of the interactions taking place in the systems we have been considering up to
now have been ‘soft’, in the sense that either the elements interact at distances of
the same order or much greater than their size, as happens in dilute or semi-dilute
suspensions and colloidal dispersions, or the elements deform easily when they come
within short range and the level of interaction increases beyond a certain point, as
happens in polymers, foams, and emulsions. As a consequence, in all these cases,
the strengths of the interaction forces increase rather gradually as the elements are
brought within closer range.
The situation is quite different when the elements are rigid and are brought to
within very short distances of one another as compared with their own dimensions.
If two solid particles are gradually brought into contact while immersed in a fluid,

DOI: 10.1007/978-3-319-06148-1_7,
232 7 Granular Materials
there is to begin with a hydrodynamic repulsion force, required to expel the fluid from
the gap between them, and this force diverges when the distance between the surfaces
tends to zero. When the particles are actually in contact, at a distance of the order of
their surface roughness, their centres can only be brought very slightly closer together,
since otherwise they will break, and even then, a very strong normal force is needed.
In contrast with the other systems, the interactions arising in a granular material can
be described as ‘hard’, in the sense that the corresponding forces eventually diverge
whenever we attempt to bring the particles too close together. The same kind of
behaviour is observed on a macroscopic scale with a granular material made from a
pile of solid objects. This constitutes a perfectly novel situation in fluid mechanics
because, under certain circumstances, the normal force imposed on the material
can be increased significantly in some specific direction without necessarily causing
the material to flow. On the other hand, two contacting particles can be displaced
tangentially by imposing a tangential force proportional to the force with which the
particles are being pushed together (Coulomb friction). A granular material can thus
be sheared by imposing a moderate tangential force, even though the same material
would appear to be almost rigid under the action of a normal force.
These are the systems we shall be considering here. We shall refer to them as
granular materials. They contain a high enough concentration of non-Brownian solid
particles to ensure that the hard interactions due to direct contact between these
particles (possibly with a thin film of liquid between them) play a primary role in
the behaviour of the system as a whole. At first glance, the relevant materials would
be powders, sands, aggregates, etc., made from solid elements immersed in a gas.
In these materials, the interactions between solid elements will naturally play a key
role in the macroscopic behaviour of the system, since this system is simply an
assemblage of these elements. However, we shall also include under this heading
materials comprising a high concentration of solid particles immersed in a liquid.
These are in fact suspensions in the compact regime (see Sect. 3.4), which lie at the
boundary between suspensions and granular materials.
In rheophysics, the investigation of systems governed by hard interactions is a
delicate matter. The reason is that the kind of interaction occurring in these systems
depends to a large extent, in fact even critically, on the spatial arrangement of the
particles, and this itself depends on the flow history. To understand how very small
changes in configuration can have a spectacular impact on the mechanical behaviour
of the system, we need only think about two solid balls. When these are very close
together, they can move freely with respect to one another, but when they are in
contact, there is no way of bringing them closer together and in order to move
them tangentially with respect to one another, a finite force is required due to the
friction between the two solid surfaces. If we extrapolate to a macroscopic level, this
phenomenon suggests that there may be a sharp transition from fluid behaviour to
solid behaviour, or conversely.
It is useful to begin by examining the main kinds of interactions in these systems,
and this will be done in Sect. 7.2. We shall then discuss in Sect. 7.3 the consequences
of the configuration effects mentioned above, and in particular the role of the system
state in its behaviour and possible future evolution during flow. In Sect. 7.5, we
7.1 Introduction 233
describe the characteristics of three kinds of behaviour associated with each of the
main types of interaction when it happens to dominate: the friction, lubrication, and
collision regimes. The key point is then to be able to determine which regime is
effective by examining the characteristics of the system. One possible approach is
put forward in Sect. 7.4. We identify the predominance of each regime by comparing
the energies coming into play on the local level for each type of interaction. Finally, in
Sect. 7.6, we discuss the behaviour of granular materials in the intermediate regimes
in which two types of interaction are involved and which we refer to as the friction–
collision regime and the friction–lubrication regime.
7.2 Main Types of Direct Interaction
In this section we shall review the main kinds of interaction operating within such
systems, distinguishing lubricated contacts for which the particle surfaces are sepa-
rated by a layer of interstitial fluid, frictional contacts for which there is prolonged
contact between two solid surfaces, and collisional contacts for which the duration
of contact is very short and the two particles exchange or dissipate part of their
kinetic energy. Each interaction is viewed here in an ideal context. This will allow
us to establish the dependence of these forces on the main physical parameters, but
of course in practice, various other phenomena will inevitably perturb these basic
trends.
7.2.1 Lubricated Contact
When two solid particles move together, this will lead to specific hydrodynamic
effects due to the flow of the interstitial fluid in directions perpendicular to the line of
approach. The physical principle is the same as when two plane surfaces separated
by a liquid layer are brought together or pulled apart (see Sect. 8.5.1). As for flow in
a duct, the radial flow of the fluid which tends to be expelled from the space between
the solid surfaces requires a pressure gradient between the central region and the
outside, and this implies that the pressure in this region must differ from the pressure
prevailing outside. One must therefore apply a normal force on the solid objects to
bring them closer together. Since the flow speed increases as the distance between
the solid surfaces gets smaller (for fixed approach speed), the pressure gradient must
increase when the distance between the two particles gets smaller, and so therefore
must the normal force. For example, for two solid spheres of radius R, the normal
force thereby induced will be given to first order by F = −μ0 (3π R 2 /2h)V .
To impose a relative tangential motion on two solid particles separated by a thin
layer of fluid, one must apply a tangential force associated with the shear of the
fluid in the direction of motion, and in the case of two spheres, a normal force
due to the fact that the rounded surfaces first move closer, then move apart during
d
ε
h
Fig. 7.1 Perpendicular cross-section of a plane surface showing how to characterise the surface
roughness. The dashed line is a plane (the envelope) passing through the peaks of the actual surface
Β
Α
ε
Fig. 7.2 Contact between two rough surfaces. The surfaces come into contact only at certain points
in regions A and B (grey circles). The thickness of the fluid film remaining between the two solid
surfaces is of the order of ξ
the motion, giving rise to an analogous effect to the one described above. These
hydrodynamic forces diverge when the distance between the solid surfaces tends to
zero since, at fixed speed, the shear rate tends to infinity as the sheared thickness tends
to zero. So for two spheres in relative tangential motion at speed VT , the tangential
component of the induced force is FT = μ0 π R ln(h/R)VT and its normal component
is FN = 1.2μ0 π R(R/ h)1/2 VT . We thus observe that, whichever component we may
consider, it will tend to infinity as R/ h tends to zero.
The divergence of the hydrodynamic force is of course entirely theoretical, since it
assumes that the solid surfaces are ideally planar or spherical. Naturally, any real sur-
face is imperfect, and in fact has a certain degree of roughness. For present purposes,
we shall assume a simple description of this idea. We shall consider an apparently
planar solid surface, at least on a sufficiently large observation scale. On the local
scale, the surface imperfections imply that the distance between the real surface and
an arbitrary plane will vary with the position (see Fig. 7.1). The surface roughness
can be treated to a first approximation as defined by the average value of this distance
h, taking the plane or envelope through the highest points of the surface as reference
plane. Note, however, that this method will not allow us to distinguish between a
rough surface characterised by sharp peaks and one comprising undulations. To deal
with that aspect of things, one can take into account the dimensions of the surface
imperfections in different directions, e.g., defining the surface roughness as the aver-
age diameter of spheres that could be placed under the envelope of the solid surface,
viz., ξ = ◦d, as shown in Fig. 7.1.
When two particles come together, their surfaces end up coming into contact at
several points (see Fig. 7.2). The force required to bring the particles slightly closer
together then depends on the local deformations of the solid phase. To a first approx-
imation, we may assume that this force increases (diverges) more quickly than the
hydrodynamic force as the distance decreases. Under such conditions, the minimum
7.2 Main Types of Direct Interaction 235
distance between two solid particles in contact remains of the order of their rough-
ness. In particular, this implies that the component of the repulsive force associated
with hydrodynamic effects will not actually diverge, since the distance between the
two solid surfaces will never tend uniformly to zero. There will always remain a
continuous fluid film (although with holes in it due to the points of contact) between
the two contacting particles (see Fig. 7.2). We can thus estimate the maximal values
of the hydrodynamic forces in the context of a lubricated interaction by appealing to
the above expressions and taking h = ξ. Under such conditions then, the maximal
value of the normal approach force will be roughly
3π R 2
F = −μ0 V . (7.1)
2ξ
7.2.2 Frictional Contact
Here we consider the interaction that results specifically from a prolonged contact
between the surfaces of two particles in relative tangential motion, which we shall
call a frictional contact. We shall thus neglect hydrodynamic effects as evaluated in
the last section. Despite its simplicity on the macroscopic scale of observation, the
physical phenomena involved here on the local scale during frictional contact are
extremely complex. In the standard temperature range, we may nevertheless identify
two important kinds of phenomena:
• The atoms or molecules of the surface interact through attractive forces such as
van der Waals interactions, sometimes metallic or ionic bonds, and so on, and this
leads to a certain adhesion between the two surfaces.
• The material undergoes elastic or plastic deformations around the points of contact,
e.g., in regions A and B of Fig. 7.2.
Other phenomena may also occur, such as cohesive capillary forces due to the pres-
ence of liquid between grains, physicochemical alteration of the surfaces by oxidation
or adsorption of polymers, surface wear, influence of the surrounding fluid, and so
on. Cohesive capillary forces can significantly modify the mechanical behaviour of
the system. Here we shall neglect all such effects.
Despite the complexity on the local level, it turns out that, on the macroscopic
scale, there is a very simple law relating the normal and tangential components N
and T, respectively, of the mutual forces between the two solid particles when their
two surfaces slip one over the other (see Fig. 7.3):
T = fN . (7.2)
In these expressions, f is the coefficient of Coulomb friction, which generally takes

values between 0.1 and 0.5 for friction between metals, between 0.05 and 1 for
polymers, and between 0.05 and 0.5 for ceramics. Qualitatively, the physical origin
N N
T T
T<fN T=fN
Fig. 7.3 Friction between two solid surfaces. Left the two solids are in contact over an area S, but
the tangential force is not strong enough compared with the normal force to induce motion. Right
the tangential force reaches the critical value allowing relative motion
of this law is as follows. When the normal force is increased, the solid surfaces are
deformed around the regions of contact (see Fig. 7.2), with the consequence that the
number of points of contact is increased, thereby increasing the resistance to relative
motion. Note that, when there is no slipping, i.e., when the relative speed of the solid
surfaces at the point of contact is zero, the tangential component is undetermined
and we just know that T ∞ fN (Fig. 7.3).
The friction force T also depends to some extent on the history of relative displace-
ment of the two solid objects (for constant contact area). In fact, the force needed to
maintain a given motion is often slightly less than the one needed to start that motion.
Furthermore, at low relative speeds, the motion often occurs in fits and starts, alter-
nating between moments of adhesive friction and slipping friction in what is known
as stick-and-slip motion. In this case, the instantaneous force fluctuates significantly
and it can be very difficult to maintain a strictly constant displacement speed. At
higher speeds, the friction force increases slightly with the speed.
7.2.3 Collision
When the contact between the two particles is very brief, the friction forces of the
last section become negligible compared with inertial forces. In this case, we speak
of a collision. This situation leads basically to an exchange or dissipation of kinetic
energy. To get a better understanding of what happens here, let us examine the
encounter between two identical perfectly elastic cubic particles with side a and
Young’s modulus E (see Sect. 2.5.3), coming into contact over one of their faces
as a result of moving at the same speed toward one another along the axis passing
through their centres, as shown in Fig. 7.4.
We analyse the situation in a frame moving with the centre of gravity of the two
particles. In this frame, each particle arrives with the same speed Vi and moves off
7.2 Main Types of Direct Interaction 237
Δa
a
Fig. 7.4 Collision between two solid elastic cubes. Left the cubes are some distance apart and not
deformed. Right the two cubes have come into contact and part of their kinetic energy has been
transformed into elastic potential energy due to the deformation they have undergone. The dashed
lines indicate the initial form of the cube
with the same speed Vf in the opposite direction after the collision. During contact,
which we assume to be frictionless, the speed of each particle decreases very quickly
to zero, whereupon its kinetic energy transforms into elastic potential energy. In
the limit of small deformations α = βa/a 1 (see Fig. 7.4), the elastic energy
gained through a deformation α is given by Eα 2 a 3 /2. The speed of each particle is
V = a α̇/2. The maximal deformation αm is reached when the initial kinetic energy
mVi2 /2 of the particle is completely transformed into elastic energy, i.e., when V = 0.
Conservation of the total energy then gives
2
1 2 α̇ 1 1
ma + Ea 3 α 2 = Ea 3 αm
2
. (7.3)
2 2 2 2
We can calculate the total time of contact
αm
dα
tc = 2 .
α̇
0
⇒
According to (7.3), this turns out to be proportional to m E/a. The prediction from
this simplified calculation turns out to be quite general: the time of contact is shorter
when the particles are light and ‘hard’, i.e., when their elastic modulus is high.
By symmetry, it is easy to show from (7.3) that the rebound speeds of each particle
will be the same but in opposite directions, i.e., Vf = −Vi . This is only true for
perfectly elastic particles. In practice, a collision also involves a certain proportion
of irreversible plastic deformation which will dissipate energy. The particles then
(a) (b)
Δx
A B A B
C D C D
Fig. 7.5 Effect of changed configuration of a granular material on its apparent concentration and
jamming state. Left loose body-centered cubic packing. Right the same granular material after
simple shear deformation, resulting in a more compact and more tightly jammed face-centered
cubic packing
lose some of their kinetic energy in the collision, whence their final speed is
Vf = −eVi , (7.4)
where e is less than 1. This factor, which varies slightly with the speed, is known as
the coefficient of restitution of the material.
7.3 Role of Configuration
7.3.1 Basic Principles
The effects of configuration on the behaviour of granular materials can be illustrated

by describing what is going on at the local scale. Consider for example four balls
stacked up very simply as shown in Fig. 7.5a. If such an arrangement is repeated in the
three main directions throughout a sample, we then have a face-centered cubic (fcc)
arrangement, which constitutes the loosest packing possible in a crystal structure, i.e.,
the one leading to the lowest solid fraction, viz., π/6 for spheres, or about 52 %. The
particles are then located one behind the other in long parallel rows (perpendicular
to the plane of the paper in Fig. 7.5), themselves arranged in parallel planes.
We now impose a simple shear on this arrangement in a direction perpendicular
to one of these planes. This induces a relative displacement of the two rows of balls.
When the balls have moved a distance equal to their radius in this direction, they find
themselves in the new configuration of Fig. 7.5b. We notice immediately that this
arrangement is more closely packed, since the height of the stack has decreased by
βx. Extrapolating to the whole sample, we may now have the most compact crystal
structure
⇒ possible, the face-centered cubic (fcc) structure, in which the solid fraction
is π/3 2, i.e., around 74 %.
7.3 Role of Configuration 239
Elastic membrane
Liquid Deformation
Fig. 7.6 Reynolds’ experiment, illustrating the dilatancy of a granular medium. Sand saturated
with water is put inside a container with deformable walls like a balloon. The level of water in the
column tells us how much water there is in the container. When the container is deformed (from left
to right), we expect the water level to increase due to the water extracting itself from the balloon
under the exerted pressure. However, the water level actually goes down. This is due to the fact that
the grains move slightly further apart in order to move relative to one another, which increases the
volume of the pores and hence the volume of water needed to fill them up
For an ideal granular material with the local structure depicted in Fig. 7.5, a simple
shear can thus induce changes in the solid fraction between the above two extreme
values. If we extrapolate this result to a large number of particles constituting a
granular medium, we would expect a material in a dense state to expand during
shear, while a rather loosely packed material will tend to compress somewhat, at
least at the beginning of the motion. In practice, for an arbitrary granular medium,
changes in configuration due to deformations of the medium tend to be rather more
complex. One must also take into account the disorder, any granulometric dispersion,
the non-sphericity of the particles, collective effects, and so on. However, the same
qualitative effects are found, associated with the same physical principles as those
brought out in this example, namely, the shear of a granular medium that is initially
at rest in a relatively densely packed arrangement involves a certain dilation of the
material.
7.3.2 Dilatancy
In the example of Fig. 7.5, in going from state (b) to state (a) as the material is
sheared, we induce an expansion of the volume occupied overall by the particles.
This phenomenon, known as dilatancy, occurs whenever we try to deform a granular
medium that has been compacted to any extent.
A simple experiment due to Reynolds (1885) illustrates the effects of dilatancy
[1]. A volume of solid grains is put in a rubber balloon and saturated with water. The
mouth of the balloon is connected to a test tube in which the level indicates the water
content of the balloon (see Fig. 7.6). When the balloon is deformed, we observe that
the water is actually sucked back into the balloon, while intuitively one would expect
the applied pressure to tend to squeeze water out. Since all the materials used here
are incompressible, this means that the overall volume of the balloon must increase.
As the volume actually occupied by the solid phase remains constant, this means
that the volume of the liquid phase in the balloon must have increased, whereas the
granular network continues to occupy the whole apparent volume. By deforming
the balloon, we have therefore expanded the granular network. Conversely, if the
interstitial fluid is not allowed to enter the granular network, it will be unable to
deform. This occurs for example with an airtight packet of coffee beans. As long as
the packet is hermetically sealed, it will be almost impossible to deform it without
breaking the beans or tearing the packet open. On the other hand, it deforms easily
as soon as a hole has been made in the packaging.
Dilatancy becomes all the more noticeable if the medium is dense in its initial
state. This effect thus arises in the deformation of granular materials in the friction
regime (see Sect. 7.5). A related effect occurs in the collision regime (see Sect. 7.6).
The normal force, which tends to dilate the material, increases with the shear rate.
7.3.3 Settling
Depending on how it was prepared, a given granular material may show varying
degrees of settlement in its initial state. For example, by causing the sample to
vibrate gently, it can be made to settle further since the particles will orientate or
displace themselves slightly in such a way as to fall into a locally more compact
arrangement, from which they will have greater difficulty in extracting themselves
during subsequent vibration. During such a process, the structure of the granular
medium becomes more and more jammed into place. This phenomenon can be illus-
trated by the following experiment. It will be relatively easy to push a rod into a
granular medium that has simply been poured into a container. If the system is then
shaken vigorously for a few moments, it will become very difficult to extract the rod
by simple traction, to the extent that even a force equal to the weight of the container
will not suffice. Here the vibrations induce settling that leads to greater jamming of
the particles.
7.3.4 State of the System
As explained above, the compactness of a granular material, in other words its solid
fraction, plays an important role in its behaviour. However, it would not suffice to
take into account only the concentration when characterising the state of such a
medium. Indeed, two parameters are essential when specifying the state. The first
is the solid fraction, which has a straightforward and precise definition: it is simply
the ratio of the volume of the solid particles to the total volume of the sample. The
7.3 Role of Configuration 241
second parameter is more difficult to define. It is the configuration of the particles,

which characterises their relative positions in space. For a given concentration, the
particles can be placed in different configurations which can produce very different
mechanical behaviour. For example, at a concentration that is not too close to ∂m ,
the particles can be piled up on top of each other or arranged so that they are all at a
small distance from one another, if they are suspended in a liquid of the same density.
In the first case, starting a flow would involve friction or jamming of the particles.
In the second case, only hydrodynamic forces would play any role and a flow would
be observed whatever the applied stress.
There is no simple way to characterise the particle configuration. In practice, to a
first approximation, one often considers the state of the material to be described by
the solid fraction alone. But as we have just seen, this will lead to some inaccuracy
for suspensions with concentrations not too close to the maximum packing fraction.
Indeed, for a given concentration, the particles could be arranged in rows, distributed
randomly, grouped into clusters, etc. On the other hand, when dealing with particles
at very high solid fractions, it is much more difficult to make a significant change in
the configuration without actually altering the concentration. This does not provide
any way to establish a direct connection but it justifies the use of a single parameter,
the solid fraction, to describe the system state to a first approximation.
7.4 Regimes of Behaviour
In order to identify the different kinds of behaviour of granular materials, it is impor-

tant to distinguish regimes in which one or other interaction tends to dominate. We
must therefore distinguish specific conditions under which certain of these inter-
actions prevail. For example, collisions between grains can only play a major role
at sufficiently high flow speeds. Saying that one specific interaction predominates
amounts to saying that the stresses in the material can be estimated by taking into
account only those particular interactions. To determine the regime to which a system
belongs, one can therefore compare the stresses (or dissipated energies) associated
with the different types of interaction. The problem with this approach is that the
energy dissipated in connection with each type of interaction is not very well deter-
mined, and in addition, it can depend on the configuration of the system, which itself
depends on the flow history. Under such conditions, in order to remain within a sim-
ple enough framework, we shall present below a somewhat approximate approach
which associates a single energy term with each type of interaction as determined
from the discussion of Sect. 7.2.
Concerning hydrodynamic interactions, we can estimate the energy required to
bring the two particles together from infinity and to within a minimum distance of
h = ξ. Assuming for simplicity that the viscous force remains at its maximal value
given by (7.1) and that the distance travelled is of the order of R, the hydrodynamic
energy is given by
μ0 θ̇ R 4
EH √ . (7.5)
ξ
For collisional interactions, we simply assume that the energy typically dissipated
during an elementary motion is a significant fraction of the kinetic energy transmitted
by the collision, which corresponds to the kinetic energy mV 2 of a particle when the
relative speed of one layer relative to the other is equal to R θ̇ :
E C = ϕS R 5 θ̇ 2 , (7.6)
where ϕS is the density of the particles. Regarding the frictional energy, it can be
estimated roughly by assuming that, during a unit deformation, two neighbouring
particles are displaced through a distance R, and that the resistance to motion is fN.
It follows that
E F = fNR . (7.7)
We may then associate dimensionless numbers with the ratios of the above energies:
EC ϕS Rξθ̇
St = = (Stokes number) , (7.8)
EH μ0
EC ϕS R 4 θ̇ 2
Ba = = (Bagnold number) , (7.9)
EF fN
EF f Nξ
Le = = (Leighton number) . (7.10)
EH μ0 R 3 θ̇
We can now identify the three basic regimes:

