Foams

Foams
This page intentionally left blank

Foams
Structure and Dynamics
I. Cantat, S. Cohen-Addad, F. Elias, F. Graner, R. Höhler,

O. Pitois, F. Rouyer, and A. Saint-Jalmes
Translated by R. Flatman
Scientific editor (English edition) S.J. Cox
3
3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
Original edition: Les mousses – Structure et dynamique
c Éditions Belin – Paris, 2010
English translation c Oxford University Press, 2013
Ouvrage publié avec le concours du Ministère français chargé de la culture – Centre national du livre
The moral rights of the authors have been asserted
First Edition published in 2013
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2013938927
ISBN 978–0–19–966289–0
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Preface
The bubbly world of foams

A foam consists of a large number of bubbles packed together (fig. 0.1), which gives
this state of matter its remarkable properties. We describe here liquid foams, which
consist of a gas and a liquid (often soapy water). They are in general opaque, remark-
ably stable, and even elastic, as is evident in whipped cream, shaving foam, and sea
foam.
A great deal of excellent work has been conducted on these mixtures of soapy water
and air: Boys’ captivating book (published in 1890), for example, which arose from
public lectures in which the author entertained the audience with his soap bubble
demonstrations, or the book of Mysels, Frankel, and Shinoda that elucidates some of
the many secrets of soap films, Bikerman’s book on industrial applications of foams,
and more recently the book of Weaire and Hutzler on the physics of foams, with its
focus on structural problems.
In this book we set out the fundamentals of current knowledge on liquid foams
and discuss some recent advances in our understanding of their structure, generation,
ageing, and rheology. We have sought to provide information at several different levels:
from simple, concrete explanations, perhaps for industrialists who find themselves
interested in foams, to an in-depth discussion of scaling laws and detailed mathematical
models, suitable for academics and students; all should, in addition, find the exercises
and suggested experiments straightforward to implement.
This is certainly not an exhaustive volume: we consider solid foams only briefly
and we avoid discussion of a number of unresolved questions. It is also the product
of team-work, so that in places we find ourselves explaining and comparing different
points of view, fundamental or applied, which is in itself a stimulating exercise. Such
polyphonic writing requires extensive proofreading and polishing: we thank in partic-
ular Christian Counillon, Agnès Haasser, and Michel Laguës for having done this and
for their excellent advice. We also thank with great pleasure all those colleagues and
students who have helped us make progress thanks to their remarks and suggestions.
Finally, we would like to acknowledge the CNRS-funded GDR “Mousses et Emulsions”
and its directors M. Adler and C. Gay for their encouragement.
vi Preface
Fig. 0.1: A liquid foam some time after its creation. It is dry at the top, with large
polyhedral bubbles, and wet at the bottom, with a smaller average bubble size. The
bubbles become spherical when they come into contact with the liquid on which the
foam floats. Photograph courtesy of S. Cohen-Addad, R.M. Guillermic, and A. Saint-
Jalmes. See Plate 1.
Preface vii
Experiments
The experiments presented at the end of each chapter aim to illustrate a number
of points within that chapter. Some experiments can be carried out in minutes in
a kitchen or bathroom, others require laboratory materials, but none are dangerous.
Nonetheless, it is necessary to take the usual safety precautions and use common sense.
At the beginning of each experiment, the theoretical prerequisites are indicated in the
form of references to the pertinent paragraphs in the chapter. The level of difficulty
and the time required for each experiment are classified according to four criteria:
Difficulty level: three levels of difficulty are identified for the experiment depending
on the required materials:
: Materials generally found in a kitchen or bathroom

: Materials that can be bought or made at home
: Laboratory materials
Cost: the cost of the experiment is divided into three levels of expense:
: cheap, less than about 5 dollars, euros, or pounds