• Lubrication regime (E H dominates): St 1 and Le 1.
• Friction regime (E F dominates): Ba 1 and Le 1.
• Collision regime (E C dominates): St 1 and Ba 1.
This analysis could be refined by noting that the Leighton number as defined here
is in fact a product of two dimensionless numbers, i.e., a hydrodynamic number
N ξ/μ0 R 3 θ̇ describing the effect that normal forces can have when particles are
brought together in the presence of viscous friction, and the friction coefficient f
describing the impact of a normal force on the resistance to relative tangential dis-
placement of particles when they are in contact with one another. In the context of
our simplified analysis, we shall nevertheless stick with a description using just the
three dimensionless numbers listed above.
In the following, we shall describe the collisional and frictional regimes in detail
and then discuss transitions between regimes. Concerning the lubrication regime, this
is simply the concentrated regime for suspensions (see Sect. 3.4). The behaviour of
such a material can be considered to a first approximation as Newtonian with viscosity
7.4 Regimes of Behaviour 243
given by one of the usual empirical formulas, such as (3.19). Note, however, that in
this regime particle migration or changing configuration can have a very significant
effect on the observed behaviour of the material. Since these phenomena have not
yet been well understood, it remains a delicate matter to predict the behaviour of a
granular material in its lubrication regime from the physical characteristics of the
medium.
7.5 Frictional Regime
7.5.1 Simple Shear
In the frictional regime, the Bagnold number is rather low and the Leighton number
rather high. Such a regime thus obtains when the shear rate is low, and/or the particles
are very small, and/or the normal stresses are very high.
When a granular material is placed without special precautions in a rheometer
geometry, it is generally found that the measured shear stress is almost independent of
the enforced angular speed, provided that the latter is not too high. This suggests that
the frictional regime is not far off, since this kind of behaviour is reminiscent of two
grains in relative displacement. However, the measured shear stress depends on the
type and characteristics of the given geometry, a situation which arises because, under
these test conditions, the normal stresses that are effectively applied to the material
may be very different from one geometry to another. To characterise the frictional
regime, it is thus essential to control or measure the normal stresses. Although this
is possible in standard rheometer geometries, it does necessitate certain technical
adjustments.
The simplest way to measure directly the effect of normal stresses on the behaviour
of such a material is to place it in two boxes as shown in Fig. 7.7a, arranged with their
open sides placed opposite one another, and induce a relative translational motion of
the boxes (direct shear box). In this way, contact is maintained between the different
layers of the material, while preventing the material from escaping from the shear
geometry, and the normal load, i.e., the force provided to hold the two boxes together,
can be controlled. Using this device, measurements show that the tangential force
is still largely independent of the speed (over quite a broad range of speeds) and
proportional to the normal force, as in the case of two solids rubbing together. More
precisely, in terms of stresses, it is found that
π = λ tan Ω , (7.11)
where π is the shear stress and λ the normal stress, which are found by dividing the
tangential and normal forces, respectively, by the horizontal cross-sectional area of
the boxes. In (7.11), tan Ω is the coefficient of friction of the material, whence we
may define Ω as the internal friction angle of the material.
Fig. 7.7 Shear tests for a (a)

granular material. a Shear N
box where the two halves
of the box are translated
relative to one another. b
Shear box where the sides tilt
over, forcing the material to
deform. Shaded particles are
the ones actually moving to
any significant extent relative
to their neighbours during the
imposed deformation
(b) N
T
From the standpoint of fluid mechanics, this test is not entirely satisfactory since
the granular medium is more or less tightly jammed into each of the boxes, which
means that the shear is necessarily localised in the contact region between the two
boxes (see Fig. 7.7a). To remedy this problem, one can use a shear box in which the
side walls can tilt over when the upper face is translated, as shown in Fig. 7.7b. By
calculating the stresses in the same way as above from the normal and tangential
forces, it is found that the shear stress does indeed vary as described in (7.11), even
though the whole of the material is now sheared. This suggests that the localised
shear in Fig. 7.7a is now uniformly distributed over the whole height of the sample
and that, in this case, the behaviour of the material is the same in each layer as the
behaviour observed in the localisation zone.
Equation (7.11) is thus a very general law for describing the simple shear of a
granular material. Since we find the same behaviour on the scale of the grains them-
selves and on the macroscopic scale, it is natural to think that there must be a simple
relationship between the coefficient of friction f of the grains and the coefficient
tan Ω of the material as a whole. In fact, the only clear result, which was rather obvi-
ous anyway, is that tan Ω increases when we increase f , while there is actually no
simple and general relation between these two coefficients. The connection between
these coefficients depends on the detailed physical characteristics of the system, and
in particular the shape and roughness of the grains.
This lack of a simple relation between the two friction coefficients comes about
because the forces employed to shear a granular material are not just required to
7.5 Frictional Regime 245
Fig. 7.8 Structure of a granular material in the frictional regime. A continuous network of contacts
extends right through the sample, forming chains of forces of different magnitudes, represented
here by lines of different thickness joining the particles. During a low speed deformation (from left
to right), the initial network breaks, only to reform immediately in another form with the same
average characteristics
overcome friction between grains. For example, a granular medium comprising

grains with almost zero friction coefficient, or even strictly zero friction coefficient
in numerical simulations, may itself have a friction coefficient that is very different
from zero. This can happen because of effects that just do not exist in the case of fric-
tion between two solid surfaces, namely, steric effects. When a granular material is
sheared, the grains must be more or less constantly unjammed, and this involves col-
lective motions of varying degrees, inducing energy dissipation related to work done
against gravity (or the pressure force) to raise the grains slightly for short periods.
The diagram in Fig. 7.5 illustrates this effect on the local level. During shear flow, the
balls in the upper layer must regularly leave position (b) to transit through position
(a), and this requires a supply of energy to lift these balls out of their equilibrium
position (b).
7.5.2 Constitutive Law
Compared with the constitutive laws discussed up to now for different kinds of mate-
rial, the one in (7.11) has a rather peculiar form since it relates two stress components.
This is a feature specific to the granular medium in the frictional regime and it results
from the existence within the system of a network of direct contacts through which
forces can be transmitted. In the frictional regime, this network gradually deforms,
breaks locally, and then immediately reforms in a new configuration (see Fig. 7.8).
For the moment, (7.11) has only been established in the context of a simple shear.
While it was relatively straightforward to extrapolate the constitutive law obtained for
complex fluids to 3D situations using standard tensor analysis, it is not at all obvious
for a granular medium, starting out with an expression like (7.11) which relates
two components of the stress tensor. A first step here is to consider a slightly more
complex situation than simple shear, but one in which the stress field nevertheless
remains fairly simple and well controlled. This is the so-called triaxial test, in which a
given pressure λ3 is applied around the walls of a cylindrical segment of the material,
Fig. 7.9 Triaxial test. A

uniform pressure λ3 is σ1
imposed around the walls
of a cylindrical sample, often
using a fluid, while an axial
pressure λ1 is also applied. If
the pressure difference is great θc Δh
enough, the sample height
changes from h to h − βh. For σ3 σ3
an initially dense sample, the h
deformation may be localised
over a surface making an angle
νc with the vertical
while at the same time a normal stress λ1 is imposed along the cylinder axis, as shown
in Fig. 7.9. When λ1 = λ3 , the system is in principle in equilibrium. However, when
λ1 > λ3 , significant deformations are sometimes observed. To study the behaviour of
the material, the deformation βh/ h of the segment is monitored for different values
of the difference q = λ1 − λ3 between the stresses in the two directions. Note that
the average pressure on the material is p = (λ1 + 2λ3 )/3.
When q is gradually increased, the deformation increases, slowly at first, and more
or less in proportion to q. It then increases significantly when q reaches a plateau
corresponding to a critical value (see Fig. 7.10). In fact, the way the system reaches
this plateau depends on the initial state of the system. For a sample that starts out
in a loose state, the system settles and the curve tends very gradually toward the
stress plateau. For a sample that starts out in a dense state, the system expands and
the curve passes through a maximum value of q (at a deformation of the order of
10 % for sands), before dropping back down to the stress plateau. In practice, it is
found that the state reached on the plateau, known as the critical state, is practically
the same whatever the initial state. If the material is initially in the critical state, an
intermediate curve is obtained, and during the deformation, the material does not
undergo any significant change in volume.
According to what we have just seen, the stress state and volume of the sample
each tend toward certain values, and these values define what is known as the critical
state of the material, independently of the deformation history. On the scale of the
grains, the existence of an asymptotic state at large deformations can be explained in
the following way. The greater the deformation, the more likely the contact network
is to break and reform in a new configuration, and this will gradually redistribute the
network of contacts in a way that is quite independent of the history of the material.
Note that this critical state is not always achieved by following homogeneous load
paths. For dense samples, the response of the material goes through a maximum
which often leads to fracture. This instability manifests itself through the appearance
σ1 − σ3
Dense
Critical
state
Loose
Δh
h
Fig. 7.10 Dependence of the pressure difference in a triaxial test on the deformation observed for
different values of the initial sample density
of a shear zone that is tilted relative to the cylinder axis and which in fact concentrates
the whole deformation (see Fig. 7.9).
A certain number of criteria, that is, relations between the stresses, e.g., Coulomb,
Drücker–Prager, etc., have been put forward to predict when the critical state will
actually be reached. Here we shall consider only the Coulomb criterion for a 2D
flow because it is relatively simple and gives a good description of what is actually
observed to a first approximation in various situations. This criterion states that the
critical state will be reached if there is a plane surface within the material over which
the magnitude of the shear stress π is proportional to the magnitude of the normal
stress λ according to (7.11).
In order to apply the Coulomb criterion in an arbitrary flow scenario, one must
determine the stresses over each plane surface within the material. In a triaxial test,
for a plane surface element with unit normal n making an angle ν with the z direction,
it can be shown from the full expression for the stress tensor (see Sect. A.4) that
λ1 + λ3 λ1 − λ3 λ1 − λ3
λ = + cos 2ν, π =− sin 2ν . (7.12)
2 2 2
This means that, for fixed applied stresses during the test, the point withcoordinates
(λ, π ) describes a circle of radius (λ1 −λ3 )/2 and centre (λ1 +λ3 )/2, 0 . The radius
of the circle, known as Mohr’s circle, thus increases with the stress. For a critical
value of q, the circle ends up being tangent to the Coulomb line defined by (7.11),
i.e., π = λ tan Ω, and the critical state is reached. We then have
λ1 1 + sin Ω
= ,
λ3 1 − sin Ω
or again
q − Mp = 0 , (7.13)
with
Ω
M = 6 sin − sin Ω .
3
This last version of the Coulomb criterion has the advantage that it constitutes an
objective formulation, in the sense that it is independent of the reference frame
used for observation, because it only involves invariants of the stress tensor. Note
that it is only valid for 2D flows, or flows with an axis of symmetry. Provided that
the critical state has not been reached (q − M p < 0), we remain in the region of
small deformations. The granular material then behaves as an elastoplastic solid. The
stress state at a given time is almost independent of the strain rate, but it does depend
on the whole history of the deformation. Once the critical state has been reached
(q − M p = 0), we enter the regime of large deformations which marks the onset
of flow. This transition is reminiscent of the yield stress transition for a yield stress
fluid. However, there are some important differences between the two scenarios. In
particular, the critical state for a granular suspension depends on the average pressure,
while the yield stress depends solely on the deviatoric stress tensor. Moreover, it is
possible in some cases, in particular with dense samples, to go beyond the critical
state without fluidisation of the material, at which point we have q − M p > 0.
7.5.3 Applications to Quasi-Static Flows
For slow enough granular flows, i.e., in the frictional regime, certain features of the
stress field can be described if the slip surfaces for the Coulomb criterion are easily
identifiable. This is the case for slow flow in a duct and for the maximal inclination
of a pile of granular material with free surfaces.
7.5.3.1 Flow in a Vertical Duct
Here we consider the slow flow of a granular suspension in a vertical tube of radius
R and with rough walls. By considering the momentum balance (see Chap. 8), it can
be shown that the shear stresses are maximal at the walls. We would thus expect
the Coulomb criterion to apply over the walls. Consider now a cylindrical segment
of the material of thickness dz, where the Oz axis, directed downwards, coincides
with the axis of symmetry of the tube (see Fig. 7.11). Along the Oz axis, this slice is
subjected to its own weight ϕgπ R 2 dz, the force
Fig. 7.11 Stress distribution

for a granular flow in a hopper. 0 σ(z)
Right dependence of the
normal stress in the material σ(z)
on the distance from the free σT tanϕ z
surface
σT
dz
σ(z+dz)

π R 2 λ (z) − λ (z + dz)
due to the normal stress within the granular medium, denoted by λ here, on the
upper and lower facets of the portion of cylinder under consideration, and also to the
friction forces over the inner wall of the tube. If this friction obeys Coulomb’s law,
the corresponding force is −2π RλT tan Ω dz, where λT is the normal stress within
the granular medium as evaluated over the wall of the tube. Assuming that, in the
frictional regime, λT is proportional to the normal force within the suspension in the
z direction, we then have λT = kλ . Under these conditions, the balance of forces
along the vertical axis in the steady-state regime is given by
dλ
− π R2 − 2π Rkλ tan Ω + ϕgπ R 2 = 0 . (7.14)
dz
With the boundary condition λ (0) = 0 (taking the origin of the z axis at the free
surface), this equation has solution

ϕg R 2k tan Ω
λ = 1 − exp − z . (7.15)
2k tan Ω R
At small depths z R/2k tan Ω, the stress thus varies linearly with the depth accord-
ing to λ = ϕgz. On the other hand, at greater depths z R/2k tan Ω, the stress is
independent of the depth, with λ √ ϕg R/2k tan Ω.
This is very different behaviour to an ordinary fluid. In the latter, the pressure
increases continually with the depth. The force at the hopper outlet is thus propor-
tional to the total height of fluid, and the flow rate increases with increasing level
of liquid above the outlet. The above result for a granular medium in the frictional
regime, originally obtained by Janssen [2], reflects the peculiar features of force dis-
tributions within a granular pile inside a tube. In particular, we have here an example
of the phenomenon known as screening. Due to friction, the particles are in part
held up by the side walls, and through this they are able to some extent to support
the particles above them, thereby reducing the force applied at the bottom of the
container.
This can be used to deduce some of the characteristics of the flow rate Q of a
granular suspension through a hopper. Since beyond some sufficient depth of material
in the hopper the pressure near the outlet does not depend on the depth, it is just as
though the particles were in free fall (independently of the mass of material situated
above them) outside the hopper from a height H that depends on the exact shape of
the hopper. This height is proportional to a characteristic length for the hopper and
some function of its shape. For example, for a cylindrical hopper of diameter D0 and
⇒ H = D f (D0 /D). As the characteristic speed of free fall
outlet diameter D, we have
under gravity is V = g D, we expect the flow rate Q = VD2 to be proportional
to (g D 5 )1/2 , with a constant of proportionality that depends on the shape of the
hopper and possibly also on the coefficient of friction of the grains. Experimentally,
we observe a relation of the form

Q = χ g(D − γ D0 )5 , (7.16)
known as the Beverloo equation, where χ and γ are two constants whose connection
with the coefficient of friction of the particles has not yet been clearly established.
7.5.3.2 Shape of a Heap of Granular Material with Free Surface
When a granular material is gradually poured from above a fixed point on a horizontal
solid surface, it forms a heap whose shape and dimensions depend on the flow history
and characteristics of the material. In particular, heaps can be obtained with free
surfaces inclined at different angles to the solid surface, but we find that these angles
can never exceed some critical angle, the maximum rest angle Ωrest . When we attempt
to produce a heap with greater angle, the material simply flows until the angle drops
back to Ωrest .
In fact the angle Ωrest is simply the internal friction angle Ω of the material. To see
this, consider a pile of material whose free surfaces are two infinite planes parallel to
some given direction (so that the problem is two-dimensional) and making an angle
ν to the horizontal (see Fig. 7.12). When the condition for instability has just been
reached, the material is expected to flow over the surface for which the Coulomb
criterion is satisfied. Suppose this is a plane surface AB making an angle i with the
horizontal. When there are no inertial effects, the balance of forces on the volume
of material situated above the plane AB is expressed by 0 = P + t, where P is the
weight of the material and t the average stress over the plane AB. These two stresses
can each be decomposed into the sum of a component parallel to the surface AB and
a component perpendicular to it, viz.,
B
t x
y
P
A i
Fig. 7.12 Heap of granular material poured from above a fixed point on a solid horizontal plane
P = (P sin i)x + (P cos i)y, t = T x + Ny .
The balance of forces implies in particular that T /N = tan i. But we know that the
Coulomb criterion is satisfied at least at the beginning of the motion across the plane
AB, so T /N = tan Ω. It follows that i = Ω. Since we must have i < ν , there can
be a volume of material within the pile that tends to slip over the surface AB only if
the angle ν of the pile is greater than Ω. Otherwise the pile remains at rest. It follows
that Ωrest = Ω. This result generalises to a heap of conical shape. The maximum rest
angle is the internal friction angle of the granular material.
Note that, in this context, the height of material plays no role. Only the angle
between the side of the pile and the horizontal is important. This is in fact valid
only when the thickness of the material is much greater than the size of the grains,
otherwise the discrete nature of the material may come into its own. Note also that
the effective rest angle depends to some extent on the way the pile was produced.
If formed gradually and carefully, the pile can reach a maximum rest angle Ωmax
slightly greater than Ω. If the pile is now slightly disturbed or a few more grains
are added suddenly at the top when an angle close to Ω has already been reached,
avalanches of grains will occur at the surface of the pile. When these avalanches have
stopped, usually at the foot of the pile, the slope of the sides will be slightly less than
Ω. Finally, depending on the preparation conditions, the angle of a heap of granular
material on a horizontal plane will vary between two values Ωmin and Ωmax which are
characteristic of the material, and Ω will lie somewhere between these two values.
7.6 Collisional Regime
By definition, in this regime, only collision-type interactions are considered impor-

tant. Since the time of contact during a collision is generally very short compared
with the characteristic time 1/θ̇ associated with relative motions of the particles, a
snapshot of the system would for the main part show the particles separated from
one another, as in a gas (see Fig. 7.13). However, there are two key differences with a
gas. In a granular medium (i) the collisions are not purely elastic and (ii) the particle
concentration is relatively high.
Fig. 7.13 Transition from the

frictional to the collisional μ
regime as viewed on the
level of the internal structure,
showing the dependence of
the friction coefficient on the
Bagnold number
Frictional Collisional
regime regime
Ba
A first approach to understanding the behaviour of a granular medium in the col-

lisional regime assumes that, under a simple shear, the layers of particles simply
slip over one another as shown in Fig. 7.5. The shear stress comes about through
momentum transmitted during collisions between particles, as described by the rela-
tion π = (1/A)βmV /βt (see Sect. 2.3.5). This expression is proportional to the
product of a factor βmV , the average momentum transmitted per collision, and a
factor 1/βt, the frequency of collisions per particle. The momentum transmitted is
proportional to the difference between the momenta of the two particles, i.e., χm R θ̇ ,
where χ depends on the coefficient of restitution, the shape of the particles, and the
collision angle. If the concentration is constant and the material remains disordered,
it is reasonable to assume that χ is independent of θ̇ . The collision frequency depends
for its part on the internal agitation of the system. Indeed, the more agitated the par-
ticles, the more collisions there will be per unit time. One can then distinguish two
asymptotic regimes depending on whether the particles are not particularly agitated,
in which case they are basically following the motion imposed on the macroscopic
level, or highly agitated, in which case the fluctuations in their motion about the
average motion will play a crucial role.
The first asymptotic regime corresponds to a low level of agitation or a concen-
tration close to Ωm . In this case, the particles come into contact essentially through
the average shearing of the material. A collision will thus occur each time two parti-
cles become neighbours for a few moments through the imposed macroscopic flow.
Under such conditions, the collision frequency is then simply θ̇ and the shear stress
is given by
χm 2
π= θ̇ . (7.17)
R
7.6 Collisional Regime 253
Furthermore, during each collision, a component of the momentum (γm R θ̇ , where