: moderate
: expensive, more than about 30 dollars, euros, or pounds
Preparation time: this refers to the time required to set up the experiment once
the materials are assembled. Depending on the experiment this could take between
five minutes and half a day.
Experimental time: once the experiment is set up, this is the actual experi-
mental time, including the analysis, required to give a complete description of the
phenomenon.
Contents
1 Uses of foams 1
1 The foams around us 1
1.1 Foams in mythology 1
1.2 On your plate and in your glass 1
1.3 Detergents and cosmetics 3
1.4 Spontaneous or undesirable foams 4
2 Foam identification 4
2.1 Physico-chemical constituents 4
2.2 Geometrical and physical properties 5
2.3 Mechanical properties 6
3 What are foams used for? 6
3.1 Desirable functions 6
3.2 Mineral flotation 8
4 Solid foams and other cellular systems 9
4.1 Solid foams 9
4.2 Other cellular structures 11
5 Experiments 13
5.1 Three ways to make a foam 13
5.2 Chocolate mousse 14
References 15
2 Foams at equilibrium 17
1 Description at all length-scales 17
1.1 At the scale of a gas/liquid interface 17
1.2 At the scale of a film 19
1.3 At the scale of a bubble 21
1.4 At the scale of a foam 22
2 Local equilibrium laws 23
2.1 Equilibrium of fluid interfaces 23
2.2 Plateau’s laws 26
3 Dry foams 30
3.1 Number of neighbours: topology 31
3.2 Bubble geometry 35
3.3 Topology and geometry 38
4 Wet foams 45
4.1 Modification of the structure 46
4.2 Osmotic pressure 51
4.3 Role of gravity 54
x Contents
5 2D and quasi-2D foams 55

5.1 3D structure of a monolayer of bubbles between two plates 57
5.2 A model for a dry 2D foam 58
5.3 Two-dimensional liquid fraction 60
5.4 2D foam flows 61
6 Experiments 63
6.1 Surface tension and surfactants 63
6.2 Creation and observation of 2D and quasi-2D foams 65
6.3 Giant soap films 66
6.4 Kelvin cell 68
7 Exercises 69
7.1 Interfacial area of a foam 69
7.2 Film tension and the Young–Laplace law 69
7.3 Plateau’s laws in 2D 70
7.4 Euler’s formula 71
7.5 Perimeter of a regular 2D bubble 71
7.6 Energy and pressure 72
References 72
3 Birth, life, and death 75
1 Foam evolution 75
1.1 The competition between different processes 75
1.2 Elementary topological processes 78
2 Birth of a foam 82
2.1 Foamability: introduction to the role of surfactants 82
2.2 Interfacial properties and foamability 82
2.3 Properties of liquid films and foamability 92
2.4 Summary of the microscopic origins of foamability 98
3 Coarsening 99
3.1 Growth rate of a bubble in a dry foam 99
3.2 Evolution of bubble distributions in a dry foam 104
3.3 Effects of different parameters 109
4 Drainage 113
4.1 What is drainage? 114
4.2 Free drainage 114
4.3 Forced drainage 115
4.4 Modelling flows in solid porous media 116
4.5 Modelling the permeability of a liquid foam 119
4.6 Drainage equations 127
4.7 Comparison of theoretical predictions with experiments 128
4.8 Summary and remarks 133
5 Rupture and coalescence 134
5.1 Rupture at the scale of a single film 134
5.2 Rupture at the scale of a foam 140
5.3 Defoamers and antifoams 140
Contents xi
6 Appendices 145
6.1 Stabilizing agents 145
6.2 Dissipation due to surfactant motion during the steady
expansion of a film 151
7 Experiments 154
7.1 Flow in a soap film 154
7.2 Free drainage in a foam and the vertical motion of bubbles 156
7.3 Forced drainage in a foam: observation of the wetting front 157
7.4 Life and death of a foam measured by electrical conductivity 158
8 Exercises 161
8.1 Exponent in the scale-invariant regime 161
8.2 Frumkin equation of state 161
8.3 Foam drainage and equilibrium height 161
8.4 Drainage in the bulk and at the wall 162
8.5 Free drainage: characteristic times and liquid fraction profiles 162
8.6 The true 3D pressure and 2D surface pressure 162
References 162
4 Rheology 167
1 Introduction 167
2 Overview of the rheological behaviour of complex fluids 168
2.1 Constitutive laws 168
2.2 Shear tests 172
2.3 Small and large strains 173
2.4 Stress tensor in a complex fluid 174
3 Local origin of rheological properties 178
3.1 Elastic shear modulus of a dry monodisperse foam 178
3.2 The elastic limit of a dry foam 183
3.3 Dissipative processes 187
4 The multiscale character of foam rheology 193
4.1 Solid behaviour 194
4.2 Transition from solid to liquid behaviour 208
4.3 Foam flow 211
5 Appendix: From the discrete to the continuous 215
6 Experiments 217
6.1 Observation of T1s 217
6.2 Visualization of the yield stress 217
7 Exercises 218
7.1 The Young–Laplace law and the stress in a spherical bubble 218
7.2 Elasticity of a dry 2D foam 219
7.3 Poynting’s law 219
7.4 Stress and strain in a square lattice 220
7.5 Elasticity and plasticity 220
7.6 Compressibility of a foam 221
References 221
xii Contents
5 Experimental and numerical methods 225

1 Experimental methods 225
1.1 Methods used to study interfaces and isolated films 225
1.2 Methods for studying foams 230
2 Numerical simulations 242
2.1 Predicting static structure 242
2.2 Predicting dynamics 244
3 Methods of image analysis 248
3.1 Image treatment 248
3.2 Image analysis 250
3.3 Image analysis, liquid fraction, and stress in 2D 254
4 Exercises 255
4.1 Measurement of the average liquid fraction of a foam 255
4.2 Pressure in the Potts model 255
References 256
Notation 259
Index 263
1
Uses of foams
Are foams really useful? We smile when scientists announce that they study foams, because
bubbles are considered child’s play, but in this first chapter we will show that the converse is
true: liquid foams exhibit a range of complex and unique properties, and are well-adapted to
diverse applications in our daily lives and to numerous industrial processes.
1 The foams around us

1.1 Foams in mythology
In the Mahabharata, Indra, having been captured by the demon Namuchi, is only freed
in exchange for the promise that he will not kill his kidnapper by day or night, with
something neither wet nor dry. Indra keeps his word: when he returns to slay the
demon (fig. 1.1), it is twilight and he uses foam! Neither wet nor dry, foam meets
Namuchi’s contradictory demands.
In Greek mythology, Cronos castrates his father Uranus and throws his genitals
into the sea. Hesiod recounts that “a white foam surrounded the immortal flesh, and
in it grew a girl. At first it touched on holy Cythera, from where it came to Cyprus,
circled by the waves. And there the goddess came forth . . . ” [6]. This goddess is named
Aphrodite, which translates literally as “born of the foam”. Aphron means foam in
Greek, so the subject of this book can be described as aphrology!
A Chinese myth describes how in order to thwart two dragons who claimed his
throne, the emperor collects the foam from their drooling jaws and locks it in a casket.
Although foams are not usually so stable, this foam remained in the casket for 2,000
years; when the casket is opened, the foam escapes and takes the form of a small
dragon. The dragon touches a young girl who, several years later, gives birth to Pao
Sze, the dragon princess renowned for her great beauty and sadness.
Although short of a full anthology, these examples show that foams, whether for
their unique properties or their (supposedly) superficial nature, appear frequently in
literature. A final example is Joyce’s Ulysses, where foam appears in diverse forms:
from the first sentence, in Buck Mulligan’s shaving bowl, to its association with the
barrels of beer flowing into the street.
1.2 On your plate and in your glass