γ is a coefficient depending on the same quantities as χ) is also transmitted in the
direction perpendicular to the average motion. As a consequence, a normal stress
is produced in the direction perpendicular to the flow, proportional to the tangential
stress:
γm 2 γ
λ = θ̇ = π . (7.18)
R χ
It should be borne in mind that this model merely provides a rough idea of the form
of the constitutive law in the collisional regime and for a particular flow context,
viz., collisional regime with a low level of agitation, and on the basis of several
simplifying assumptions, viz., binary elastic collisions.
The second asymptotic regime corresponds to the situation where the particles are
highly agitated, to the point where collisions can now occur between neighbouring
particles at a very different frequency to θ̇ . This can be investigated theoretically
using the same tools and methods as for the kinetic theory of gases (see Chap. 2). We
thus find that the shear stress has the form

λ ≈ Tg θ̇ , (7.19)
where Tg is the so called granular temperature, defined as the mean squared grain
speed fluctuation: Tg = ◦u∝2 . This expression is formally identical to the one found
for perfect gases in (2.21).
The main difference between a gas and a granular material lies in the origin of
the temperature. For a gas, it is of thermal origin. Simply putting a gas in contact
with a surrounding medium will provide it with the agitation energy associated
with that temperature. For a granular material, when there are no effects of thermal
agitation on the dynamics of the particles, only the flow can supply it with this
internal agitation energy. On the basis of energy considerations (first and second
laws of thermodynamics), the kinetic theory of granular media provides expressions
for the granular temperature as a function of the system variables. The constitutive
law is then nonlocal in a certain sense, since one must fully solve the equations of
conservation of momentum and energy in order to find the stresses or the viscosity.
Under simple shear, the temperature is generally found to be proportional to θ̇ 2 . It
follows that the shear stress given by (7.19) is still proportional to θ̇ 2 . This pattern,
i.e., the fact that the stress should be proportional to the square of the shear rate,
which appears in both asymptotic flow regimes, seems therefore to be the signature
of the behaviour of a granular material in the ‘purely’ collisional regime.
7.7 Intermediate Regimes
7.7.1 Transition from Frictional to Collisional Regime
In this section we consider systems which begin in a frictional regime but in which
collisional effects gradually begin to play an increasing role, while lubricated contacts
remain negligible. This situation corresponds to a Leighton number that remains high
(dominant frictional contacts) and the Bagnold number gradually increases. This
happens for instance in dry granular media with increasing flow speeds.
In this context, the most natural approach is to use the tools set up to describe
behaviour in the frictional regime, in particular the friction coefficient, and consider
the dependence of this coefficient on the Bagnold number. Indeed, numerical simula-
tions [3] confirmed by experiments show that the behaviour of these systems can be
described over a range of Bagnold numbers up to Ba √ 10−2 by a friction coefficient
of the form
⇒
tan Ω = tan Ω0 + k Ba, (7.20)
where k is a parameter depending on the material. At the same, it turns out that
the compactness of the system,
⇒ i.e., its apparent density, decreases with the speed
according to τ = τ0 − k ∝ Ba, whence the system expands as the speed increases.
During this transition, the fraction of prolonged contacts between particles gradually
falls. So at low speeds there is a network of chains of forces, but as the speed increases,
this network disappears (see Fig. 7.13). Note that, for higher values of the Bagnold
number, the model in (7.20) is no longer valid. In fact, tan Ω then tends to a constant
value.
7.7.2 Transition from Frictional to Lubricated Regime
Here we consider the behaviour of a granular material that starts in the frictional
regime and evolves gradually to a lubricated regime. This corresponds to a situation in
which, starting from the frictional regime the Leighton number gradually decreases.
This situation is typical when the flow speed of a compact granular suspension is
increased.
7.7.2.1 General Expression for the Stress Tensor
The regimes considered here are such that the hydrodynamic and frictional interac-
tions can play a significant role in the constitutive law. In this context, it is useful to
note that, over an arbitrary plane surface element, the stress can be calculated from
the sum of hydrodynamic and frictional interactions (see Appendix A). This result
7.7 Intermediate Regimes 255
can be extrapolated to the stress tensor, which can thus be decomposed into a sum of
one tensor associated with hydrodynamic interactions alone and another associated
with frictional interactions alone, viz.,
= f (Ω) + g . (7.21)
This result was established in the field of soil mechanics (Terzaghi, 1943) to describe
the behaviour of heaps of granular materials saturated by some liquid, but also in fluid
mechanics [4] to describe the behaviour of highly concentrated suspensions. Note
in passing that this illustrates the fact that a granular material bathed in a liquid can
indeed be considered, depending on the circumstances, as a concentrated suspension
or as a granular heap.
To a first approximation, we may neglect the consequences of contacts between
grains on hydrodynamic effects within a concentrated suspension. The tensor f (Ω)
is the stress tensor in a concentrated suspension of solid particles and therefore
depends essentially on the particle concentration (see Chap. 3). In contrast, the term
g can vary significantly depending on the particle configuration and in particular
depending on the number of contacts between particles, which itself depends on the
flow history of the system as a whole and the initial state of the material. We would
thus expect that, at low speeds Le 1, the contacts between particles would be able
to hold out, leading to a situation in which g is dominant. At higher speeds, hydro-
dynamic repulsion terms tend to repel the particles, whereupon friction eventually
becomes negligible and the corresponding stress is negligible compared with that of
hydrodynamic origin. It is as though the material has fluidised.
7.7.2.2 Behaviour for High Leighton Numbers
When the interstitial fluid is Newtonian, the first stress term in (7.21) tends to − pI
and the constitutive law is therefore = − pI + g . The system behaves in a
qualitatively identical way to a granular material in the purely frictional regime.
However, in most cases, the pressure of the interstitial fluid is not uniform, so
the friction coefficient of a granular paste differs from that of the corresponding
dry granular material. The physical explanation for this phenomenon is as follows.
Consider two grains of volume η placed next to one another. The normal force
applied to the upper grain is equal to the weight ϕS ηg of the grain, and the minimal
tangential force able to shift the upper grain sideways with respect to the other
is fϕS ηg. If the two particles are now immersed in a liquid of density ϕL , the
apparent weight of the upper particle is reduced by the buoyancy ϕL ηg due to
Archimedes’ principle. We assume that each particle is immersed in a volume of
liquid η ∇ = η(1 − ∂)/∂, where ∂ is the particle concentration. The tangential force
required to induce a relative motion of the grains is then proportional to the normal
force T = f (ϕS − ϕL )ηg effectively applied to the grain. For very low speeds,
the flow of the interstitial liquid induces negligible forces. T is therefore the force
required to shear the whole system of two grains and surrounding liquid. However,
the effective normal force between the two parts of the system, i.e., between the two
particles and the liquid, is here the sum N = ϕS ηg + ϕL η ∇ g of the weight of the
grain and the weight of the liquid. Consequently, the effective friction coefficient of
the system, equal to the ratio of the tangential force and the effective normal force,
is given by
f (ϕS − ϕL )
,
ϕS + ϕL η ∇ /η
and this is less than f . The above result can be generalised to granular material,
giving a friction coefficient of the form
T ϕS − ϕL
tan Ω ∇ = = tan Ω , (7.22)
N ϕS + ϕL (1 − ∂)/∂
which is less than the friction coefficient of the same granular material when dry.
Suppose now that the liquid flows vertically at average speed V around the two
grains, from the lower grain toward the upper one. The apparent weight of the upper
grain will then be reduced by the hydrodynamic drag force exerted on the grain due
to the liquid flow, viz., k LμV . The tangential force required to shear the system will
be reduced in proportion:

T = f (ϕS − ϕL )ηg − k LμV . (7.23)
Since we also know that the normal force associated with the weight of the liquid
and particles does not vary, the effective friction coefficient of the system is therefore
lower than for the liquid at rest. It may even vanish if the flow speed of the liquid
is high enough. For this to happen, it would just have to exceed the critical value
Vc = βϕηg/k Lμ. A more standard approach to the problem, considering the gran-
ular material as a whole and using Darcy’s law, which describes the flow of a fluid
through a porous medium (here the grain network) to estimate the pressure gradient
from one layer of grains to another, leads to qualitatively similar conclusions.
As a consequence, water rising at low speeds, of the order of a few cm/s, can
threaten the equilibrium of a granular ground material, making the stresses within
its solid backbone zero or negative. The material then loses all consistency. This is
what happens in some quicksands and also in the phenomenon of hydraulic uplift at
the foot of earth-built dams. In the latter case, water circulates through the base of
the dam and gradually removes the solid materials situated downstream of the dam.
When the interstitial fluid is air, there may still be significant interstitial pressure
effects despite the low density of air. For example, for very fine sand (grain sizes
less than 0.3 mm), the flow in an hourglass can be significantly perturbed, becoming
intermittent, or even stopping altogether. In contrast, if a very fine sand is injected
into a vertical tube, it is found that the flow rate increases linearly with the length of
the tube. In both cases, it is the air pressure difference on either side of the constriction
or endpieces of the tube which controls the flow.
7.7 Intermediate Regimes 257
This phenomenon is also used in process engineering in the context of fluidised

beds. A fluid is injected into a granular medium to hold the grains in suspension,
thereby allowing chemical reactions to occur.
7.7.2.3 Behaviour Transition for Small Leighton Numbers
In the transition from a frictional regime to a lubricated regime, the continuous

network of contacts gradually disappears, and with it the frictional interactions. At
high Leighton numbers, the shear stress is proportional to the normal stress, while
at low Leighton numbers, it is simply proportional to the shear rate. An approx-
imate match between these two limiting situations leads to the following general
expression:
π = λ tan Ω0 + μ(∂)θ̇ , (7.24)
which can be reexpressed in the form

π
tan Ω = = tan Ω0 + χ/Le .
λ
The first term on the right-hand side dominates for high values of Le, while the
second dominates for low values of Le.
This is roughly what is observed for example in an experiment using a Couette
geometry, with a granular suspension made by piling up grains in a less dense liquid.
At low apparent shear rates, the shear stress levels off (see Fig. 7.14) at a plateau that
rises with the height of material in the geometry. Since the normal stress within the
material is likely to rise with the height of material, the above result is compatible
with an interpretation of the behaviour in terms of a Coulomb model. When the shear
rate is increased beyond a certain value, the shear stress begins to increase linearly
and is independent of the height of material, as predicted in the lubrication regime.
These characteristics conceal a more complex reality. Indeed, even though such
a material may appear to behave like a yield stress fluid as suggested by (7.24), the
material rarely remains homogeneous. The particles tend to migrate from the most
strongly sheared regions towards ones that are less sheared. A particle concentration
gradient therefore builds up in the material [5], associated with what may be a sig-
nificant non-uniformity in the shear rate. This means that the true behaviour of the
material at a given particle concentration may differ significantly from its apparent
behaviour, which assumes an average concentration and does not take into account
heterogeneities in the shear rate throughout the material.
Fig. 7.14 Transition from

the friction to the lubrication μ
regime from the point of view
of internal structure, showing
the dependence of the effective
friction coefficient on the
Leighton number
Frictional Lubricated
regime regime
Le -1
7.7.2.4 Behaviour Transition for a Yield Stress Interstitial Fluid
When the interstitial fluid is a yield stress fluid, the tensor f (∂) must be replaced
by an expression of the kind proposed by the Herschel–Bulkley model. Here we
assume that the components of this fluid are much smaller than the particles of the
granular phase. At low flow speeds, since the term depending on the shear rate in the
constitutive law of the yield stress fluid is negligible, we have
⇒
f √ − pI + πc (∂)D/ −D ,
leading to a yield stress of the form
πc (∂) + λ tan Ω . (7.25)
The functional dependence πc (∂) of the yield stress of a paste containing grains was
discussed in Chap. 3.
References
1. Reynolds, O.: On the dilatancy of media composed of rigid particles in contact. Philos. Mag.
20, 469–481 (1885)
2. Janssen, H.A.: Versuche über Getreidedruck in Silozellen. Zeitschr. d. Vereines deutscher Inge-
nieure 39, 1045–1049 (1895)
3. da Cruz, F., et al.: Rheophysics of dense granular materials: Discrete simulation of plane shear
flows. Phys. Rev. E 72, 021309 (2005)
4. Batchelor, G.K.: The stress distribution in a suspension of force free particles. J. Fluid Mech.
41, 545–570 (1970)
References 259
5. Ovarlez, G., Bertrand, F., Rodts, S.: Local determination of the constitutive law of a dense
suspension of noncolloidal particles through magnetic resonance imaging. J. Rheol. 50, 259–
292 (2006)
Further Reading
6. Andreotti, B., Forterre, Y., Pouliquen, O.: Granular Media—Between Fluid and Solid. Cam-
bridge University Press, Cambridge (2013)
7. Bideau, D., Hansen, A. (eds.): Disorder and Granular Media. North-Holland, Amsterdam
(1993)
8. Bowden, F.P., Tabor, D.: Friction et Lubrification. Dunod, Paris (1959)
9. Brown, R.L., Richards, J.C.: Principles of Powder Mechanics. Pergamon Press, Oxford (1966)
10. Coussot, P.: Rheometry of Pastes, Suspensions and Granular Materials. Wiley, New York (2005)
11. Duran, J.: Sands, Powders and Grains—An Introduction to the Physics of Granular Media.
Springer, New York (1999)
12. Johnson, K.L.: Contact Mechanics. Cambridge University Press, Cambridge (1985)
13. Nedderman, R.M.: Statics and Kinematics of Granular Materials. Cambridge University Press,
Cambridge (1992)
14. Schofield, M.A., Wroth, C.P.: Critical State of Soil Mechanics. McGraw-Hill, London (1968)
15. Terzaghi, K.: Theoretical Soil Mechanics. Wiley, New York (1943)
Chapter 8
Rheometry
Abstract The aim of rheometry is to measure the rheological properties of materials,

that is, the relationship between stresses within the material and deformations it has
undergone. Commonly used techniques of rheometry use rather simple flows (mainly
simple shear), partly so that the constitutive law of the material under the given
conditions is expressed by a relation between a small number of variables, and partly
so that these variables can be measured directly by macroscopic measurements,
in other words, without having to make local measurements within the material.
We exhibit the calculations for the various simple shear geometries, along with the
experimental procedures for studying the behaviour of the main kinds of complex
fluid. In practice, many experimental problems can arise to complicate interpretation
of the measurements, such as sedimentation, evaporation, slip on the walls, inertial
effects, the appearance of shear bands, and so on.
8.1 Introduction
The aim of rheometry is to measure the rheological properties of materials, in other

words (see Appendix A) the relations between the stresses within the material and the
history of deformations it has undergone. However, taken literally, such a task would
of course be impossible, because in practice there is no way one could monitor
the deformation history at every point of a fluid and measure all the components
of the stress tensor. The standard techniques of rheometry make use of simple flow
scenarios, i.e., simple enough to ensure that the constitutive law of the material under
such conditions can be written in the form of a relation between a small number of
variables, and simple enough to allow direct measurement of these variables at a
‘macroscopic’ level, in other words without the need for local measurements inside
the material. The main type of flow that meets these requirements is simple shear.
In this case, the layers of material tend to slip along parallel to one another (see
Fig. 1.5) under the action of a shear stress τ which induces a shear rate ξ̇ . Different

DOI: 10.1007/978-3-319-06148-1_8,
262 8 Rheometry
geometries can implement such a flow, including cone–plate, parallel plates, coaxial
cylinders, ducts, and inclined plates. In the first part of this chapter (see Sect. 8.2), we
review the basic calculations used in these geometries to deduce the constitutive law
of a material under simple shear from the associated macroscopic measurements.
These calculations involve several assumptions:
• The material remains homogeneous.
• Its volume remains constant.
• The sample occupies the geometry in the way assumed by the theory.
• The shear is uniform in certain directions.
In practice, many experimental difficulties can arise to undermine the above hypothe-
ses and compromise interpretation of the measurements. Examples are the sedimen-
tation or creaming of elements initially suspended in the mixture, evaporation of the
interstitial liquid, slipping at the walls, inertial phenomena, the development of shear
bands, and others. The second part of this chapter (see Sect. 8.3) will review the main
phenomena that may perturb the measurement process.
Establishing a measurement protocol is a delicate matter. It must do everything
to prevent the appearance of perturbing factors and it must be adjusted to suit the
specific features of the behaviour and material under investigation. Such a protocol
is developed by first examining the way the material reacts to various kinds of stress.
One must then analyse this data making use of a clear corpus of established knowl-
edge, but appealing also to accumulated knowhow and even intuition to some extent.
In this sense, rheometry can be considered something of an art, and especially for
fluids such as yield stress fluids whose peculiarities distinguish them so sharply from
simple liquids. We shall review a certain number of standard procedures and discuss
their main advantages, but also the precautions that need to be taken when they are
employed.
For yield stress fluids, a range of practical techniques have also been developed
in industry for estimating the yield stress without the need for a rheometer. Some of
these techniques will be presented in the last part of the chapter.
8.2 Basic Geometries
The simplest geometry able to produce simple shear is obtained directly from the
definition of this type of flow, i.e., it comprises two parallel plates in relative motion
along a direction lying in the plane of the plates (see Fig. 2.7). The distance between
the solid surfaces in relative motion is called the gap. This geometry is used in
geotechnics (see Fig. 7.7). In fact it can only be used for tests during which the
maximal deformations are rather small, and in particular, solids or yield stress fluids
in their solid regime. If the material is a liquid, then under the action of a given
force, the relative motion will proceed until the two plates are no longer opposite
one another (see Fig. 2.7), and interpretation of the measurements in terms of the
8.2 Basic Geometries 263
Fig. 8.1 Parallel disk

geometry
Solid
disk
Material H
0
r R
constitutive law of the material is based on calculation hypotheses which are no

longer relevant. To study complex fluids, it is therefore preferable to use systems that
allow the relatively moving layers to remain in contact with one another.
8.2.1 Parallel Disk Rheometer
The most natural idea for keeping the layers in contact even when the material suffers
significant deformations is to have them rotate about an axis through their centre.
This is the basic construction of the parallel disk geometry, which consists of two
solid disks of radius R, a distance H apart, and rotating at different angular speeds
about their common axis (see Fig. 8.1). Here we shall assume for simplicity that the
lower disk is fixed. If the angular speed of the upper disk is α , a point on this disk at
distance r from the axis will rotate about the axis with a tangential speed equal to α r.
Material elements along the straight line between this point and the same point on the
lower disk will thus be subjected to a simple shear between the two solid surfaces,
with shear rate ξ̇ = α r/H. Likewise for all the elements located at the same distance
r from the axis. Hence, with this geometry, the shear rate varies from 0 at the central
axis up to a maximal value αR/H at the outer edge. Because of this shear, at a height
h above the lower disk, the material elements will be moving at speed α rh/H. The
layer of material at height h is thus rotating at angular speed α h/H and we do indeed
obtain a relative slipping of the fluid disks around the central axis.
This shear geometry is very simple to set up and use, especially as regards insertion
of the sample and cleaning, but the relationship between macroscopic measurements
and the constitutive law of the material is somewhat complex. We noted above that
the shear rate was not uniform throughout the gap. Naturally, the same goes for the
shear stress τ associated with resistance to the relative slipping of the layers which,
owing to its dependence on the shear rate, will vary with the distance r. In practice
264 8 Rheometry
the resistance to shear is measured through the total torque M that must be applied
to the upper disk to impose the given rotation. This torque is found by summing the
force elements τ (r)2β rdr over circular bands of width dr, each multiplied by its
distance r from the axis, whence
R
M= 2β r 2 τ (r)dr . (8.1)
0
To a first approximation, we may assume that the average shear rate is close to the
maximum shear rate obtained at the outer edge of the sample:
αR
ξ̇R = , (8.2)
H
and also that the stress is constant across the gap, whence (8.1) implies that
3M
τ= . (8.3)
2βR3
These approximations generally provide quite reliable results as far as the flow curve
is concerned owing to the fact that a significant part of the torque results from
forces applied at the outer edge of the fluid. Indeed, in this region, both the area
2β rdr of the band and the distance r at which the force is applied increases in
proportion to r. It follows that the ‘average’ stress specified in (8.3) is close to
the value of the stress at the outer edge. For example, for a Newtonian fluid, the
local stress is τ (r) = μξ̇ = μα r/H, and integration of (8.1) shows that the stress
τR = μξ̇R = 2M/β R3 at the outer edge is equal to just 4/3 of the ‘average’ value in
(8.3), whereas a torque distribution proportional to the distance would give a factor
of 2.
The pair of variables given by (8.2) and (8.3) can thus provide a first approximation
to the shear stress and shear rate at the outer edge of the sample, and these values
can be deduced directly from the measured values of the torque and angular speed.
The exact values of these variables can also be deduced by making full use of (8.1)
and varying α or M. To do this we make the change of variable r ◦ ξ̇ = α r/H,
which leads to
ξ̇R
2β
M= ξ̇ 2 τ (ξ̇ )dξ̇ .
(α /H)3
0
Differentiating this equation with respect to ξ̇R = α R/H, we obtain