Foam is, of course, associated with certain beverages. Beer foam, stabilized amongst
other things by proteins, appeals both to the eyes and to the palate. Champagne foam,
which is more fragile, is so distinctive that a discerning drinker is able to predict the
2 Uses of foams
Fig. 1.1: The properties of foam inspired Ganesh, the mythical author of the
Mahabharata, in the scene in which Indra attacks Namuchi.
c Amar Chitra Katha
Pvt. Ltd.
quality of the wine from the foam alone. In champagne and certain beers, carbon diox-
ide, which is a product of fermentation in the bottle, is concentrated by an overpressure
(6 bars for champagne). The liquid is supersaturated at ordinary pressure but degasses
when the bottle is opened; bubbles appear, rise, and form a foam. Some draught beers
are dispensed with air or nitrogen in order to slow the ageing of the foam. Fizzy drinks
and carbonated water are produced by adding carbon dioxide, but the resulting foams
are not very stable due to the lack of proteins. The long-lived froth on a cappucino
is stabilized by milk proteins. This leads us naturally to questions about foaming,
drainage, and ageing which are dealt with later in this book (chap. 3).
Not only are foams highly drinkable, but their consistency and lightness make them
a pleasure to eat as well. Chocolate or fruit mousses (fig. 1.2) immediately spring to
Fig. 1.2: Three-dimensional imaging experiment of the air bubbles in a chocolate

mousse. The technique used is X-ray tomography, which is described in chap. 5. Image
courtesy of P. Cloetens, ESRF.
The foams around us 3
mind, but air is also incorporated into ice cream, altering its texture and making it
easier to serve, but without causing it to melt too quickly in the mouth. Air, of course,
is a raw material which doesn’t cost anything . . . whilst making a large contribution to
the volume of the final product.
Edible foams are usually solid, obtained by solidifying a liquid foam either by refri-
geration (chocolate mousse) or cooking (bread). The foaming or stabilizing agents for
edible foams are natural, for example sugar (for meringue) or gelatine (for marshmal-
lows, where it also serves as a gelling agent). Industrially produced edible foams tend
to be stabilized by lecithin (from egg or soya) or xanthan gum, which makes them very
viscous. Some foams, like bread and cake, have no need of a stabilizing agent because
they are solidified before they collapse.
Baked alaska is an unusual combination of foams: ice cream is thermally stabilized
by two other foams: the ice cream is placed on a solid foam, the sponge cake, and
then covered by a liquid foam, an uncooked meringue, and placed in the oven. The ice
cream is sufficiently well-insulated that it doesn’t melt while the meringue is gently
cooked!
There are also situations where foams appear without being sought, as a con-
sequence of either boiling or simply agitation, as in the froth that is generated when
cooking pasta or potatoes in water, where the proteins present enable the bubbles to
persist. Stable milk foam can be seen when a baby is feeding from a bottle or when a
glass of milk is shaken, while that which floats on coffee or tea is very short-lived.
1.3 Detergents and cosmetics

The presence of soap in a cleaning product has a consequence beyond the required
property (removal of fat and solid particles): the appearance of a foam. It is widely
believed that the presence of foam is a measure of cleaning efficacy but in fact the op-
posite is true: manufacturers try to suppress foams in mechanical cleaning (in washing
machines and dishwashers). The addition of only a few drops of washing-up liquid to
a dishwasher results in the machine overflowing due to a large quantity of stable foam.
In order to avoid these problems, antifoam agents are added (cf. §5.3, chap. 3).
There are, however, advantages of using foams in some cases. For example, while
tests show that the cleaning performances of non-foaming and foaming oven cleaners
are hardly distinguishable, the latter will be desirable as the cleaner will cling to
the vertical walls of the oven for longer. The ability of a foam to stick to a vertical
surface, owing to its elasticity, also makes it useful for shaving; a small quantity of
product covers the cheeks and stays there until the motion of the razor blade renders
it sufficiently fluid to wash off. This leads to questions of rheology which will be
discussed later in the book (chap. 4).
Generally speaking, foam products are very common in cosmetics: their ability to
cover and cling to surfaces explains their use in hair gels or conditioners. When you
wash your hair, the shampoo generates more foam in the second wash, indicating that
the first application has succeeded in its role. But the shampoo foam is also there for
your enjoyment. Similarly, it is in their role of providing comfort and pleasure that
foaming bath products are successful.
4 Uses of foams
1.4 Spontaneous or undesirable foams

As we have seen, foams arise from the presence of a gas in a liquid, for example the
carbon dioxide in a fizzy drink or the alkanes in a shaving foam. A magma which
degasses causes bubbles to accumulate, and therefore a foam forms in the magma
chamber of a volcano. Depending on the conditions (gas flow rate, bubble size, cavity
shape), the foam may collapse gradually or, as in Strombolian volcanoes, suddenly
and spectacularly!
More common in nature is a foam that forms in a liquid that is agitated in such a
way to allow air to be incorporated, as in a waterfall or a breaking wave. These foams
are weak and transient if nothing stabilizes them, but in the presence of surfactants
from pollutants or plankton, they become the long-lasting “white horses” seen along
the coast where spectacular foam mountains can reach several metres high, as can
be found for example in Australia. So the presence of a foam is a good indicator of
contamination (industrial or natural) by surfactants.
In the same way, pouring washing-up liquid into a public fountain produces a
splendid display of foam, which can be made more spectacular still if a few drops
of food colouring are added. At the cinema, foams have a role in special effects, for
example representing snow. In the final scene of The Party by Blake Edwards, a house,
the scene of a lavish party, is overrun with soap bubbles and chaos ensues.
Undesirable foams are also encountered in the industrial world, where the focus is
on developing products to get rid of them, a subject to which we will return later (§5.3,
chap. 3). Such foams appear for example in settling tanks or during the manufacturing
of steel, glass, pulp, sugar, watercolour paints, or products obtained by fermentation
(like wine or penicillin). An over-acidic stomach secretes foam, and the recommended
medicines contain an antifoam that allows them to reach the stomach wall before
reacting.
2 Foam identification
As we have seen, liquid foams are commonly encountered all around us. We now
describe the properties which make these materials unique and potentially useful in
various industrial applications (see §3). Furthermore, a foam, under carefully controlled
conditions, often serves as a model system for a cellular material or a biological tissue.
2.1 Physico-chemical constituents

A foam is a dispersion of gas in a liquid. The tightly packed gas bubbles occupy most
of the volume (fig. 0.1). The liquid phase, which consists of the soap films and their
junctions, is continuous, unlike the gas phase. It contains special molecules, known
as surfactants, which stabilize the bubbles by arranging themselves at the liquid/gas
interfaces (see §2, chap. 3).
Different properties can be obtained by changing each constituent of a foam.
Consider the gas: air, for example, which is easily available in large quantities and
is non-flammable, is used to inflate the aqueous foams in fire extinguishers, enabling
Foam identification 5
them to act as a barrier between the combustible material and oxygen from the atmo-
sphere. The choice of gas also influences the ageing of the foam: carbon dioxide, which
is soluble in water, makes drinks effervesce, while nitrogen or alkanes, which are less
soluble, delay the ageing of beer or shaving foams (§3, chap. 3).
The liquid also exerts an influence on the ageing of a foam. It drains (because
of gravity for example) through the network of films and their junctions, which has
the effect of drying the foam (§4, chap. 3). The liquid presents countless possibilities.
It can, for example, provide a vehicle for some active ingredient: varicose veins are
treated by injecting a foam incorporating a sclerosing medicine, which has the same
effect as for a traditional treatment but with much less liquid, and therefore less med-
ication. Or the foaming liquid could be made to harden, through a chemical reaction
or a change in temperature, to allow the generation of a solid cellular structure (a
metallic or polyurethane foam for example) that is much more durable than a liquid
foam.
The stabilizing molecules in a foam also influence its properties and, due to their
particular chemical affinity for one material over another, foams are used in separation
methods (like mineral flotation) or in detergency. As the impurities compete with the
surfactants, they weaken and even destroy the liquid films (§5, chap. 3): the foam thus
serves as a marker during cleaning (dishes, shampoo), since it only appears once the
main contaminants have gone.
2.2 Geometrical and physical properties