M α ∂M
τR = τ (ξ̇R ) = 3+ . (8.4)
2β R2 M ∂α
The value of ∂M/∂α must be measured by plotting the measurement curves M(α ).
For each value of α , we can then calculate ξ̇R and the associated value of τR . Note
that this analysis assumes a permanent flow scenario, i.e., that the state of the material
is fully specified by the stress or the local shear rate. Finally, we obtain the exact
relation τR = f (ξ̇R ), giving the flow curve of the material.
8.2.2 Cone–Plate Rheometer
This arrangement looks very like the parallel disk geometry except that the upper
disk is replaced by a cone, slightly truncated to avoid frictional contact between the
two solid surfaces, but whose removed apex would otherwise have fallen precisely
on the lower disk. The angle θ between the cone and the horizontal is very small,
being at most a few degrees. This geometry has the great advantage that it exactly
compensates for the dependence of the shear rate on the radial distance which occurs
in the parallel disk setup. Indeed, like the relative speed of the two solid surfaces, the
distance between the two solid surfaces also varies linearly with the distance from
the central axis, since H ≈ rθ . In this arrangement, the shear rate is thus ξ̇ = α /θ,
which is independent of r. Under these conditions, the shear stress does not depend on
the distance and the integral in (8.1) can be carried out directly, giving the expression
for the shear stress in (8.3), which is in principle uniform throughout the sample.
This setup has the great advantage that one can obtain the effective and instan-
taneous values of τ and ξ̇ directly by measuring M and α . In particular, transient
phenomena involving temporal changes in τ or ξ̇ can be monitored and measured
directly. On the other hand, it is less convenient for loading the sample. The dis-
tance between the cone and plate must be very accurately controlled and the material
must not contain any suspended particles that might get trapped in the gap, which is
necessarily very narrow, especially close to the central axis.
8.2.3 Concentric Cylinder Rheometer
Also known as the Couette rheometer, in honour of one of its main advocates, this
arrangement comprises two coaxial cylinders of radii R1 < R2 and height h. The
material is loaded between the two cylinders which are then rotated at different
angular speeds (see Fig. 8.2). The material is thus sheared by the relative motion of
the cylinders about their central axis. We shall assume here that the inner cylinder
rotates at angular speed α , while the outer cylinder remains fixed. In this form of
simple shear, we now have cylindrical fluid layers slipping over one another.
The shear stress τ (r) represents the friction between two layers located on either
side of the distance r. The torque M on the inner cylinder is the force due to friction
at distance r, i.e., 2β rhτ , multiplied by the distance r. We thus find the stress as a
function of the distance in the form
266 8 Rheometry
Fig. 8.2 Concentric cylinder

geometry Γ
Material
R2
R1
M
τ (r) = . (8.5)
2β hr 2
This shows us how the stress varies across the gap. Indeed, it decreases from
τ1 = M/2β hR12 near the inner cylinder to τ2 = M/2β hR22 near the outer cylin-
der. It is interesting to consider the case when these two bounds are very close, since
we may then treat the stress as uniform across the gap, so that τ1 ≈ τ2 = τ , which
implies that the same is true for the shear rate. To justify this assumption, the maxi-
mum stress must not exceed the minimum stress by more than 10 %, which implies
that R2 /R1 < 1.05. In this case, the shear rate can be estimated by dividing the
relative speed of the two solid surfaces by the width of the gap, whence
α R1
ξ̇ = . (8.6)
R2 − R1
This situation leads to a simple way of deducing the shear rate and shear stress, as
given by (8.5) and (8.6), respectively, from a couple of measurements of the angular
speed and the associated torque. Things are clearly much more complex when the
shear stress can vary significantly across the gap. However, this case can nevertheless
be dealt with by a similar approach to the one used in the parallel disk setup, even
though it is slightly more complex here. We shall describe this briefly in the following.
The layers of material are rotating about the central axis at an angular speed ϕ that
depends on the distance r. Two adjacent layers of thickness dr around the distance
r have relative motion with relative angular speed ϕ∞ dr, corresponding to a relative
tangential speed of rϕ∞ dr. The local shear rate is therefore ξ̇ = rϕ∞ (r). The angular
speed of the inner cylinder can then be expressed in the form
R2
ξ̇
α = dr . (8.7)
r
R1
This can be transformed by the change of variable r ◦ τ (r), using the expression
(8.5) for the local stress, whence
τ2
1 ξ̇ (τ )
α =− dτ . (8.8)
2 τ
τ1
Differentiating this with respect to M, we obtain
∂α (t)
2M = ξ̇ (τ1 ) − ξ̇ (τ2 ) . (8.9)
∂M
Finally, the shear rate can be calculated from a series of expressions of the same kind
corresponding to experiments with stresses (or equivalently, torques) that decrease
by a factor of π = (R1 /R2 )2 at each step. Summing over all the relations like (8.9)
and taking into account the fact that the shear rate associated with a zero stress is
itself zero, we find
∞

∂α
ξ̇ (τ1 ) = 2M . (8.10)
∂M π p τ1
p=0
Note that, when the outer cylinder is very big, i.e., R2 ◦ ∞, we have ξ̇ (τ2 = 0) = 0 ,
whereupon (8.10) simplifies to
∂α
ξ̇ (τ1 ) = 2M . (8.11)
∂M
Once again, this method will only work if the temporal evolution of the behaviour
can be neglected. In this context, the constitutive law can thus be determined exactly
from measurements made with a cylinder immersed in a large volume of the sample,
then using (8.11) to calculate the shear rate at the wall and (8.5) to calculate the
corresponding shear stress.
8.2.4 Flow in a Duct
Here we consider the flow in a cylindrical duct of radius R (see Fig. 8.3). By axial
symmetry around the central axis of the duct, elements located at a given distance
268 8 Rheometry
Fig. 8.3 Flow in a duct
θ
r
z
L
Flow
r have the same speed v(r). The flow thus occurs in the form of a relative slip-
ping of cylindrical layers and the shear rate is −v∞ (r). The friction between these
layers induces a shear stress τ (r). Assuming that the pressure is uniform in each
cross-section, the balance of forces on a cylinder of material of radius r and length
L in the steady-state regime leads to the expression

p(0) − p(L) β r 2 + 2β rLτ (r) = 0 .
The local stress is therefore given by
λp r
τ =− , (8.12)
L 2
where λp = p(0) − p(L), which is positive if the flow occurs from 0 to L. This
relation shows that the stress is highly non-uniform across the material. It varies
from a finite value along the wall of the cylinder to zero at the centre. Clearly, the
same must go for the shear rate, whose value varies significantly from the wall to the
centre of the duct. A rough estimate of the apparent shear rate can be obtained by
dividing the average speed V = Q/β R2 , where Q is the flow rate through the duct,
by the radius. This yields
Q
ξ̇app = . (8.13)
β R3
To a first approximation, we can thus analyse the data by assuming that the con-
stitutive law under simple shear is obtained by associating ξ̇app with the stress
τ (R) = −Rλp/2L at the wall. An analysis of this kind clearly does not take into
account the heterogeneity of the variables and can yield results for the shear stress and
shear rate that bear little resemblance to the actual constitutive law of the material.
A more complete analysis is still possible here. The starting point is the definition
of the flow rate. Integrating by parts, this can be expressed as a function of the local
shear rate:
R R 2
r
Q = 2β rv dr = −2β ξ̇ dr . (8.14)
2
0 0
The change of variable r ◦ −2τ L/λp leads to

RΩ/2
8β
Q= 3 ξ̇ (τ )τ 2 dτ , (8.15)
Ω
0
where Ω = −λp/L. Differentiating (8.15) with respect to Ω, we obtain the following

expression for the shear rate at the inner wall of the duct as a function of measurable
macroscopic variables:
1 d(QΩ 3 )
ξ̇R = . (8.16)
β R3 Ω 2 dΩ
Note. The calculations presented above aim to use macroscopic measurements to

estimate the constitutive laws of materials. They assume that the behaviour under
simple shear is unambiguously defined, i.e., that a single value of the shear rate is
associated with each value of the shear stress. This is not generally true in practice,
because the permanent flow regime of a material is never achieved instantaneously
under the action of an imposed stress or shear rate. In the case of thixotropic or vis-
coelastic materials, the behaviour will vary as time goes by and the relation between
the stress and the shear rate will also evolve almost continually, because it depends
on the whole flow history. Despite these possible problems, and because it would
be very difficult to proceed otherwise, the above formulas are assumed in practice
to reliably describe the instantaneous value of the relation between stress and shear
rate. The way this relation depends on the flow history is subsequently examined in
a second step.
8.3 Perturbing Factors in Rheometry
The errors arising in rheometry come from a mistaken interpretation of the mea-
surements in terms of the intrinsic rheological behaviour of the material. The kind
of factors that can perturb these measurements are always ones that disallow the
calculations outlined above to deduce the constitutive law of the material from the
measured macroscopic variables. This happens if the form of the sample or its flow
characteristics do not match the basic assumptions justifying the above calculations.
270 8 Rheometry
It also happens if the sample is not homogeneous, or if its behaviour depends on the
flow history and this history differs at different points of the material in the gap.
8.3.1 Perturbations of the Sample Volume
In the calculations of Sect. 8.1, we assumed a precise theoretical shape for the sample
in each geometry. When the sample is loaded, the ideal shape is usually obtained to
a good approximation, but during the flow, various effects tend to make its shape or
volume evolve in those geometries which involve a free surface. Globally speaking,
these effects have greater impact on the measurements when the ratio of the free
surface to the solid surface in contact with the material is higher. So this kind of effect
does not arise with the capillary geometry because there is no free surface, and it is
usually negligible in concentric cylinder geometries with narrow gap R2 /R1 < 1.05.
When the gap in the latter geometry is large, there is nevertheless a risk of inertial
effects at high shear rates because these tend to tilt the free surface down toward
the central axis and thus reduce the contact area between the material and the inner
cylinder.
Volume effects can be critical in the cone–plate and parallel disk geometries.
This can be understood from a simple calculation. Suppose the actual radius of the
sample in the gap is R − ν, with ν ⇒ R, and we carry out the rheometric calculations
leading to the constitutive law τ (ξ̇ ) using the theoretical radius R. We then obtain
the apparent shear rate (8.2) and the apparent stress (8.3). The effective shear rate
α (R −ν)/H differs only slightly from the apparent shear rate. However, the effective
stress in the material, also calculated using the true shape of the sample, is equal to

3M 3ν
τeff = ≈ 1 + τapp .
2β(R − ν)3 R
For example, a fractional difference of ν/R = 10 % between the true radius and the
assumed radius of the sample leads to a 30 % error in the stress if no correction is
made.
A change in the effective radius of the sample can easily occur during flow in these
geometries because the free surface is subject to various stresses, such as interfacial
tension with the surrounding air or with the solid surfaces, shear induced by the upper
solid surface, inertial effects which tend to eject the material, gravitational effects,
and evaporation of interstitial liquid. The material may also tend to spread outside
the geometry, or the interface may hollow out toward the interior. The effects on the
effective stress and the corrections required to deal with them are no simple matter,
whatever the shape of the sample, and the effective shear is never completely under
control within the sample, and especially toward the periphery of the sample, in cases
where one of the above phenomena is operating. In practice, the best approach seems
to be to try to control such phenomena. This can be done by observing (qualitatively)
any change in the shape of the free surface so as to make the measurements when
8.3 Perturbing Factors in Rheometry 271
it appears to have a roughly constant shape, possibly making some approximate

correction to the measurements to take into account a reduced sheared volume.
8.3.2 Slipping on the Walls
One perturbing factor that often occurs with devices that have very smooth surfaces
is slipping between the sample and the walls. In this case, the velocity profile is not
what would be expected for a homogeneous shear between two solid surfaces, with
the speed varying very rapidly close to one wall. At a rather large scale of observation,
this may give the impression that there is a discontinuity in the speed, although this
would not actually happen in reality. The slip zone is a layer of thickness χ which
remains very small compared with the overall size of the sample and in which the
shear rate is much higher than in the rest of the sample (see Fig. 8.4). This slip layer is
often a region in which the concentration of interstitial liquid is much higher than in
the rest of the material. It may be that, in certain materials, this liquid tends to extract
itself from the material. The thickness and composition of this layer may change with
the flow history. Apart from certain specific cases, the physical mechanisms leading
to the development of the slip layer are not very well understood.
It is worth nothing that, in a fluid comprising a dispersion of elements in a liquid,
the concentration of these elements will naturally decrease as one approaches a
solid wall. Indeed, suppose we try to distribute the centres of these elements in a
statistically uniform manner throughout the sample. Over the solid surface, there
will then remain a region of thickness equal to half the size of the elements in which
there will be no centre. The concentration of elements will thus decrease steadily as
one approaches the solid surface (see Fig. 8.5a). It follows that the simplest way to
avoid slipping is to use rough solid surfaces, with a scale of roughness much greater
than the size of the elements. In this case, if the centres of the elements are again
distributed uniformly, the concentration will remain roughly uniform across the outer
envelope of the roughness (see Fig. 8.5b).
In practice, when the solid surfaces cannot be made rough enough, the only option
is to observe the slipping and measure its impact on the measurements. Considering
what was said above, when the shear rate ξ̇eff is homogeneous outside the slip layer,
i.e., in the homogeneous fluid (see Fig. 8.4), the speed at a distance y is given by
V (y) = US + ξ̇eff y , (8.17)
where US is the slip speed. The standard description assumes that US depends solely
on the shear stress, which is constant in the given region, to a first approximation.
Consider now the flow of a fluid between two plates in relative parallel motion a
distance H apart. If slipping occurs along each wall, the relative speed of the plates
is V = 2US + ξ̇eff H, so the apparent shear rate ξ̇app = V /H is given by
272 8 Rheometry
Fig. 8.4 Velocity profile in a (a)

flowing material when there
V
is slipping at the solid wall.
a On the scale of observation,
the speed has a finite value
(the slip speed) where the
Solid .
profile intersects the wall.
b On a scale of observation N wall γ eff
times smaller, in the slip layer,
the speed is seen to decrease Slip
rapidly to zero at the point of velocity
intersection with the wall
y
(b) V
Slip Homogeneous
layer fluid
.
γ eff
Slip
velocity
δ y/N
2US
ξ̇app = + ξ̇eff . (8.18)
H
If slipping is neglected, this leads to an underestimate of the effective viscosity γeff
of the material, since the apparent viscosity γapp deduced from the macroscopic
measurements is given by γapp = τ/ξ̇app < τ/ξ̇eff = γeff .
It is interesting to note that a series of tests with increasing thicknesses of material
and the same shear stress should in principle lead to the same slip speed, whence
the slipping term in (8.18) will eventually become negligible when the characteristic
size of the sample is large enough, i.e., ξ̇app ◦ ξ̇eff when 1/H ◦ 0.
Furthermore, if we carry out two analogous tests with the same stress τ and two
different thicknesses H and H √ , we obtain two values ξ̇app and ξ̇app √ of the apparent
shear rate, but in principle the same value of ξ̇eff , since the stress remains unchanged.
The system of two versions of the Eq. (8.18) can then be solved to give the slip speed
and effective shear rate, and hence the effective constitutive law of the material
outside the slip layer:
H ξ̇app − H √ ξ̇app
√
ξ̇eff (τ ) = , (8.19)
H − H√
(a) y
Solid surface
(b)
Solid surface
Fig. 8.5 Concentration of suspended elements as a function of the distance y from the furthest
solid point. a Smooth surface. b Rough surface
√
ξ̇app − ξ̇app
US (τ ) = . (8.20)
2/H − 2/H √
This idea can be applied in the parallel disk and concentric cylinder geometries,
possibly with some refinements to deal with the non-uniformity of the shear rate in
the gap.
Paradoxically, even though the cone–plate geometry does in principle provide a
uniform shear, it is not as simple to analyse the measurements when there is slipping
on the walls. Indeed, an incompatibility arises in the velocity field if we try to assume
that slipping depends only on the shear stress and that this shear stress is homogeneous
in the gap. Equation (8.18) gives rϕ(r) = 2US + ξ̇eff rθ in this case, but the rotation
of the cone also requires that this speed should be equal to α r. This equality cannot
be upheld for all values of r, implying that the assumptions are not justified and the
stress distribution must be more complicated.
8.3.3 Migration
When a sample is prepared, a number of components are generally mixed together in

as uniform a manner as possible. The material is then necessarily discontinuous on
274 8 Rheometry
a scale slightly greater than the scale of the molecules in the interstitial liquid. The
elements in suspension never have exactly the same density as the interstitial liquid.
In addition, on the scale of the molecules in the interstitial liquid, the flow will be
disturbed by the presence of these elements. These phenomena can introduce force
imbalances on the elements, leading to migration effects, as already described for
suspensions (see Chap. 3). For example, density differences induce sedimentation of
dense particles or creaming of bubbles or droplets that are less dense than the liquid
in which they were originally immersed. Migration effects can also be observed in a
particular direction within a heterogeneous flow. This happens for example for flow
in a duct or in the concentric cylinder rheometer, where the suspended particles often
tend to move toward regions of lower shear, i.e., toward the centre of the capillary or
toward the outer cylinder. The consequences for the apparent viscosity were discussed
in Sect. 3.6. In general, if the elements tend to group together in some specific region,
this leads to greater flow in those regions that have become more liquid due to the
departure of the elements, giving an apparently less dense fluid.
8.3.4 Shear Bands
Some materials, despite being homogeneous, appear to have unstable behaviour even
at low speeds. This kind of behaviour has mainly been observed in attractive colloidal
suspensions (see Sect. 5.9). The flow curve of these materials is such that there is no
possibility of homogeneous flow below some critical shear rate ξ̇c (see Fig. 8.6). Such
behaviour is closely related to the thixotropic nature of the material. The instability of
flow below ξ̇c is due to competition between restructuring effects intrinsic to the fluid
and destructuring effects induced by the flow. It is thus a delicate matter to speak of
a single flow curve. When the stress is ramped up, the liquid regime is reached when
the stress exceeds a certain value which increases with the length of time the material
was previously left to rest. The apparent flow curve then has a plateau which gets
higher with increasing rest time (see Fig. 5.24). The critical shear rate associated
with the solid–liquid transition, which corresponds roughly to the transition from
the stress plateau to the rising flow curve, also increases with the rest time. If the
stress is now very gradually reduced, starting from the liquid state, for example after
imposing a strong shear on the fluid, we obtain flows over a broader range of shear
rates, and in the end the fluid stops flowing for some specific shear rate. This value ξ̇c
is the critical value for which it is impossible to obtain a stable homogeneous flow,
whatever the flow history.
If we now start with the material in a homogeneous shear state with apparent shear
rate ξ̇app greater than ξ̇c , and if we impose a value of ξ̇app below ξ̇c , the flow cannot
remain homogeneous. If the material remains homogeneous, the only solution is for
the velocity profile to be heterogeneous. In a simple shear scenario, we naturally
expect to obtain bands of different thicknesses, sheared at different shear rates. The
simplest solution is to have one band of thickness h sheared at ξ̇c , while the rest of the
material, over a thickness H −h, remains rigid with ξ̇ = 0. The exact location of these
Shear-band Homogeneous
flow
τc
. .
γc γapp
Fig. 8.6 Apparent behaviour of a material in which shear bands have formed
bands depends on the precise characteristics of the stress distribution. Whatever the
geometry, there are always some small heterogeneities that allow the fluid to evolve
into the solid regime wherever the stress is very slightly lower than the yield stress,
but to continue to flow freely wherever the stress is very slightly higher than the yield
stress.
Under these conditions, the relative speed of the moving parts of the device is
V = hξ̇c and the sheared thickness is
ξ̇app
h= H. (8.21)
ξ̇c
We observe that this thickness decreases steadily as the shear rate goes down. One
might imagine a more complex shear distribution, but the above distribution is the
most likely because it minimises energy dissipation. Indeed, if the shear takes place
over a thickness h∞ < h, the shear rate in this region is ξ̇ = (H/h∞ )ξ̇app and the power
dissipated per unit area is τ ξ̇ h∞ = τ ξ̇app H, whereas it is equal to τc ξ̇c h = τc ξ̇app H
in the simpler case described above. Since τ (ξ̇ ) > τc , the dissipated power is higher
when h∞ < h.
If an apparent shear is imposed at very low values of ξ̇app , the width h can reach
values of the same order as the size of the suspended elements. In this case, the
assumption that the medium is continuous is no longer justified in the sheared region.
We then observe different kinds of behaviour depending on the structure of the mate-
rial (see Fig. 8.6). The apparent flow curve may tend to get displaced from the plateau
associated with the yield stress, depending on the nature of the suspended elements
(and in particular the mutual interactions they develop) and their concentration.
276 8 Rheometry
8.3.5 Instability Associated with a Decreasing Flow Curve
In practice, decreasing flow curves are sometimes measured. In other words, there is
a range of shear rates over which the stress decreases when the shear rate increases.
This kind of phenomenon can be observed with thixotropic materials like those
described above, i.e., if we impose a shear rate below the critical shear rate, the
material tends to reorganise itself in such a way that the stress required to maintain
the flow becomes higher than the stress associated with a stable homogeneous flow
at a high shear rate. We saw above that, for these materials, at a given imposed shear
rate, this phenomenon is in fact associated with the existence of shear bands with
low shear rates. More generally, it can be shown that any decreasing region of the
apparent flow curve cannot correspond to stable flow in a homogeneous material.
Consider the ideal simple shear of a material between two infinitely long and
infinitely wide parallel plates, moving parallel to one another. To show that this
flow is unstable, we carry out a standard linear stability analysis. We assume that
the ‘theoretical’ stable and homogeneous velocity field is perturbed and we observe
the theoretical evolution of this perturbation. In stable flow, the magnitude of the
perturbation decreases as time goes by and the system recovers the expected flow
characteristics for the imposed boundary conditions. On the other hand, if the mag-
nitude of the perturbation increases as time goes by, that means that the system is
unstable. Here we consider that the ideal shear given by (A.28) is perturbed, so the
principal component of the velocity field is given by
u(y, t) = ξ̇0 y + ν(y, t) , (8.22)
where ξ̇0 is the (constant) shear rate of the unperturbed theoretical flow. We disregard
any perturbations of the other components of the velocity field and any perturbations
of u and the pressure in the x and z directions, which amounts to assuming that the
shear surface (the surface of the plates) is much bigger than the sheared thickness of
fluid (lubrication hypothesis).
Under these conditions, the equations of motion (A.19) are given to first order by
∂u ∂τ ∂ ξ̇ ∂τ ∂ 2 u ∂τ
τ = = = 2 . (8.23)
∂t ∂y ∂y ∂ ξ̇ ∂y ∂ ξ̇
We now assume that the perturbation can be expressed as a sum of periodic per-
turbations (independent modes) of the form ν = ν0 exp(ϕt + iky), where ν0 ⇒ 1.
Inserting this into (8.23), we obtain
k 2 ∂τ
ϕ=− , (8.24)
τ ∂ ξ̇
which shows that the magnitude of the perturbation, proportional to exp(ϕt), grows
indefinitely as time goes by, i.e., the flow is unstable (ϕ > 0), when
∂τ
<0. (8.25)
∂ ξ̇
Therefore, if the flow curve of a material has a decreasing part, the corresponding
flow is unstable. This holds for all kinds of simple shear flow as discussed in this
chapter.
8.3.6 Other Perturbing Factors
Inertial effects arising when the material and/or the shearing device is set in motion
can also disturb the measurement process. The impact of these effects can be esti-
mated by calculating the time required to impart a certain speed to an object whose
moment of inertia with respect to an axis of rotation corresponds to the total mass of
the material and the device. However, in some cases, one must also take into account
the response of the material to a given load, especially in the case of viscoelastic
materials. Measurements carried out in the transient regime, i.e., under a sudden
change in the level of stress or speed, can be significantly affected during the first
few moments of the flow.
Recall that measurements made to establish the constitutive law of a material
assume that the flow occurs in the laminar regime (see Appendix A.8). It is hard to
predict the exact conditions under which turbulence will begin to play a significant
role. Indeed, turbulence is still poorly understood in the case of non-Newtonian fluids,
and the presence of suspended elements may underlie more complex mechanisms
than those envisaged in fluids that are simply composed of small molecules. To a first
approximation, referring to the transition ranges for capillary flows, for instance, we
may consider that turbulence effects, if any, will come into play at Reynolds numbers
above about 100.
There is also a possibility of the temperature rising during shear as a result of
viscous dissipation (see Appendix A.1). In general, such temperature rises are rel-
atively small and slow to occur, so this is an effect that only needs to be taken into
account when making very accurate measurements at a preordained temperature.
The liquid or liquids contained in the samples under investigation can evaporate at
the interfaces with the air. Since evaporation results from a vapour density gradient
between the outer surface of the sample and the surrounding air, it will happen
all the more quickly if the sample is sheared rapidly because vapour can then be
more efficiently evacuated into the air. Evaporation of the liquid phase in the sample
can be a perturbing factor if the sample volume changes significantly as a result.
When this evaporation leads to a non-uniformity in the distribution of liquid in the
sample, possibly with total drying in some regions, the problem becomes delicate.
If dry regions can attach themselves to the solid walls, this will increase the stresses
needed to shear the material. This phenomenon may give the impression that the
fluid gets more viscous as time goes by. To make allowances for this effect, one can
maintain a high vapour density around the sample by placing it in a casing that limits
278 8 Rheometry
contact with the surroundings, or simply insulate the interface from the surrounding
air by placing a film of oil over the free surface of the sample.
8.4 Experimental Procedures
In this section, we shall review various aspects of the experimental procedures