The presence of a large quantity of gas makes a foam a markedly less dense material
than the liquid it contains. Thus, a foam of water and air floats on water (as in white
horses on the sea, or fig. 0.1). It contains less material and thus fills the same volume
or covers the same surface for a lower cost. When a foam is used to decontaminate
a nuclear power station or asbestos, on the one hand less active product is used, but
there is also much less waste to reprocess afterwards. In the same way, treating a
material with a foam (to make it impermeable or grease-resistant, for example) rather
than with a liquid not only enables you to economize on product but also to dry the
material more quickly.
A foam also contains a large number of interfaces (§3, chap. 2), and therefore
an enormous surface area per unit volume: with 50 or 100 grams of water, one can
easily make a litre of foam with 10 m2 of interface! The possibilities for molecular
transfer are multiplied, which is valuable in foods where foams are strong flavour
enhancers (of chocolate or spices for instance) and allows salt or sugar content to be
reduced.
Its numerous interfaces make the foam refract and reflect light in all directions,
preventing light from propagating in a straight line. A foam absorbs light (although
liquid and gas are generally transparent), and it blurs vision as much as fog, frosted
glass, or milk. Because of this, probing the bulk structure is made easier with 3D
tomography. Sound travels more slowly (and less far) in an aqueous foam than in
6 Uses of foams
water or in air, due to the alternation between liquid and gas. Moreover, a liquid foam
conducts electric current, particularly when the liquid fraction is high. These physical
properties are exploited when probing foam structure (§1.2, chap. 5).
2.3 Mechanical properties

A foam’s mechanical response is remarkably different from that of both the liquid and
the gas from which it is made. In fact, a foam behaves either as a solid or a liquid
(see chap. 4)! It can be a weak (visco-)elastic solid, capable of returning to its original
shape if the deformation is not too great, or a (visco-)plastic solid, to be sculpted at
leisure. It insulates against knocks and shock waves and therefore offers remarkable
mechanical isolation.
A foam can also flow like a liquid, seeping into pores, filling cavities, and coating
fibres. It can be poured into containers and run through tubes of various shapes. It
behaves as a “yield stress fluid” and because its viscous resistance increases less quickly
with flow rate than that of a normal fluid it is able to reduce frictional losses. It also
keeps small particles in suspension, even when flow stops. All these characteristics are
crucial for the use of foams in drilling for oil.
3 What are foams used for?

Foams have novel properties owing to their lightness, their very large specific surface
area, and the fact that they exhibit both solid and liquid behaviour. These properties
often meet the somewhat contradictory requirements of a number of industrial ap-
plications [1]. We characterize some of the sought-after functions (fig. 1.3) and give
examples of applications in which they are used.
3.1 Desirable functions

Why are foams preferable to other fluids in industrial processes? There are at least
seven good reasons:
1. Economize on product, reduce waste
Using a foam offers the possibility to employ a smaller quantity of active ingredient
or raw material without diminishing its effect. It often reduces the amount of liquid
that must be discarded or treated following an industrial operation, for example
in decontamination and cleaning. So radioactive tanks could be decontaminated,
following the decommissioning of a nuclear power station, by cleaning them with a
specially formulated foam.
2. Quickly fill large spaces
The high rate of expansion of a foam (the ratio of foam volume to the volume
of liquid used to generate it) allows it to rapidly fill large spaces. For example,
fire-fighting foams must cover large areas as quickly as possible. If an aeroplane’s
landing-gear fails, a layer of foam can be spread across more than a kilometre of
runway within half an hour to allow the aeroplane to land. On a more playful note,
“foam parties” also require large volumes of foam.
What are foams used for? 7
Confer the
Economize on Isolate, Give a fluid
Quickly fill Trap Absorb or structure
product, confine, the behaviour
large spaces material exert pressure of a foam
reduce waste smother of a solid
on a solid
Cleaning
Surface treatment
Building materials
Reducing pollution
Fire-fighting
Extraction of
natural resources
Cosmetics
Food
Army, Police
Fig. 1.3: The relation between the different functions that aqueous foams perform
and their main uses and applications.
3. Isolate, confine, smother

In fire-fighting it is necessary both to smother the fire and to isolate the source of
combustion. Foams are able to extinguish burning hydrocarbons more effectively
than water due to their low density and because they float and thus isolate the
fuel from oxygen in the air. Other examples include the foam extinguishers kept
in chemical factories in case of accidents, the sticky anti-riot foams used by police
or soldiers, and even certain insects (like froghoppers) whose nymphs secrete a
cocoon of foam around themselves to protect against drying out and temperature
variations.
4. Trap substances
The interfaces of a foam are able to capture very small particles of solid matter
(for example in purification), ions (during water purification or in the separation of
uranium), organic molecules (enzymes, proteins, polymers), or small molecules like
the surfactants themselves. This forms the basis of numerous separation processes,
like sewage treatment or mineral flotation, which we describe later in this chapter.
Even a gas can be trapped, in the bubbles, as in methane recovery from coal mines.
5. Absorb or exert pressure
Foams are used to muffle explosions (for example in mine clearance). A foam is
injected into rock or concrete in a controlled way to fracture and excavate them.
Conversely, foam also limits the problems of overpressure in the bore-holes used in
drilling for oil.
8 Uses of foams
6. Give a fluid the behaviour of a solid

Creating a foam is one way to increase the viscosity of a fluid to such an extent
that it acquires the mechanical properties of a solid, as in the case of cosmetics or
food. A foam which is sufficiently elastic doesn’t flow under the effect of gravity,
and can therefore be used to cover vertical surfaces, in order to clean or treat them
for example. The elasticity of foams is also useful in displacing oil to extract it from
a reservoir.
7. Confer the structure of a foam on a solid
Finally, liquid foams serve as precursors of solid foams for the construction of cellular
materials, including bread and cakes, and polyurethane, glass, and metal foams [2].
The structure of the solid material is determined by that of the liquid foam, which
is itself governed by the laws of capillary action (chap. 2), as long as it solidifies
before ageing processes like drainage render it heterogeneous. Likewise, cellular
concrete is obtained by mixing a foaming agent into the concrete, then degassing
and solidifying it.
3.2 Mineral flotation

Among the many applications cited above, the process of mineral flotation is one of
the most important. It has been used for more than a hundred years, and now uses
close to 90% of the world’s surfactant products. After it has been extracted from the
ground, the ore-containing rock is ground up and then treated in order to separate the
mineral, such as copper, zinc, nickel, talc, or carbon, from the gangue. This separation
step must be inexpensive and rapid—typically of the order of one hundred tonnes of
mineral per hour.
The principle of flotation is based on the wetting properties at the liquid/air in-
terfaces. It is possible to optimize the physico-chemical conditions (by adjusting the
pH, and the nature and concentration of the surfactants) so that the mineral grains
are trapped by the interfaces while the gangue remains in the liquid. The attraction
of using a liquid foam is of course that it has very high interfacial area.
The ground-up rock and mineral are placed in large tanks (height and diameter of
several metres) (fig. 1.4). Two types of surfactants (at least) are added: those that
control the affinity of the mineral for the interface and those that cause the mixture
to foam. Air is injected at the base while the mixture is vigorously mixed in a way
that favours encounters between bubbles and particles. A foam is thus continually
formed and rises to the surface of the tank before spilling out. It brings with it the
mineral, attached to the bubbles (fig. 1.4). During the ascent, a large proportion of
the liquid falls back down, making the foam dry and fragile. During the collisions
between bubbles, the gangue particles may be trapped in the liquid phase. They are
discharged downwards with the draining liquid, a process that can be enhanced by
spraying the tank with water from above. Other properties of the foam contribute to
the efficacy of the process. Its elasticity limits convection in the column, its low density
allows it to float on the surface, where it can be skimmed off, and its fragility and low
water content are advantageous for recovering only the mineral.
Solid foams and other cellular systems 9
Mineral
particle Gas
Mineral-laden
dry foam
Foam
Bubble
Circulation of ground Recirculation