involved in measuring certain rheological characteristics. This is a very broad sub-
ject and we shall thus limit the discussion to a general outline. Moreover, to be as
effective as possible, the experimenter must adapt these procedures in real time to
the behaviour of the material and perturbing effects that come into play to differ-
ing degrees depending on the method chosen. To discuss these procedures, we shall
assume here that we are dealing with an arbitrary material, and consider the various
possible ways to probe it. The discussion follows the different stages involved in
carrying out a measurement, starting with the choice of measurement geometry, then
going on to the preparation and loading of the sample, the first assessment of the
behaviour of the material on the basis of its flow curve, and finally, more specific
measurements of its rheological properties.
8.4.1 Choosing the Geometry
The choice of geometry depends first and foremost on the apparent viscosity of the
material. If it flows very easily, parallel disks are ruled out because the fluid will
simply flow out of the gap, unless of course a very small gap can be used, but in
this case it is better to use a cone–plate geometry. If the elements contained in the
fluid are very small compared with the distance between the truncated apex of the
cone and the plate, this becomes a real option. In the opposite case, jamming effects
may occur and then it is better to switch to a concentric cylinder device. If on the
other hand the material has a paste-like consistency, the parallel disk geometry is
well suited because it is particularly easy to load the sample. The concentric cylinder
geometry is more difficult to use with a pasty fluid, because during loading there is a
risk of air pockets getting trapped between the fluid and the bottom of the container.
To avoid slipping on the walls, it is essential to use rough surfaces whenever
the material contains elements with dimensions greater than the micron, this being
the order of magnitude of the roughness of metal surfaces in commercial rheometer
geometries.
8.4 Experimental Procedures 279
8.4.2 Preparing the Sample
No specific procedure is needed to measure the viscosity of a Newtonian fluid, since

this parameter does not depend on the flow history. The situation is quite different for
complex fluids whose behaviour depends on the instantaneous state of the material,
which itself depends on the flow up to that point in time. For example, for a yield
stress fluid, it is essential to know in which regime the material happens to be (solid
or liquid) or which deformations it may have undergone in the solid regime. For a
viscoelastic fluid, it is crucial to know whether the material was previously subjected
to stresses that have not yet been fully relaxed. For a thixotropic fluid, it is essential
to know how long the material has been flowing or how long it has been at rest,
because this will contribute to the structural condition of the material. If we wish our
measurements to reflect the true behaviour of the material, we must at least in part
control the changes in the state of the material since its preparation.
What makes this such a difficult question is the fact that the state of the material
can depend upon its whole mechanical history, that is, on all the deformations it
has undergone since it was first prepared, i.e., since its components were first mixed
together. Since the preparation is a stage during which we do not fully control the
flow characteristics, we cannot expect to prepare the material in a given state. But this
preparation phase must be reproducible, so that, even if the resulting state is unknown,
it will at least be the same at the beginning of each experiment. When the sample
is loaded into the rheometer geometry, it will once again undergo deformations that
may vary from one test to the next. For this reason it is not ideal to study the behaviour
of the material just after it has been loaded into the rheometer. For example, for a
yield stress fluid, the deformations it undergoes at this stage may be significant, and it
may flow as a result of even a small further deformation. To get round this problem,
one method often used is to impose a simple shear flow at a high shear rate just
after it has been loaded, the idea being to get the material into an almost completely
‘destructured’ state. From there, we may then impose a new flow history, but this
time carefully controlled, e.g., leaving it at rest for a certain time before shearing it
once again. The aim of this operation is to make it forget its whole prior flow history
and then control its subsequent history. This kind of approach is particularly useful
with thixotropic fluids.
8.4.3 Flow Curve
When we are dealing with a fluid that can flow under its own weight, its flow curve,
i.e., the relation between stress and shear rate, can generally be determined by mea-
surement. In practice, the first procedure to apply for a quick evaluation of the shape
of the flow curve involves imposing an increasing stress over a certain range and
measuring the resulting shear rates. The basic problem with this measurement is that
the shear rate is recorded without really knowing whether a permanent flow regime
280 8 Rheometry
has been achieved at each new level of stress. To make the results somewhat more
reliable, but without fully resolving this difficulty, we can ramp up the shear rate in
stages, ensuring that each stage lasts for a time inversely proportional to the corre-
sponding level of the shear rate. In this way, the deformation of the material can be
made identical at each level.
The simplest way to check that the measurement points obtained during the
increasing ramp are close to the permanent flow regime is to impose a decreas-
ing ramp just afterwards. If this second curve matches the first, we may conclude
that the measurement points, which coincide even though they result from different
flow histories, do indeed correspond to the permanent flow regime. If the points are
shifted, i.e., if we find that the measurement curves resulting from increasing and
decreasing ramps form a loop, we may conclude that the permanent flow regime was
not reached during either the increasing or the decreasing ramp, or both. In general,
one is in principle closer to the permanent regime during the downward ramp, simply
because in this case the flow history corresponds to a previous shear at high shear
rate, followed by a gradual reduction in the shear rate. In some cases, this discrep-
ancy between the increasing and decreasing curves is solely due to perturbing effects
such as the changing shape of the free surface or migration of suspended elements.
To check that one is indeed dealing with an effect intrinsic to the behaviour of the
material, one can impose a second upward–downward ramp just after the end of the
first test, imposing the same rest period beforehand. If there are no perturbing effects,
the loop obtained should be the same as the first.
Note that these hysteresis loops are naturally obtained with thixotropic fluids since
these materials gradually liquefy as time goes by and this effect is all the more marked
for higher imposed stresses. These curves are therefore qualitative characteristics of
the thixotropy of a material. The longer the thixotropic material has been at rest, the
higher the stress required to set it back in motion, leading to a curve at high stress
levels for the increasing ramp, while the decreasing curve corresponds to a material
that has been destructured, thus leading to much lower stress levels (see Fig. 5.24).
However, such a hysteresis loop is not an intrinsic characteristic of the material since
its exact shape depends on the details of the upward and downward ramps, i.e., the
range covered and the timing.
To avoid the various problems mentioned above and hence obtain a precise mea-
surement of the flow curve of a material, it is better to follow a simpler but longer
protocol. The idea is to start from a destructured state and impose a given stress τ
for a long enough time to obtain a permanent flow regime. The associated shear rate
ξ̇ is measured and we obtain a point (τ, ξ̇ ) on the flow curve. This operation is then
repeated in similar tests for other stress levels, thereby constructing the flow curve
point by point.
Whatever technique is used, the flow curve is determined over some restricted
range of shear rates or stresses. These limitations in terms of maximum stress are
due to the limited capabilities of the device. The minimum stress depends not only on
the accuracy of the device, but also on the fact that the slightest perturbing effect on the
free surface of the sample can induce a shift or a temporal variation in the measured
stress level, to the point where the stress effectively imposed on the material is in
8.4 Experimental Procedures 281
fact screened. Limitations concerning the maximum shear rate arise due to possible
inertial effects like turbulence or ejection, as described above. Measurements carried
out at very low shear rates raise the problem that measurement times required to
impose any notable deformation on the material may turn out to be very long, thus
allowing perturbing factors such as slip, migration, or drying to play a significant role.
For this reason, apart from certain specific cases, e.g., simple liquids or through very
careful experimental conditions, it is unreasonable to expect reliable measurements
at shear rates below 10−2 s−1 .
Note. It is essential use a logarithmic scale when representing the flow of a fluid. To
begin with, one can then detect the existence of a yield stress directly by looking to
see if there is a stress plateau at low shear rates. In a plot with a linear scale focusing
on high values of the shear rate, it can be hard to make out the value of the yield stress
if all the points associated with low shear rates are clustered against the ordinate axis.
A logarithmic plot also has the advantage that it does not favour any particular range
of shear rate, which is important when we need to carry out a fundamental analysis
of the behaviour of the material.
8.4.4 Solid–Liquid Transition

Given the particular behaviour of yield stress fluids, specific procedures are needed in
this case. To begin with, it is essential to clearly identify the two regimes of behaviour
of the material, i.e., the solid regime and the liquid regime, in order to make reliable
measurements. This distinction is necessary as soon as we wish to measure the flow
curve by the procedure described above. In this case, the upward ramp gives a curve
with nonzero slope at low stresses, followed by a plateau and a further increasing part
(see Figs. 5.18 and 5.24). This happens because a rheometer records an average shear
rate by dividing the observed deformation by the duration. For this reason, we may
record a nonzero shear rate in the solid regime because the deformation increases
as the strain increases. For example, for an elastic material (τ = Gy), if we apply
a linearly increasing stress ramp τ = kt and estimate the shear rate over a constant
time lapse λt, we observe a constant shear rate since ξ̇ (τ ) = λξ /λt = k/G.
To distinguish the different regimes, one must monitor the deformations of the
material using the criterion that a material is solid if it deforms only a finite amount
when subjected to a given stress. One then imposes different stress levels and observes
the changes in the deformation as time goes by (see Fig. 8.7). For a yield stress fluid,
there are two different regimes. Below a critical stress, the yield stress of the material,
the curves showing the deformation as a function of time eventually reach a plateau,
or at least the apparent shear rate continues to decrease, suggesting that the material
is tending to come to rest and that the observed flow results from some residual
viscoelastic or plastic deformations. Above this stress, the deformation curves tend
asymptotically to straight lines with nonzero slopes, indicating that the material is
in permanent flow at a shear rate given by the slope of these straight lines.
282 8 Rheometry
γ
t
onstan .
τ=c γ
γc
τ/G
τ/μ
t
Fig. 8.7 Temporal evolution of deformation during creep tests at different stress levels for a yield
stress fluid. The solid regime corresponds to finite deformations, less than the critical value ξc .
The liquid regime is obtained for deformations above this critical value, the rate of change of the
deformation (shear rate) then tends to a constant value at the imposed stress
8.5 Practical Measurement Techniques
Yield stress fluids have a key rheological property, the existence of a yield stress where
the behaviour of the material changes radically, and in fact there is a solid–liquid
transition. This has an interesting practical consequence: the yield stress of such a
material can be estimated by observing this transition under simple flow conditions
and measuring the associated characteristics. The simplest conditions for observing
this transition are those in which the material goes from the liquid to the solid state.
In practice, the idea is to impose a flow by applying a certain stress to the material,
then gradually reduce this stress until the flow ceases. The main setups for this type
of measurement are squeeze tests, where the fluid is contained between two surfaces
which are brought toward each other, but also spreading over a solid surface, and the
displacement of an object through the fluid. The last situation has been discussed in
Sects. 3.2 and 3.6.
8.5.1 Squeeze Tests
In this kind of technique, which is used in particular for civil engineering materials
and in the food industry, a sample of the material is squeezed by bringing together
two parallel solid walls. Since the normal force required to maintain the motion
increases as the thickness of the material is reduced, there are two possibilities for
measurement here. Either we can impose a given force and measure the final thickness
hc associated with this force, or we can squeeze the sample down to a certain thickness
hc and measure the associated normal force. In both cases, we obtain the critical force
corresponding to the arrest of flow at a thickness hc .
8.5 Practical Measurement Techniques 283
z z
O 2b
r
R
Mid-plane
V Material
Fig. 8.8 Squeeze flow between two parallel disks
When the thickness of the sample is much smaller than its dimensions in the
plane of the solid surfaces, we can use the lubrication hypothesis to estimate the
relation between the force and the critical thickness. By bringing the two solid plates
together, we induce a radial flow which tends to squeeze fluid out of the gap between
the two plates. To achieve this, we must impose a higher pressure at the centre than
at the periphery, to ensure that the fluid flows in this direction. The force that must be
applied to bring the two plates closer results from this pressure. It increases rapidly
when the thickness of the layer is decreased.
It is easy to obtain an expression for this force in the case of two disks of radius R,
a distance 2b apart, moving together at speed V (see Fig. 8.8). From the conservation
of mass, we deduce the average radial speed at distance r to be given by
b
1 r
U(r) = vr (r, z)dz = − V. (8.26)
b 2b
0
Over the plane midway between the plates, the shear stress is zero by symmetry.
Considering the balance of forces on a volume located between the heights 0 and
z and radii r and r + dr, we find that the radial force due to shear over the upper
surfaces at height z, i.e., 2β r drτ , must be balanced by the force

2β rz p(r) − p(r + dr)
due to the drop in pressure, whence
∂p
τ (r, z) = z. (8.27)
∂r
284 8 Rheometry
The pressure gradient is negative when the plates are brought together because the
pressure is higher at the centre than at the edge. On the other hand, it is positive when
they are pulled apart, creating a pressure dip in the fluid at the centre.
For a Newtonian fluid, we have τ = μdvr (r, z)/dz, so integrating (8.27) between
b and z and using the boundary condition vr (r, b) = 0, we obtain the velocity
distribution
1 ∂p 2
vr (r, z) = (z − b2 ) . (8.28)
2μ0 ∂r
A second integration gives the average speed at distance r in the form
b2 ∂p
U(r) = − . (8.29)
3μ0 ∂r
From (8.26) and (8.27), we deduce that ∂p/∂r = 3μrV /2b3 , which can be integrated
between r and R to give the pressure distribution
3μ(r 2 − R2 )V
p(r) = . (8.30)
4b3
Here we have neglected effects due to surface tension and atmospheric pressure,
assuming that p(R) = 0. Integrating (8.30), we obtain the total force exerted on
the plates and resulting from the motion of the fluid associated with bringing the
plates together when the radius R of the layer remains constant, i.e., when the fluid
is allowed to flow out of the gap:
R
3β μR4 V
F=− 2β rp(r) dr = . (8.31)
8b3
0
The normal force thus diverges rapidly when the distance tends to zero.
For a yield stress fluid with constitutive law in the liquid regime given by (A.39)
(see Sect. A.10.7), taking into account the characteristics of the flow, we obtain

τ (r, z) < τc =⇒ ξ̇ = 0 ,
(8.32)
ξ̇ ≈= 0 =⇒ τ (r, z) = ν τc + F |ξ̇ | ,
where ν is equal to 1 for squeezing and −1 for separation of the plates. From (8.27)
and (8.32), we deduce that there is a sheared region and a non-sheared region, i.e.,
ξ̇ = 0 for |z| < z0 and ξ̇ ≈= 0 for |z| > z0 , with
τc
z0 = ν . (8.33)
∂p/∂r
In the limiting case where the non-sheared region occupies almost the whole volume
of the sample, i.e., z0 = b, so conditions are such that the flow is just beginning or
has come to a halt, integration of (8.33) between R and r while neglecting the effects
of surface tension and atmospheric pressure leads to the pressure distribution
τc
p(r) = ν (r − R) . (8.34)
b
By integrating again, we finally obtain the total normal force on the solid plate:
β τc R3
Fc = ν . (8.35)
3b
This can be used to estimate the thickness at the point where the squeezing motion
ceases for a given force when the volume η0 of the squeezed sample remains con-
stant: 1/5
τc2 η03
bc = . (8.36)
18β F02
The critical force needed to reach a certain thickness can also be deduced from this
equation.
8.5.2 Inclined Plane
This rheometric technique is the most natural when studying the behaviour of yield
stress fluids since their key property is that they form deposits of significant thickness
on nonzero slopes. We begin by discussing the use of the inclined plane to determine
the flow curve of a material. We then consider the procedures and calculations used
to estimate the yield stress of a fluid from the form of the deposits on an inclined
plane.
Consider first the uniform flow of a layer of fluid down an inclined plane making
an angle i with the horizontal. By symmetry, the velocity has only one nonzero
component parallel to the plane, and the velocity profile v(y) is independent of
the cross-section taken along the x axis. Everything happens as though fluid layers
parallel to the solid plane slipped over one another. We thus have a simple shear in
which the shear rate at height y is given by v∞ (y).
We now consider momentum conservation [see (A.19)] applied to the fluid ele-
ment bounded by the free surface, the plane surface at height y, and two cross-sections
perpendicular to the plane and located a distance L apart (the region indicated by
dashed lines in Fig. 8.9). Projecting in the x direction, this equation reads
τ (y) = τg(h − y) sin i , (8.37)

286 8 Rheometry
Fig. 8.9 Fluid flow on an inclined plane
where h is the thickness of the fluid layer. This tells us that the shear stress depends
linearly on the depth and reaches its maximum value at the bottom:
τw = τgh sin i . (8.38)
Projecting in the y direction, we obtain the pressure distribution, which is of the

hydrostatic form:
p = p0 + τg(h − y) cos i . (8.39)
The shear rate cannot be controlled in this kind of flow, but the fact that the stress
distribution (8.37) is known means that one can obtain the constitutive law from a
set of measurements of macroscopic variables. In practice, it is easy to measure the
rate of flow q (per unit width) through a cross-section and the speed U of the free
surface of the fluid. These quantities can be related to local characteristics of the
motion, since we have
h τw
h2
q= v dy = 2 τ ξ̇ (τ ) dτ , (8.40)
τw
0 0
h τw
h
U = v(h) = ξ̇ dy = ξ̇ (τ ) dτ . (8.41)
τw
0 0
Differentiating (8.40) and (8.41), we obtain expressions for the shear rate at the wall
ξ̇w = ξ̇ (τw ) in terms of the stress at the wall:

1 dq
ξ̇w = , (8.42)
h dh i

dU
ξ̇w = . (8.43)
dh i
Theoretically, (8.42) or (8.43) could be used to determine the flow curve in the form
of a set of pairs (ξ̇w , τw ), derived from a broad range of pairs (h, U) or (h, q).
The above calculations are more generally valid when the layer of fluid has a very
large area of contact with the plane as compared with its thickness. The expression
(8.37) for the stress shows that, beyond a certain height relative to the bottom, the
shear stress is less than τc . No flow is possible in this region, implying that there is
a region that is not sheared close to the surface, with thickness given by
τc
hc = . (8.44)
τg sin i
When the height of fluid is less than this thickness, there is no uniform permanent
flow with a yield stress fluid. This observation proves useful in practice since it shows
that the yield stress can be determined by measuring this critical thickness. To do
this, since (8.40) is only valid at the transition between the solid and liquid regimes,
one must begin by getting the material to flow, then gradually reducing the flow rate
or the slope until the flow ceases altogether. Another solution would be to spread
a layer roughly uniformly over a plane and then tilt the plane gradually until flow
begins, but it is not easy to detect the beginning of flow in the liquid regime, since it
may be confused with deformations in the solid regime.
Further Reading
1. Bird, R.B., Stewart, W.E., Lightfoot, E.N.: Transport Phenomena. Wiley, New
York (2001)
2. Coleman, B.D., Markowitz, H., Noll, W.: Viscometric Flows of Non-Newtonian
Fluids. Springer, Berlin (1966)
3. Coussot, P.: Rheometry of Pastes, Suspensions and Granular Materials. Wiley,
New York (2005)
4. Ferguson, J., Kemblowski, Z.: Applied Fluid Rheology. Elsevier, Amsterdam
(1991)
5. Piau, J.M.: Fluides non-newtoniens. Techniques de l’Ingénieur A710, 1–16,
(1979) (A711, 1–24)
6. Tanner, R.I.: Engineering Rheology. Clarendon Press, Oxford (1985)
7. Truesdell, C.: A First Course in Rational Mechanics. Academic Press, San Diego
(1991)
Appendix A
Fluid Mechanics: Physical Principles
A.1 Introduction
The aim of fluid mechanics is to describe the characteristics of flows using systems of
equations that account for the forces acting on the material and its intrinsic behaviour.
In the case of simple fluids, i.e., Newtonian fluids, in the laminar regime, i.e., when
the flow velocities are not too high, the theoretical basis is by now perfectly well
understood and many kinds of flow can be successfully described. However, the
mechanics of complex fluids, i.e., non-Newtonian fluids, provides a vast field for
future research that has still only been touched upon. The difficulty here is twofold: the
constitutive laws are not well established and their complexity makes them difficult
to handle from the point of view of calculating flows.
In this appendix, we shall review the general principles of fluid mechanics, which
actually sit within the broader framework of continuum mechanics. To simplify the
description, we shall focus on tools developed to deal with fluid materials, since
these are the main subject of the present book. However, it should be noted that
these considerations fall somewhat short when dealing with yield stress fluids, since
such materials may be solid or liquid depending on the circumstances. We shall aim
in the following presentation to emphasise the physical origin of the principles of
continuum mechanics in a way that is consistent with the rest of the book.
Once we can provide an accurate description of the interactions between con-
stituent elements, we can in principle predict the flow characteristics of a fluid made
up of an ensemble of such components as a function of the forces applied to it. If the
material only contains a few components, the task is relatively straightforward. When
the system under investigation comprises a very large number of components, which
is clearly the case when we consider an arbitrary fluid, it is extremely difficult, even
with the help of modern computing, to predict the motion of each component and
deduce from that the macroscopic features of the flow. So in order to obtain a good
physical grasp of the relevant phenomena and prepare for real practical applications,
it is crucial to develop analytical tools able to describe the collective properties of
an ensemble of components, i.e., a fluid, in motion. This is the whole point of the
formalism and the tools of continuum mechanics.
DOI: 10.1007/978-3-319-06148-1,
290 Appendix A: Fluid Mechanics
A.2 Flow Variables
Since fluid mechanics was developed mainly in order to describe the flows of simple
fluids such as air and water, whose properties do not depend on their deformation
history, the usual starting point is to consider the velocity of the fluid at a given
time and place. This is the so-called Eulerian description. This velocity u is the
velocity of the component that happens to be passing through this point at this
precise time. The aim then is to describe the time dependence of the local velocity
distribution within the material. Another possibility is to follow the velocity of each
fluid component. This is the Lagrangian description of the motion. The aim in this
case is to describe the trajectory of each of these components, starting from their
initial position. This kind of description is clearly well suited to studying limited
deformations, or materials with some preferred configuration which never stray far
from that configuration. In fact certain classes of material such as viscoelastic fluids
or yield stress fluids have instantaneous properties that depend on deformations they
have undergone relative to a specific configuration. For these materials, it may be
useful to use both kinds of description, depending on the state in which the material
happens to be. Here we apply a rather general definition of what constitutes a fluid,
according to which they are materials capable of deforming at will without losing
their mechanical properties, even though their instantaneous behaviour may depend
on the history of deformations they have undergone previously.
Two other variables characterise the state of a fluid: the density, which indicates
the number of components per unit volume, and the temperature, which gives an idea
of the degree of agitation of the components. We shall also use a Eulerian approach to
describe these variables. In short, we seek to predict the evolution in space and time
of three variables characterising the state and motions of the material, viz., its density
ρ, its velocity u, and its temperature T . These variables are functions of the time t
and the local position x with components (x, y, z) relative to a fixed orthonormal
frame (i, j, k). The components of the velocity in the same frame are (u, v, w).
A.3 Continuity of the Medium
It is not as easy as one would think to specify the velocity of a given material com-
ponent located at some precise point. This implicitly assumes that the material is a
juxtaposition of infinitely small components, which is obviously never really the case.
In practice, when we use ordinary methods to measure the value of one of the above
variables at a given point, we actually measure the average value of this variable
over a small volume containing a certain number of fluid components. In an ordinary
liquid, this means that we record the average velocity of some 1015 molecules. More-
over, a measurement of the density must necessarily concern a volume containing at
least enough components to ensure that the result does represent the average value
in the neighbourhood of the given point, while remaining small enough relative to
macroscopic variations of this variable, on the scale of the observer. In continuum
mechanics, the values of the velocity, density, and temperature variables are therefore
Appendix A: Fluid Mechanics 291
Macroscopic
Local discontinuites variations
l
Fig. A.1 Dependence of an average physical property q over a certain volume of material on the
size l of this volume
ensemble averages over representative volume elements Ωe . It would be difficult to

specify this representative volume element a priori because its dimensions must jus-
tify the preconditions for the continuous medium hypothesis, and these preconditions
depend in part on the mechanical behaviour of the material.
To facilitate the mathematical description, we require these variables, which are
also functions of the space and time coordinates, to be continuously differentiable
(and hence also continuous) with respect to those coordinates. In this way we can use
a single set of equations involving space and time derivatives to describe the motion
of the fluid at each point (appending specific relations between the variables on either
side of any discontinuity surfaces). For this to work, the difference ξι between the
values of variables associated with neighbouring points, i.e., associated with two
neighbouring volume elements, must be very small compared with the difference
ξI in the values of this variable at two points separated by a typical macroscopic
length scale of the flow. Likewise, time variations in the value of a variable at a
specific point must be very small compared with variations over the same time of
the macroscopic difference mentioned above.
From this definition, we thus find that the continuous medium hypothesis is nec-
essarily associated with a range of observation scales. Indeed, if we observe on our
scale the flow of a simple liquid made up of a very large number of molecules, it is
easy to divide space up into a multitude of very small volumes, each containing many
molecules. It seems natural then to consider that, from one small volume to the next,
the change in the variable is likely to be small compared with macroscopic variations
in this variable. But if we use such a small observation scale that the groups of mole-
cules used to measure a variable no longer contain more than a few components, the
medium is likely to be discontinuous, because from one group to another, the density,
velocity, or temperature may vary significantly (see Fig. A.1). For ordinary liquids,
just as for colloidal suspensions, this problem will clearly only arise in very specific
flow situations, since the scale of observation is generally very big compared with
the size of the components making up the fluid. In the most favourable cases, the
spatial variations of the given variable ι are linear. We then have ξι = ξI /N , where
N is the number of volume elements of the fluid making up a distance equal to the
characteristic scale L of the flow. In this case, the continuity hypothesis ξι ◦ ξI
for the variable ι is valid whenever N > 10.
Up to here we have only considered a geometric criterion for the continuity of the
medium. Significant differences in the velocity from one group of components to a
neighbouring one can be observed even when these groups are very small compared
with the volume of suspension, simply due to the collective behaviour of these groups,
i.e., due to the mechanical behaviour of the suspension. For example, if the given
material is organised in such a way that a flow separates the material into two more
or less rigid parts which slide over one another, the relative speed of neighbouring
components situated on either side of the slip surface will be of the same order as
the macroscopic flow speed. The continuum assumption is no longer valid.
For simple liquids under ordinary conditions, this kind of problem occurs very
rarely thanks to the disorder and thermal agitation that prevail in such media.
For certain complex fluids, the problem is more delicate because the spatial vari-
ations of a variable ι, which depend on the distribution of forces over the fluid
components and their reactions against these forces, are not linear. A case in point is
the yield stress fluid in certain flow configurations where the velocity may vary very
quickly over a relatively small thickness. Under such conditions the macroscopic
variations in the velocity are of the same order as those at the boundaries of this
region of rapid variations and if the continuous medium hypothesis is to be justified,
there must be a sufficient number of volume elements in the fast variation zone of the
velocity. The problem is that the criterion depends on the flow conditions because the
thickness sheared tends to zero when the stress at the wall tends to the yield stress of
the suspension. With a confined granular suspension in frictional flow (see Sect. 7.5),
the situation is much more critical because, whatever the dimensions of the sample
(for a given particle size), rupture surfaces, or more precisely, shear bands will tend
to form beyond a certain deformation. In this case, there is no simple criterion for
the continuity of the material.
These examples show that it is not possible to state universal criteria for the
continuity of a fluid, especially if they must encompass concentrated suspensions.
Only necessary conditions, in particular geometric conditions, can be put forward.
Continuity depends on characteristics of the material and the flow, which means
that the continuity hypothesis must always be confirmed with hindsight. In the rest
of this book, we assume in principle that the material is continuous, treating any
discontinuity surfaces as singularities in the flow.
A.4 Forces
The motion of a fluid is determined in part by the external forces on it and by the
mutual forces between its constitutive components. These forces fall into two cate-
gories. The first includes the so-called body or volume forces, such as electromagnetic
forces, gravitational forces, or inertial forces, which usually vary slowly with distance
Fig. A.2 How forces between (a)

fluid components lead to
surface forces
ds
i
Fii' (b)
i'
(relative to the size of the fluid components) and have very long range. Consider a
representative volume element Ωe containing a certain number of fluid components,
i.e., liquid molecules and solid particles for a suspension, and such that the medium
can be treated as continuous when we describe the changes in the values of the vari-
ables by dividing space up into like volumes. The external force on this volume is
obtained by summing all the external forces Fi on each component of Ωe :

FΩe = Fi . (A.1)
We then define the external force density at a point of space by b = FΩe /Ωe . An
arbitrary volume Ω can be divided into a certain number of these volume elements.
When b is constant throughout Ω, the external force applied to this volume is given by
Fvol = bΩ. (A.2)
When the only body force is gravity g, we have b = ρg.

The second category of forces are surface forces Fsurf , which act between two
components very close together but decrease rapidly with distance. For an ordi-
nary liquid, this mainly concerns van der Waals forces, which are essentially forces
between neighbouring molecules. For a suspension, these may be forces between
molecules of the interstitial liquid (see Chap. 2), or indeed interaction forces between
the particles, some of which may be electrostatic (see Chap. 5) and could then have
effects over a certain distance through the liquid. However, the magnitude of these
forces still decreases rapidly with distance and it seems reasonable to neglect their
action beyond the nearest neighbouring particle. Otherwise, it would still be possible
to specify volumes (which would then be called elements) which encompass several
fluid components such that the force due to this kind of interaction and exerted by such
a volume on the surrounding volumes would not extend beyond nearest neighbours.
Consider a surface element ds within the fluid, centred on a space point x. We
would like to evaluate the force exerted by the fluid situated on one side (B) of
this surface on the fluid situated on the other side (A). Given the short range of the
surface forces discussed above, only those fluid components located very close to this
surface will suffer any force due to fluid components located along the surface on the
other side (B). To a first approximation, each component i suffers a force Fii due to
each component i situated opposite it on the other side of the surface (see Fig. A.2).
Fig. A.3 Decomposition

of the stress vector into a σn
tangential component τ and t
a normal component σn with
respect to the surface element ds
The total force exerted by the fluid situated on side B on the fluid situated on side A
can then be written in the form of a vector dF such that

dF = Fii . (A.3)
i
If Fii is roughly constant, then dF is proportional to the area ds. It thus makes
sense to consider the variable defined by the vector tσ (x) = dF/ds. When the area
is big enough, tσ is the average value of the force per unit area exerted by each
component on the component opposite it along the given surface. However, as for
the other variables, if the variable tσ (x) is to be continuously differentiable, the area
considered around x must be neither too big, in which case the changes in tσ from one
point to another would be of the same order as the changes between the boundaries
of the sample, nor too small (of the order of the area of a few fluid components), in
which case the changes in tσ from one point to another would be associated with the
intrinsic discontinuities in the fluid. The minimum area dse that can be used for this
purpose will be referred to as a surface element.
Analogously with the volume element, some geometric arguments can be put
forward regarding the evaluation of this surface element, but its exact value will also
depend on the mechanical properties of the fluid. However, assuming this surface
element known, we can now define the stress vector t at a point x in the following way:
dF
t = lim . (A.4)
ds∞dse ds
Note that we have assumed here that the components do not exert torques on one
another.
The stress vector defined in this way can be decomposed into a term along the
normal n to the surface element and a term perpendicular to n, viz., t = σn + τ (see
Fig. A.3). The first term is a normal stress expressing forces that tend to separate
the elements from one another or push them together. The magnitude of this term
is proportional to the way the fluid components are packed together and also to the
attractive or repulsive forces between these components. The second term is the shear
stress, which arises because fluid components situated on one side of the surface tend
Fig. A.4 Small cube of

material for calculating the i
torque exerted on the z axis
(along the vector k) which
passes through the centre O of
the cube σ xy
σyx
j
O
σyx
k σxy
to drag along components located on the other side in a slipping motion in the plane
perpendicular to the direction of n.
Decomposing relative to the three axes of the frame, the stress vector given by
(A.4) on the facet with normal i can be written
tx = σx x i + σx y j + σx z k, (A.5)
where σx x = σ. Likewise for the facets with normals j and k, the stress vectors on
these are t y = σ yx i + σ yy j + σ yz k and tz = σzx i + σzy j + σzz k, respectively.
Consider now a volume of fluid with characteristic length r . Applying Newton’s
second law ma = Fext to this volume and taking into account the two kinds of force
identified above, we obtain
ρr 3 a = Fvol + Fsurf = r 3 b + r 2 ξt, (A.6)
where a is the average acceleration of the volume elements and ξt is the sum of
the stress vectors on the outer surface of the given volume (the surface forces inside
the volume cancelling out in pairs). When r tends to zero in the limit allowed by the
continuous medium hypothesis, (A.5) gives ξt = 0 to first order. Since this is true
for any parallelepiped, we deduce that the stress t−x on the facet with normal −i is
minus the stress on the opposite facet (with normal i). An analogous result can be
obtained in the other directions. We thus deduce that t−x = −σx x i − σx y j − σx z k,
with analogous expressions for t−y and t−z .
Consider now a cubic volume for which the normals to the different faces are
parallel to the three axes of the frame (i, j, k) (see Fig. A.4). The total torque on
this volume along the k axis is the sum of the torques due to the stresses on each
face. The torques due to all the normal stresses have zero resultant because these
stresses induce forces on either side of the z axis. The shear stresses σx z , σzx , σzy ,
and σ yz induce forces along the z axis which do not give rise to any torque about
Fig. A.5 Tetrahedron used to

deduce the stress vector on a j
surface element with arbitrary
normal n 1/b
n
ds 1/a
k i
1/c
this axis. The total torque is due solely to the shear stress σx y and σ yx , and it is
proportional to 2(σx y − σ yx ). There is no reason to expect a nonzero couple on the
material components resulting solely from the stresses. This would cause them to
spin in some particular direction. We deduce therefore that
σx y = σ yx . (A.7)
This argument can be repeated for the torque around the other axes, so we deduce of
course that σx z = σzx and σzy = σ yz .
To find the stress vector on a facet with arbitrary normal ai + bj + ck, we
consider the stress vector t = σa i + σb j + σc k on the tetrahedron constructed by
joining the origin and the plane with equation ax + by + cz = 1 by means of three
straight lines along each of the three axes (see Fig. A.5). If we know the different
components of the stress identified above, we can calculate the stress vector for each
of the three facets lying in the planes x = 0, y = 0, and z = 0. When the volume of
the tetrahedron tends to zero, (A.6) tells us once again that ξt = 0, and with (A.7)
and similar, this implies that
σa = aσx x + bσx y + cσx z ,

σb = aσx y + bσ yy + cσ yz , (A.8)
σc = aσx z + bσ yz + cσzz .
This shows that we can express the force exerted on a small element ds of the material
with arbitrary outward unit normal n in the form
dF = t ds = (Σ · n) ds, (A.9)
where Σ is a symmetric tensor according to (A.7) and the analogous relations. This
is the stress tensor. Its components in the upper right triangle including the diagonal
are σx x , σx y , σx z , σ yy , σ yz , and σzz . Note that (A.9) specifies a sign convention for
the stress because we chose the outward normal, oriented from the inner face to the
outer face, to calculate the stress exerted by the outside on the given surface. It should
also be noted that the expression for this tensor depends on the choice of frame. In
another frame obtained by a rotation of the initial frame represented by a matrix R,

it takes the form Σ R = RΣRT . However, the tensor Σ has three invariants, which
do not depend on the choice of frame, namely, its determinant, its trace, and another
invariant defined by
1⎡ ⎢ ⎣
Σ = (trΣ)2 − tr Σ 2 . (A.10)
2
Simplifying, we can say that this variable expresses the strength of the shear. There
is always a special frame in which only the diagonal components of the stress tensor
are nonzero. These components are in fact the eigenvalues of this tensor, referred to
as the principal stresses. Naturally, these do not depend on the frame, which explains
the frame independence of the three invariants named above.
Note finally that in general a volume of liquid will only deform significantly if
different stresses are applied to its various facets. For this reason, the stress tensor is
decomposed into a pressure term which deals with stresses that cannot (in general)
induce a deformation of the fluid and a term called the deviatoric stress tensor which
deals with stresses that tend to deform the fluid. Bearing in mind the above remarks,
we define the pressure term p in the general case as the average of the principal
stresses:
1
p = − trΣ. (A.11)
3
In this case we have
Σ = − pI + T, (A.12)
where T is the deviatoric stress tensor.
A.5 Mass Conservation
The principle of mass conservation expresses the fact that, when there are no mass
sources, matter is conserved during any thermodynamic transformation. Taking into
account the fluid motions, this means for example that, within a fixed volume Ω of
space with surface area Σ, the rate of change of the total mass is equal to the flux of
fluid components through the outer surface of the given volume. The volume of such
components crossing a surface element ds during a time dt is (u dt·n)ds. It follows
that
 ⎨
⎪ ⎪
d ⎧ ⎩
ρ dv = − (u · n)ρ ds. (A.13)
dt
Ω Σ
Since (A.13) holds for each volume element of the fluid sample, we deduce that
Fig. A.6 Main forces acting

on fluid components: volume
and surface forces
n
(Σ.n)ds
Ω ds
dv
Σ
g
∂ρ
+ ∇·(ρu) = 0. (A.14)
∂t
This is the local expression of mass conservation. For a gas, there may be very large
changes in the density, but for a liquid, and a fortiori a solid, the density varies rela-
tively little and a simplified form of (A.14) is often used in practice, namely ∇·u = 0.
This simplification is generally valid also for liquid–solid mixtures that are not in
two-phase flow (i.e., the average velocity of the solid does not differ from the aver-
age velocity of the liquid), since the density of each phase is roughly constant. The
situation for gas–solid mixtures is somewhat different. The behaviour of the mixture
and the density are often dictated by the solid phase, whereupon the presence of the
gas can often be ignored. As a consequence, the density of these mixtures may vary
significantly when the solid particles move closer together or move further apart,
whether this be due to migration through the gas or because the gas is compressed
or expanding.
A.6 Momentum Conservation
The principle of momentum conservation results directly from Newton’s second

law according to which any change in the motion of a body induces a force, and
conversely. More precisely, the acceleration of a rigid body is equal to the sum of the
forces acting on it divided by its mass. For a fluid, this principle applies to each of
its constitutive components. Consider a volume Ω of fluid subject to a body force of
density b and all the stresses exerted on the various fluid facets that can be identified
within this volume and on its outer surface (see Fig. A.6). In this case, the sum of
external forces acting on the various components of this volume is

Fext = bΩi + Σ · n ds. (A.15)
i facets
In the second term on the right-hand side, the stresses in the fluid which act on either
side of a facet lying entirely within the given volume will of course cancel out in
pairs, leaving only those stresses acting on the boundary surface Σ of the volume.
To calculate the acceleration of the given volume element, one must take into
account the fact that the velocity u is not necessarily the velocity of the same fluid
element from one moment to the next since it is the velocity associated with a given
position at a given time. The acceleration of a fluid element is in fact
u(x + dx, t + dt) − u (x, t)

a(x, t) = lim , (A.16)
dt∞0 dt
where dx = u dt is the path travelled by the fluid element located at x at time t during
the time interval dt. In this way we can follow a given fluid element, with the result
∂u
a(x, t) = + (u · ∇)u. (A.17)
∂t
Summing the expressions of Newton’s second law for each component within the
given volume and using (A.15) and (A.17), we find eventually
⎪ ⎪ ⎪ ⎪
∂(ρu)
dv + ρu(u · n) ds = b dv + (Σ · n) ds. (A.18)
∂t
Ω Σ Ω Σ
This expression is valid for any fluid volume. We thus deduce the local form of the
principle of momentum conservation:
∂u
ρ + ρ(u · ∇)u = b + ∇ · Σ. (A.19)
∂t
A.7 Temporal Fluctuation
Since fluids comprise a disordered assemblage of components which may in principle

have many different forms, the changes in the variables associated with volume
elements will not be regular, but will fluctuate around average values. In practice,
however, these values are measured over finite periods of time so the fluctuations
will not generally be observed. For example, the instantaneous velocity at a given
point can be expressed in the form
u = u + u , (A.20)
where u is the time-averaged value of the velocity and u the fluctuation. We assume
that u is the value measured by the experimenter if the measurement takes a finite
time to carry out, longer than the minimum time θ∗ required for the average of the
variable over this period to be very close to the average over a much longer time, but
shorter than a time beyond which significant changes would occur due to the flow.
For a flow in which the characteristic time for temporal evolution of the average
velocity is α, we may adopt the following definition:
⎪θ
1
u = u(θ) = u dt, (A.21)
θ
0
which is constant for θ∗ < θ < α. Averages and fluctuations of other variables can
be defined in a similar way.
The equations for mass and momentum conservation are valid in principle only
when we introduce instantaneous values for the variables. The corresponding expres-
sions for the variables are found by time-averaging these equations over a suitable
length of time. During this operation, all the terms proportional to one variable or to
two variables whose fluctuations are assumed to be uncorrelated give similar terms
involving the average value of the variable. On the other hand, in the momentum
conservation equation there is a term which is not linear in the velocity or one of its
derivatives. This is the term ρ(u · ∇)u in (A.19). Its average is not the product of the
average of each factor. Indeed, we have
(u · ∇)u = (u · ∇)u + (u · ∇)u . (A.22)
Under suitable assumptions (the system is ergodic and the time-averaging opera-

tion commutes with the differentiation ⎣ it can be shown that ρ(u · ∇)u
⎢ operators),

can also be written in the form ∇ · ρu ⇒ u . We then observe that, in terms of
the average values of the variables (denoted hereafter without the bar to simplify),
the momentum conservation equation can be written in the same form as the local
expression, provided that we use the modified stress tensor defined by
Ξ = Σ − ρu ⇒ u . (A.23)
It should be remembered that the variables are understood here as average values over
representative volume elements. The tensor Ξ is a case in point. Quite generally, the
definition (A.23) allows one to include in the stress tensor a term due to momentum
transport through the fluid. A single expression can thus express the resistance to
flow due to friction or contacts between fluid components which give rise to the stress
tensor Σ, and at the same time the resistance due to turbulence in the flow, thermal
agitation (see Sect. 2.2.2), or Brownian motion (see Sect. 5.2) of these components.
Note finally that a pressure term due to velocity fluctuations, viz.,
1 ⎢ ⎣
tr ρu ⇒ u ,
3
can be included in (A.23).
A.8 Turbulence
When the speed of flow of a fluid is gradually increased, we find that the temporal
fluctuations in the local values of the various variables around the average become
ever greater until they eventually predominate. This is the phenomenon known as
turbulence. Since in particular the local velocities fluctuate in all directions around
the average velocity, this tends to accelerate diffusion within the fluid. Turbulence is
then easy to observe with the help of this effect. A dye injected into a turbulent flow
will diffuse much more quickly in directions perpendicular to the direction of the
average velocity of the fluid. Turbulence occurs when the inertia of the fluid com-
ponents becomes too great in comparison with the viscous energy dissipation due
to the average motion. The phenomenon can then be quantified using a dimension-
less number, the Reynolds number, calculated from the ratio of the corresponding
energies. A general expression for the Reynolds numbers is ρV 2 /τ , where V is the
average flow speed and τ a typical stress value within the system. Note that the limit-
ing value of this number, beyond which turbulence becomes significant, depends on
the initial conditions and boundary conditions of the flow, because this phenomenon
has a distinctly nonlocal nature.
The problem with turbulence is that, in the present state of knowledge, there is no
direct way to determine the evolution of the fluctuation terms in the velocity in terms
of the other (average) variables. In particular, due to the nonlocal nature of turbulence,
it is no longer possible to determine any counterpart of the constitutive law of the
fluid in the turbulent regime by means of independent experiments and to predict
the flow characteristics under other conditions. The ‘state’ of turbulence cannot be
dissociated from the flow characteristics. In the general case, in order to model a
flow in a new geometry, the second term of Ξ must be estimated using relatively
complex procedures. The constitutive laws of materials like those discussed in this
book correspond to laminar flow, i.e., scenarios in which turbulence can be neglected.
A.9 Solution of the Flow Problem
In addition to the above equations, one must also know sufficient initial conditions
and boundary conditions for the problem to be well posed and possess a unique
solution from the mathematical standpoint. Simplifying somewhat, one must know
at each point a total of three components among the components of the velocity or
the stress tensor, taken relative to independent axes. In practice, this is often the case.
For example, we know, or assume given, the flow rate upstream of a free surface flow
and also the geometry in which the flow takes place (the velocity is therefore zero
along the walls). Likewise for a flow in a duct, the flow rate or pressure difference
between the ends of the duct is given and we assume in general that the velocity
vanishes on the walls.
We now consider a flow problem at constant temperature, for which the temporal
fluctuations in the velocity are negligible and the initial conditions and boundary
conditions are given. In this case, we may assume that the motion obeys (A.14) and
(A.19). These constitute four equations in the ten unknowns ρ, u, v, w, σx x , σx y ,
σx z , σ yy , σ yz , and σzz . It is not therefore possible in principle to solve any flow
problem without further input, i.e., without further relations between these variables.
The relations in question are known as the constitutive law of the material.
When there are no temperature variations, the constitutive law is often given in
the form of six expressions specifying Σ as a function of ρ and u. Here we shall be
mainly concerned with the characteristics of the constitutive law when there is no
heat exchange. The problem is of the same kind, although more complex, when we
consider flows in which temperature variations cannot be neglected. New variables
T , e, and q then come in and one requires the principle of energy conservation to
express their dynamics (see Appendix B). These must also be related together and
with the other variables by the constitutive law. When the temporal fluctuations in
the velocity due to internal agitation of the fluid components cannot be neglected,
the extra tensor in (A.23) can be determined directly as a function of the local state
of the material and the flow characteristics. In this case, (A.23) is the constitutive
law of the material. This is no longer true when the temporal fluctuations are due to
turbulence (see Sect. A.8).
When the constitutive law of the material is known, there is then no theoretical
reason why the flow problem should not be solved. In practice, however, the relevant
system of equations is very complex and even a numerical solution can be very
difficult to find. It is often necessary to simplify the problem by assuming that certain
phenomena or characteristics of the flow play only a minor role.
A.10 Constitutive Laws
A.10.1 General Considerations
When heat exchange can be neglected, the constitutive law of a material as we

have identified it is a relation between the stresses and the fluid motion. From the
physical standpoint, we thus need to determine the forces between two groups of
fluid components located on either side of a surface as a function of the motion of
these groups. These forces will depend on the motion of the two groups, but also on
the history of their motion which may have influenced their state. This notion of state
generally plays no role when we are dealing with an ordinary liquid under typical
flow conditions, and in particular, when considering sufficiently long periods of time
compared with the characteristic time of thermal agitation, for then the molecules
will forget their prior movements almost instantaneously, they are in the same state
of thermal agitation, and they interact among themselves in a similar way.
On the other hand, the state of the system may play a crucial role for materials
in which the components are much bigger than the molecules of an ordinary liq-
uid, exerting interactions on one another that vary significantly depending on prior
deformations or relative positions. This state can in principle be characterised by a
local variable. Note that for some kinds of material it may be preferable to use a
tensor rather than a scalar. But whatever the situation, given the short range of the
mutual actions of the different fluid components, the stresses on a small fluid volume
will depend in particular on the history of the motion in the close vicinity of this
volume. This is the principle of local action.
Many different kinds of constitutive laws have been proposed in theory, some very
simple and some very complex. Among the latter, very few have been confirmed by
experiment. In practice, it seems more useful to refer to a simple and physically
reasonable scheme which can include the main laws actually observed. To this end,
we assume that the local stresses at a point depend only on the state of the system and
the deformations (or strains) at the given time in a representative volume element
containing this point. We shall thus require mathematical tools for quantifying the
deformations.
A.10.2 Strain Rate Tensor
The key feature of the components of a fluid is their ability to move relative to one
another. When the fluid flows and there is no slipping on the walls, mass movement of
the whole fluid is impossible and it is compelled to deform. In this case, the velocity
differs from one point to another, i.e., the velocity gradient tensor ∇u is nonzero.
This tensor contains terms associated with elongation of the material. These are
terms on the diagonal, viz., ∂u/∂x, ∂v/∂ y, and ∂w/∂z. For example, when a fluid
is subjected to a flow such that only these terms are nonzero, and in addition they
are actually constant, this implies that each velocity component is proportional to
the distance of the given point from the origin. As a consequence, a fluid sample
undergoes steady extension or contraction in the various directions as time goes by.
In contrast, the off-diagonal terms of the velocity gradient tensor are associated with
shear. If for example only the term ∂u/∂ y is nonzero, the velocity only varies in
the y direction and fluid layers parallel to the plane O x z slip over one another. The
material is thus sheared in the y direction. However, if ∂v/∂x is also nonzero and
equal to −∂u/∂ y, the fluid motion is in fact a rotation of all the fluid components
about a fixed point. This type of mass movement does not in principle give rise to any
particular interactions between neighbouring components and should not therefore
affect the stresses within the fluid.
When we study the constitutive law, it is thus natural to focus on the strain rate
tensor D, defined as the symmetric part of ∇u, which only retains terms correspond-
ing to the relative motions due to shear:
1⎢ ⎣
D= ∇u + ∇uT . (A.24)
2
The components of this tensor are not affected by rotations since, any time two
terms symmetric relative to the diagonal of ∇u have opposite values, as in the above
example, the term corresponding to them in D is equal to their sum and hence
vanishes. Note that, when the fluid density is constant, (A.14) implies that
tr D = ∇ · u = 0. (A.25)
A.10.3 Simplified Form of the Constitutive Law
Finally, taking into account the various simplifications made up to now, the consti-
tutive law of a material should express the stress tensor at each point as a function
of the strain rate tensor and the instantaneous state of the fluid. The latter could also
be expressed using a tensor that would take into account any anisotropy in this state,
but it usually suffices to describe it with a single variable λ. Then we have
Σ = f (D, λ). (A.26)
To complete the constitutive law, a further equation must be adjoined to (A.26) to

describe how λ depends on the flow history. The latter thus comes in through the
state the fluid has reached at the given time.
In fact, in order to determine or at least characterise the constitutive law of a
material experimentally, one must in principle measure the velocity field and the
stress field within the material during a flow. In practice, it is often only possible to
measure on the one hand an average flow velocity in a cross-section or the angular
velocity of a given surface, and on the other, the torque on a device, a pressure
difference, or a force applied to the fluid as a whole. Under such conditions, we
cannot deduce the tensor form of (A.26) directly. It is therefore particularly useful to
arrange for conditions such that the form of the tensor D is as simple as possible. In
this way, the number of relations between the components of D and the components
of Σ that actually need to be measured can be drastically reduced. It is precisely the
aim of rheometry (see Chap. 8) to realise simple flows of this kind, in which the fluid
deformations are controlled in the best possible way by varying them over as wide
a range as possible. In the following we shall consider two kinds of simple flow:
simple shear, which is a viscosimetric flow, and elongation, which is not.
A.10.4 Simple Shear
We are often interested in the situation known as simple shear, for which there is a
frame of reference in which the tensor D has the form
 ⎨
0 1 0
γ̇ ⎧
D= 1 0 0⎩, (A.27)
2 0 0 0
with γ̇ the shear rate. This kind of expression is obtained for example for the perma-
nent shear of a fluid between two parallel planes in relative translational motion (see
Fig. 1.5) at speed V . Indeed, when there are no other external forces on the fluid (and
in particular, gravity can be neglected), it is easy to show, by expressing the balance
of forces on parallelepiped test volumes with two faces parallel to the slip plane, that
the shear stress is constant in all planes parallel to the solid surfaces. In addition, by
symmetry, the velocity cannot have a nonzero component in any other direction than
the relative velocity of the planes. There is no particular reason why this velocity
should vary within a given plane, so the fluid layers associated with these planes slip
over one another in the direction parallel to the planes. Finally, as the stress tensor
does not vary from one plane to another, the strain rate tensor will also be constant,
since these two tensors are related by (A.26). The only nonzero term in D, namely
∂u/∂ y, is constant between the two solid surfaces, and we thus deduce the following
velocity distribution
u = γ̇ y, v = 0 = w. (A.28)
Moreover, according to (A.28), the shear rate is γ̇ = V /H , where H is the thickness

of the sheared fluid. It can also be shown that the shear stress along each plane is
given by τ = F/A, where F is the force applied to the planes in the direction of
motion and A the sheared area.
When D can be written in the form (A.27), it can be shown using the principles
of rational mechanics that, in the same frame, the stress tensor has the form
 ⎨
σ11 σ12 0
Σ = ⎧ σ12 σ22 0 ⎩ . (A.29)
0 0 σ33
The pressure is related to the density by an equation of state (especially for gases) or
else we assume that the density is constant, whence the pressure can be treated as an
unknown variable instead of ρ. Under these conditions, assuming that the pressure
p = (σ11 + σ22 + σ33 )/3 has been determined independently, the stress tensor only
involves three unknowns. We then define the shear stress
τ = σ12 , (A.30)
and the first and second differences between the normal stresses
N1 = σ11 − σ22 , N1 = σ22 − σ33 . (A.31)
The apparent viscosity η is defined by

τ
η= . (A.32)
γ̇
For this simple shear, we thus have conditions such that the different variables char-
acterising the stress tensor, viz., τ , N1 , and N2 , are functions of γ̇ alone, which
for its part completely determines the tensor D. It is now easy to understand the
underlying aims of rheometry, which reduces the search for a tensorial expres-
sion like (A.26) to finding three relations between scalars. In this particular case,
we can in principle fully determine the form of (A.26) by carrying out suitable
measurements of tangential and normal forces. The main flow geometries provid-
ing a way to carry out such measurements are described in detail in Chap. 8. It
is important to note that this simplifying approach can conceal a certain number
of problems. For example, the simplest general tensor relation (A.26) is usually
inferred from the expression determined under simple shear conditions, when in
fact several tensor relations compatible with this expression may in principle be
possible.
A.10.5 Elongation
Elongational flows are sometimes also used to characterise the behaviour of a fluid.
In this case, all the tangential components of the strain rate tensor (i.e., those related
to a shear) are zero, and in an appropriate frame D can be expressed in the form
 ⎨
b 0 0
D = ⎧0 c 0⎩, (A.33)
0 0 d
where b, c, and d are three constants such that b + c + d = 0 according to (A.14).

For a simple uniaxial elongation, the velocity profile is
a a
u = ax, v = − y, w = − z, (A.34)
2 2
where a is a constant. This type of flow is obtained if we succeed in stretching a cylin-
der of fluid at exponential speed. We than have b = a and c = −a/2 = d. Moreover,
the stress tensor has no nonzero tangential components and the elongational viscosity
is by definition
N1 (a)
ηE = . (A.35)
a
Note that a simple shear can also be described as the composition of a rota-
tional motion at speed γ̇/2 [according to the decomposition of the velocity gra-
dient tensor in (A.24)] and a plane elongation at rate γ̇/2 in the direction x + y,
since
γ̇ γ̇
D·(x + y) = (x + y), D·(x − y) = − (x − y).
2 2
A.10.6 Energy Dissipation
If we know the forces and deformations within a fluid, we can calculate the power
dissipated by the internal forces in a given volume Ω. This is given by

P= dFi ·ui , (A.36)
i
where i labels a contact between two contiguous surface elements of area dσ and nor-
mal ni in the volume Ω and ui is the relative velocity of these surfaces. By definition,
the local force can be written in the form dFi = Σ · ni dσ and the relative velocity
is ui = ∇u · dxi , where dxi is the vector joining the centres of the fluid elements
on either side of the given surface. The power dP dissipated within a representa-
tive volume element around a given point can be calculated by summing the powers
dissipated over surfaces with normals oriented in the three possible directions i, j,
and k. This gives dP = tr (Σ · D) dΩe . From this we deduce the power dissipated
within an arbitrary volume in the form
⎪
P= tr (Σ · D) dv. (A.37)
Ω
A similar result is obtained by a more rigorous method in which the principle of

energy conservation is written out in detail (see Sect. B.1). This expression takes
on a particular meaning when we study constitutive laws since it represents energy
dissipation due to the resistance to relative motion of the fluid components, so it does
indeed correspond to viscous dissipation.
A.10.7 Main Types of Behaviour

The constitutive law of a Newtonian fluid has the general form
Σ = − pI + 2μD, (A.38)
where μ is the (constant) viscosity of the fluid.

The constitutive law for a simple (non-thixotropic) yield stress fluid in the steady
state can now be written quite generally as

−T < τc √⇒ D = 0,
D (A.39)
−T > τc =⇒ Σ = − pI + τc + F(D)D,
−D
where F is a positive function of the second invariant D. The Bingham model
corresponds to F = 2μB , where μB is called the plastic viscosity. The Herschel–
Bulkley model corresponds to
2n k
F(D) = 1−n ,
−D
where k and n are two parameters associated with the material. Under simple shear,
the constitutive law (A.39) has the form
τ < τc =⇒ γ̇ = 0,
(A.40)
τ > τc =⇒ τ = τc + f (γ̇),
where f is a positive function of γ̇. With the Herschel–Bulkley model, we have

f (γ̇) = K γ̇ n . We also sometimes use the Casson model, for which

f (γ̇) = K γ̇ + 2 K τc γ̇.
Note that the above expression (A.39) neglects possible elastic and viscous effects
in the solid regime, and elastic effects in the liquid regime. These aspects may nev-
ertheless play a significant role in the fluid behavior, in particular for transient flows.
Further Reading
1. Batchelor, G.K.: An Introduction to Fluid Dynamics. Cambridge University Press,

Cambridge (1967)
2. Truesdell, C.: A First Course in Rational Mechanics. Academic Press, San Diego
(1991)
3. Coleman, B.D., Markowitz, H., Noll, W.: Viscometric Flows of Non-Newtonian
Fluids, Springer Tracts in Natural Philosophy, vol. 5. Springer, Berlin (1966)
4. Piau, J.M.: Fluides non-newtoniens, Techniques de l’Ingénieur A710, 1–16,
A711, 1–24 (1979)
5. Ferguson, J., Kemblowski, Z.: Applied Fluid Rheology. Elsevier, Amsterdam

(1991)
6. Tanner, R.I.: Engineering Rheology. Clarendon Press, Oxford (1988)
Appendix B
Elements of Thermodynamics
B.1 First Law
In thermodynamics, we assume that the total energy of a system can be written as

the sum of the macroscopic kinetic energy K associated with average motions of
the components of the system with respect to some given frame, and its microscopic
internal energy. This internal energy U is the sum of the potential energies of inter-
action between the constituents and their kinetic energy on the local scale. U is a
function of the state of the system, but since it is difficult to describe this state in
detail, we generally describe only the changes in this function during transformations
of the system.
The first law of thermodynamics postulates that the change in the total energy
during a transformation is the sum of the heat supplied to the system and the work
done by external forces on the system. The heat Q is the energy associated with a
rise in temperature or a change of phase (and hence state) of the system components.
The work done W is the energy transmitted by application of a force when it gives
rise to a displacement of the macroscopic elements of the system. The power is the
rate of change of the work per unit time.
The first law can be written in the form
dU dK δQ
+ = Pe + , (B.1)
dt dt δt
with
⎪ ⎪ ⎪ ⎪
1
U= ρe dv, K = ρu dv, Pe =
2
ρb · u dv + (β · n)·u ds,
2
Ω Ω Ω Σ

DOI: 10.1007/978-3-319-06148-1,
312 Appendix B: Elements of Thermodynamics
where e is the internal energy density and Pe the power due to work by external
forces. The rate of flow of heat δ Q/ δ t into the volume Ω can also be written in
the form
⎪ ⎪
δQ
= r dv − q · n ds, (B.2)
δt
Ω Σ
where r is the rate of heat production per unit volume, e.g., due to chemical reactions,
and q the local heat flux (heat per unit area and per unit time). We may now rewrite
(B.1) using these variables, with the result
⎪ ⎪ ⎪
d u2 ⎛
ρ e+ dv = (b · u + r ) dv + (β · n) ·u − q · n ds. (B.3)
dt 2
Ω Ω Σ
Now the kinetic energy theorem, which is a reformulation of the conservation of

momentum, tells us that the change in kinetic energy K is equal to the sum of the
work done by external forces and the work done by internal forces, so
dK
= Pi + Pe , (B.4)
dt
where
⎪
Pi (u) = − tr (Σ · D) dv (B.5)
Ω
is the power generated by the work done by the internal forces.

Equations (B.2), (B.3), (B.4), and (B.5) imply the following local form of the
principle of energy conservation:
de
ρ = tr (Σ · D) + r − ∇ · q. (B.6)
dt
The left-hand side of (B.6) is the rate of change of the internal energy of the system.
The last two terms are related to heat transfer in the usual sense, i.e., resulting from
temperature gradients. The first term on the right-hand side plays an important role
when we study the flow of viscous fluids, since it corresponds to the power dissipated
by viscous friction per unit volume.
For a material at rest, the change in the kinetic energy of the system is zero and,
for a small transformation, the first law takes the form
dU = δ Q + δ W, (B.7)
where W is the work done by external forces.

Appendix B: Elements of Thermodynamics 313
For a simple fluid, and in particular for Newtonian fluids, undergoing a very slow
transformation, the quantity δ W can be expressed more precisely. Indeed, the term
associated with body forces in the power generated by external forces is zero because
the speeds are negligible. In addition, since the deformation rates are also very small,
only the pressure term in the constitutive law is significant, i.e., Σ = − pI, and the
power generated by external forces is therefore
⎪
dΩ
Pe = (− pI · n) ·u ds = − p ,
dt
Σ
whence we have δ W = − p dΩ. This no longer holds for complex fluids, and in
particular, yield stress fluids, for which anisotropic stress may remain when the
system is at rest or undergoing slow deformations.
B.2 Entropy
For a system in a given macroscopic state, the microscopic state, i.e., the distribution
of energy states of the microscopic components of the system, is not precisely defined.
In other words, many different microscopic states will be compatible with the given
macroscopic state. If the number of these microscopic states is Z , the entropy of the
system is defined by
S = kB ln Z . (B.8)
Given the equivalence of states for which similar components could exchange energy
levels, statistical information suffices to describe the changes in the entropy of the
system. Note in passing that the entropy is additive. This follows directly from the
above expression as a logarithm of the number of states. Consider two systems A and
B, with entropies S A and S B , respectively, and involving Z A and Z B microscopic
states, respectively. The total system A + B has Z A Z B possible microscopic states,
so it has entropy
S = kB ln(Z A Z B ) = kB ln Z A + kB ln Z B = S A + S B .
When the states are defined by a probability distribution ψ(r ) such that the probability
of having a state between r and r + dr is equal to ψ(r ) dr , the number of states
between r and r + dr is nψ(r ) dr , where n is the total number of states. The entropy
of the system is therefore kB ln [nψ(r ) dr ], whence, up to addition of kB ln(n dr )
which may be considered constant, the entropy is given by
S = kB ln ψ(r ). (B.9)
314 Appendix B: Elements of Thermodynamics
B.3 Second Law
The temperature can be defined from the entropy and energy of the system:
⎝
1 ∂ S ⎝⎝
= . (B.10)
T ∂U ⎝Ω
It follows that, during a reversible infinitesimal transformation, we have dS = δ Q/T ,

where δ Q is the heat transfer to the system. Furthermore, in an isothermal system,
i.e., at constant temperature, we have the following expression for the entropy, up to
a constant:
Uisoth
Sisoth = . (B.11)
T
The second law of thermodynamics tells us that the entropy of a closed system
increases to a maximum when equilibrium is established. This implies in particular
that, for a system in contact with a single heat source, we have Q/T < 0.
B.4 Free Energy
The free energy F , also called the Helmholtz free energy, is a function of state,
changes of which correspond to the work that must be done to get the system from
the initial state to the final state by a reversible transformation at constant temperature.
So if we carry out an operation during which the infinitesimal work δ W is done to
the system at constant temperature, the resulting change in the free energy is
dF = δ W. (B.12)
Since dU = δ Q + δ W and dS = δ Q/T , it follows that dF = dU − T dS.

Consider now a system in contact with a thermostat which ensures that the tem-
perature is constant and the combined system plus thermostat is isolated. In this
situation, the total entropy Stot = S + Sisoth is maximal by the second law. The total
energy of this ensemble is the sum of the energies of the system and the thermostat,
so Utot = U + Uisoth . We deduce that the expression Stot = S + (Utot − U )/T is
maximal at equilibrium. Since the total energy of the combined system is constant,
this means that the free energy function F = U − T S is minimal, or put another
way, at constant temperature,
dF = dU − T dS = 0. (B.13)
Since the total differential of a function f (x, y) has the form

⎝ ⎝
∂ f ⎝⎝ ∂ f ⎝⎝
df = dx + dy,
∂x ⎝ y ∂ y ⎝x
Appendix B: Elements of Thermodynamics 315
we may deduce several useful relations from the above results. We note first that
dU = T dS − pdΩ, and also dS = dU/T + pdΩ/T , whence
⎝
∂ S ⎝⎝
p=T , (B.14)
∂Ω ⎝U
and since dF = dU − T dS − SdT = −SdT − pdΩ in the general case, we deduce

that
⎝
∂F ⎝⎝
p=− . (B.15)
∂Ω ⎝T
B.5 Energy Distribution
Consider a small part A of a large system A0 which has energy E 0 and which is
thermally isolated. We assume that the interaction energy between A and the rest of
the large system A is small enough to be able to consider that E 0 is the sum of the
energies E and E of each of the subsystems, so that E = E 0 − E. We seek the
probability p(E) that the small system is in a microscopic state of energy E ◦ E 0
when equilibrium is reached. The number of possible states in this situation results
from the degrees of freedom left to the rest of the system, i.e., the number Z (E 0 − E).
This implies that
1 1 E dS
ln p(E) = ln Z (E 0 − E) = S(E 0 − E) ≈ S(E 0 ) − . (B.16)
kB kB kB dE 0
Since we know by definition that dS/dE 0 = 1/T , we find

E
p(E) = α exp − . (B.17)
kB T
This is known as the Boltzmann distribution. The constant α is determined by the

normalisation condition, i.e., the sum of the probabilities of occupying the different
available energy levels must be equal to 1.
In the above discussion, we assumed the existence of a discrete distribution of
energy levels. For a continuous distribution of energy states E, we must introduce
a probability density. The probability that the system is in one of its energy states
between E and E + dE is then given by

E
p(E) dE = α exp − dE. (B.18)
kB T
Index
A flocculation, 178
Adhesion energy, 41, 85, 203 pasty–hydrodynamic transition, 198
Amphiphilic molecule, 214 reaction limited aggregation, 191
Anisotropy effects, 104, 109 repulsive system, 158, 179, 180, 190
Apparent viscosity, 53 concentrated pasty regime, 185, 188
Archimedes’ principle, 90, 93, 116 concentrated regime, 183, 184
dilute regime, 184
glassy regime, 184, 185
B liquid regime, 188, 190
Bagnold number, 242 semi-dilute regime, 184
Bancroft’s rule, 216 sedimentation, 165, 167
Body forces, 292 shear thinning, 197
Born repulsion, 40 solid–liquid transition, 187
Brittleness, 3, 72, 74 stability, 16, 173, 175, 178, 179
Brownian diffusion, 162 thermal agitation, 157–160
Brownian motion, 18 thixotropy, 158, 198
in colloids, 158, 159, 167, 181, 183 van der Waals forces, 167, 169
Buoyancy, 90, 93, 116 yield stress, 189, 194, 196–198
Colloidal interaction, 13, 16, 180, 198
Compressibility, 71
C Compression modulus, 71
Capillary number, 211 Concentration, 12
Chemical bonds, 40 critical, 88, 100, 103, 109
Coalescence, 26, 202, 212, 213, 216 effects, 97, 103
Cohesive energy, 65, 203 non-uniform, 109, 111–113
Colloid, 15, 21, 157, 199 of polymers, 136
aggregation, 158, 190–192 Concentration regimes
attractive system, 158, 179, 190, 198 in colloids, 180, 199
concentrated pasty regime, 194, 195 in emulsions and foams, 220, 230
concentrated regime, 192 in polymers, 137, 140, 147, 155
dilute regime, 194 in suspensions, 99, 103
liquid regime, 195, 198 Concentric cylinder rheometer, 265, 267
semi-dilute regime, 194 Cone–plate rheometer, 265
depletion interactions, 176, 177 Configuration effects, 15, 238, 241
diffusion limited aggregation, 191 Conformation, 122
elastic modulus, 194, 195 Constitutive law, 302–308
electrostatic forces, 170, 172 Newtonian fluid, 307

DOI: 10.1007/978-3-319-06148-1,
318 Index
simplified, 304 solid regime, 226–228

yield stress fluid, 308 stability, 212, 218
Continuous medium hypothesis, 82, 83, 95, van der Waals forces, 214
96, 291, 292 yield stress, 26, 28, 228, 229
Coulomb criterion, 247 Entropy, 42, 46, 48, 313
Coulomb model, 31 of polymer chain, 21
Cox–Merz law, 152 Evaporation, 66
Creaming, 94, 165 Eyring model, 63, 64
in emulsions, 207
Cross-linking, 142, 144
F
First law of thermodynamics, 311–313
D Floc, 192, 193, 195
Debye length, 172, 178 Flocculation, 178
Density, 84, 85, 290 Flory’s theorem, 140
Depletion interaction, 176, 177 Flow curve, 264, 265, 274, 275, 279, 281,
Deviatoric stress tensor, 297 286, 288
Diffusion, 162, 165 decreasing, 276, 277
rotational, 164 Foam, 29, 30, 201, 230
Diffusion coefficient, 162 compact regime, 222
Dilatancy, 100, 239, 240 continuous phase, 201
Dissipation, 55, 307 dispersed phase, 201
Double layer, 170, 172 dry, 222
Drag, 90 of emulsion, 212, 218
in yield stress fluid, 117 stability, 212, 218
on droplet or bubble, 206 Fractal, 192, 193, 195
on particle, 89, 91 Fracture, 72
torque, 91 brittle, 5, 74
Duct, flow in, 268, 269 ductile, 5
Ductility, 3, 72, 73 Free energy, 314–315
of polymer configurations, 127, 132, 133
of polymer interactions, 133, 135
E Friction coefficient, 242, 244, 252, 254, 255,
Elastic regime, 61 258
Elasticity, 3, 5, 67, 194, 195 effective, 256
of polymers, 21, 22, 143, 145, 146, 152–
155
Electrostatic forces, 170, 172 G
Elongation, 68, 69, 306–307 Gaseous state, 36, 42, 43, 55
Emulsification, 208, 215 Gel, 195
Emulsifier, 216 Glass, 4
Emulsion, 26, 28, 201, 230 transition, 76, 79
adhesion energy, 203 temperature, 76
coalescence, 202, 212, 213, 216 viscosity, 79
cohesive energy, 203 Glassy state, 37, 43, 76, 79
concentrated regime, 221, 225, 230 Granular materials, 30, 33, 231, 258
continuous phase, 201 Bagnold number, 242
creaming, 207 collisional regime, 242, 251, 253
dilute regime, 221, 222, 224 collisions, 236, 238
dispersed phase, 201 configuration, 238, 241
liquid regime, 226, 229 Coulomb criterion, 247
ripening, 202, 204, 212, 216, 218 critical state, 246–248
sedimentation, 207 dilatancy, 239, 240
semi-dilute regime, 221, 224, 225 frictional contacts, 235, 236
Index 319
frictional regime, 242, 243, 251 of sublimation, 75

internal friction angle, 243 Leighton number, 242
jamming, 238, 240 Lennard-Jones potential, 40, 41
Leighton number, 242 Liquid, 5, 10
lubricated contacts, 233, 235 interstitial, 82
lubrication regime, 242 polar, 170, 172
maximum rest angle, 250, 251 Liquid state, 36, 42, 55, 66
restitution coefficient, 238
roughness, 232, 235, 244
settling, 240 M
solid fraction, 240 Mass conservation, 297–298
Stokes number, 242 Maximum packing fraction, 87–89, 98, 111
surface roughness, 234 Mean free path, 46
Migration, 109, 110, 274
and slipping, 113
H induced concentration effect, 112
Heat, 311 under flow, 111, 112
Herschel–Bulkley model, 116, 117, 189, under simple shear, 110
229, 308 Momentum conservation, 298–299
Hydrodynamic interaction, 82, 94, 99–101, Monomer, 121
103, 108 Monomer volume fraction, 136
Hydrogen bond, 41
Hydrophilic–lipophilic balance (HLB), 215
Hydrostatic pressure distribution, 93, 116 N
Newtonian fluid, 6, 55, 63, 65
constitutive law, 307
I Normal stress, 294
Ideal gas, 38, 47–49
Inclined plane, 286, 288
Instability of flow, 274, 276–277 O
Interaction potential, 38 Orientation effects, 13, 14, 96, 104, 107
Interface, 202, 203 Osmotic pressure, 165
pressure difference across, 203
Interfacial tension, 66, 203
Internal energy, 311 P
Interstitial liquid, 82 Parallel disk rheometer, 263, 265
Paste, 115, 117, 118
stability, 116, 117
J Péclet number, 181
Jamming, 100, 103 Percolation threshold, 11
in colloids, 185 Phase separation, 10
in granular materials, 238, 240 Plastic, 21
Plasticity, 5, 72
Poisson coefficient, 71
K Polymer, 21, 26, 121, 156
Kinetic theory, 50, 55 adsorption, 173, 175
Krieger–Dougherty model, 103, 109, 111, apparent chain length, 123, 126
118, 181, 225 chain extension, 127, 128
concentrated regime, 136, 139, 140, 150,
154
L conformation, 122
Laminar flow, 9 correlation length, 138
Latent heat cross-linked, 142, 144
of evaporation, 66 dilute regime, 136, 137, 147, 150
320 Index
effective volume fraction, 136 Settling, 240

elastic modulus, 143, 145, 146, 152–155 Shear bands, 274, 276
entanglement, 140, 141 Shear box, 244
free energy, 127, 132, 136 Shear modulus, 69–71, 195
gel, 142 in emulsion, 227
good solvent, 136 Shear rate, 52
in solution, 23, 132, 136 Shear stress, 6, 53, 294
interactions with solvent, 132 Shear thickening, 15, 113, 115
mass fraction, 136 Shear thinning, 22
melt, 23 in colloids, 197
persistence length, 129, 132 in polymer solutions, 146
persistence time, 132 Silicone oil, 23
poor solvent, 136 Simple shear, 7, 8, 52, 305–306
radius of gyration, 126, 127, 175 for rheometry, 261
relaxation modulus, 149, 156 Slip on walls, 271, 273
relaxation time, 146, 148, 154 Solid, 2, 5
repeat unit, 121 Solid fraction, 12, 89, 97, 100, 101, 104, 111
reptation, 25, 153, 154 in granular materials, 240
segment, 129 non-uniform, 109, 111, 113
semi-dilute regime, 136, 138, 139, 154 Solid state, 36, 42, 66, 75
temperature effects, 155, 156 Squeeze test, 282, 286
theta solvent, 135 Stability
viscoelasticity, 23, 25, 145, 146, 148, of colloid, 16, 173, 175, 178, 179
150, 151, 154 of emulsion, 212, 218
viscous modulus, 146, 150 of paste, 116, 117
Power, 311 of suspension, 92, 95
Pressure, 48, 49 Stokes number, 242
hydrostatic distribution, 93, 116 Strain rate tensor, 303–304
in kinetic theory, 50, 52 Stress tensor, 296
invariants, 297
Sublimation, 75
R
Surface forces, 293
Relaxation time, 25, 78
Surface tension, 66
liquid, 61
Surfactant, 18, 26, 207, 212, 214–216, 219
polymer, 146, 148, 154
Bancroft’s rule, 216
Representative volume element, 83, 84
Suspension, 10, 15, 81, 118
Reptation, 25, 153, 154
concentrated regime, 100
Restitution coefficient, 238
dilute regime, 99–101, 108, 117
Rheometer
in yield stress fluid, 115, 118
concentric cylinder, 265, 267
cone–plate, 265 semi-dilute regime, 99, 108, 118
Couette, 265 stability of, 92, 95
gap, 262 viscosity, 95–98, 101–105, 108, 109,
parallel disk, 263–265 112, 113, 117
Ripening, 202, 204, 212, 216, 218
Roughness, 232, 234, 235, 244, 271, 278
Rubber plateau, 151, 153 T
Temperature, 38, 49, 155
of glass transition, 76
S Thermal agitation, 36–38
Second law of thermodynamics, 314 in colloids, 157–160
Sedimentation, 94, 95 Thermodynamics, 311–315
in colloids, 165, 167 first law, 311–313
in emulsions, 207 second law, 314
Index 321
Thixotropy, 18, 21 Viscous dissipation, 55, 307

in colloids, 158, 198 Volume forces, 292
Torque, 91 Vulcanisation, 25
Turbulence, 301
Turbulent flow, 9
W
Weissenberg effect, 26
V Wetting, 86
Van der Waals equation of state, 58 Work, 311
Van der Waals forces, 6, 16, 39
in colloids, 167, 169
in emulsions, 214 Y
Velocity gradient tensor, 303 Yield stress, 16, 18, 20, 275, 276, 281
Viscoelasticity, 23, 25, 78 colloid, 189, 194, 196–198
of polymer, 145, 146, 148, 150, 151, 154 emulsion, 26, 28, 228, 229
Viscosity, 6 Yield stress fluid, 16, 185, 226
apparent, 53 constitutive law, 308
of gas, 52, 55 drag on particle, 117
of glass, 79 rheometry, 262, 279, 281, 288
of liquid, 63, 65 solid–liquid transition, 281
of suspension, 95–98, 101–105, 108, suspension in, 115, 118
109, 112, 113, 117 Young’s modulus, 143

Rheo Physics

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Rheo Physics

Caricato da

Copyright:

Formati disponibili

Soft and Biological Matter

For further volumes:

ISSN 2213-1736 ISSN 2213-1744 (electronic)

Library of Congress Control Number: 2014941116

Springer International Publishing Switzerland 2014

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Many materials cannot be strictly categorised as solid or liquid, and this is

engineering, soil science, agronomy, chemical engineering, planetary science, and

understanding through a physical approach that emphasises the mechanisms

Paris, France, March 2014 Etienne Guyon

Our understanding of the physical or physicochemical processes underlying the

Paris, March 2014 Philippe Coussot

2 Simple Materials . . . . . . . . . . . . . . . . . . . . . . . . . ............ 35

2.2.8 Hydrogen Bond and Hydrophobic Forces . . . . . . . . . . . 41

3.4.2 Concentration Regimes. . . . . . . . . . . . . . . . . . . . . . . . 99

6 Emulsions and Foams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

7 Granular Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

Appendix A: Fluid Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

Appendix B: Elements of Thermodynamics . . . . . . . . . . . . . . . . . . . . 311

b Distance between the centres of two molecules or particles

/m Maximum packing fraction

P. Coussot, Rheophysics, Soft and Biological Matter, 1

through sprays of droplets in pesticide applications, to vapour condensation produc-

with particular types of organisation. It is also possible to predict the behaviour of

1.4.1 Phase Separation

1.4.2 Effect of Particles on Behaviour of the Mixture

Here we are specifically concerned with the effects of adding particles to

1.4.3 Other Effects

1.4.3.1 Size Effects

1.4.3.2 Orientation Effects

1.4.3.3 Configuration Effects

1.5.1 Colloidal Interactions

1.5.2 Yield Stress

1.6.1 Properties of Polymer Chains

To understand the properties of polymers, we must first examine the properties of

Fig. 1.15 a Representation of a polypropylene molecule in a barely stretched conformation.

1.6.2 Polymers in Solution

In various applications or when manufacturing certain materials, polymers are

1.6.4 Other Properties of Polymers

Other characteristic properties of polymer materials are observed in flow.

Fig. 1.18 a Internal structure of a foam. b Shaving foam. c Chocolate mousse

1.9 Granular Materials

1.10 Real-Life Materials

In nature as in industry, materials rarely comprise a single type of element (polymer,

Here we shall be concerned with materials comprising a single type of elementary

P. Coussot, Rheophysics, Soft and Biological Matter, 35

2.2 Interactions Between Elementary Components

2.2.1 Elementary Components

2.2.2 Thermal Agitation

where T is called the temperature and kB = 1.38 × 10−23 J K−1 is Boltzmann’s

2.2.3 Interaction Potential

To describe the interaction forces between arbitrary objects such as molecules, it is

Differentiating this expression, we obtain

Fig. 2.1 Interaction force F

2.2.4 Van der Waals Forces

2.2.5 Chemical Bonds

Within a crystal structure, molecules give up something of their independent exis-

2.2.6 Born Repulsion Force

r < d =⇒ (r) = ∞, r > d =⇒ (r) = 0 . (2.3)

2.2.7 Balance of Forces

(a) (b) (c)

This model can be simplified by representing the repulsive part by a hard-sphere

where w is the maximum attractive potential, or adhesion potential, obtained when

When τ b3 2kB T , this expression simplifies to first order, yielding

F = (r) ≈ (b − r) (b) . (2.35)

f = Ψ (x) = 3εb (b) .

f = Ψ (x) = x (b) ,