mineral and gangue of gangue
+ +
foaming solution foaming solution
Fig. 1.4: Schematic diagram of the process of mineral flotation that is used to separate
the gangue from the particles of the mineral, which are trapped at the air/water
interfaces and then recovered at the top of the column.
4 Solid foams and other cellular systems

4.1 Solid foams
A closed-cell solid foam consists of pockets of gas surrounded by solid walls but, in
contrast to liquid foams, solid foams can also exist if the walls have holes in them
(so-called semi-open-cell foams), or are even entirely missing, as shown in fig. 1.5.
4.1.1 Natural foams Many foam-like materials are found naturally in the world
around us. Sea sponges are like open-cell foams, with a structure that allows water to
penetrate, and an extremely large surface area which retains this water by capillary
action. Radiolaria are single-celled organisms, typically a hundred microns in size,
that multiply rapidly in marine plankton. Their skeletons (fig. 1.5a) are ordered
open-cell foams. Honeycombs also have very well-ordered structures and have become
symbolic of a perfectly crystalline arrangement, although in reality they contain cells
with different sizes. This, incidentally, poses problems for the bees when trying to join
these cells together (fig. 1.5b).
Closed-cell structures are also found naturally in bones, which are consequently
light yet strong, and in cork, which because of its tiny air bubbles is light, supple, and
provides good sound insulation. We make use of the lightness and roughness of pumice
stones, that is rocks which contain bubbles, which usually have closed cells, but may
be open, as in reticulite, an extreme form of pumice containing at least 95% air.
4.1.2 Artificial foams Since these properties are so desirable, many artificial
foams have been created to mimic them. In the automobile industry, for example,
decreasing weight leads to a reduction in manufacturing and running costs. Vehicles
therefore contain polyurethane foams (cushions), rigid plastic foams (dashboard), and
aluminium foams (bumper), to name just three.
10 Uses of foams
(a) (b)
Fig. 1.5: Examples of natural solid foams. (a) The radiolarian skeleton can be seen
as an open-cell solid foam (from [5]). (b) A bees’ honeycomb is an example of a closed-
cell solid foam. The cells which appear in white on the photo are filled with wax: they
contain either honey or young bees undergoing metamorphosis. The others are either
empty, occupied by larval worker bees, or full of pollen (in grey). The large cells are
for the males (from [4]). c Pierre Deom, www.lahulotte.fr.
Some of these foams are obtained by solidifying a liquid foam. To better understand
what controls properties such as the average bubble size, the bubble-size distribution,
the volume fraction of solid, and the homogeneity of the material, it is necessary to
model the liquid foam before solidification. This is a complicated problem which links
many of the topics covered in this book: coarsening, drainage, rupture, rheology, and
physical chemistry.
The solidification may follow either direct injection of a gas into a molten liquid
(physical blowing agent), or the production of bubbles within a liquid through a chem-
ical reaction which produces gas (chemical blowing agent), for example by adding a
powder that decomposes at high temperature. Both aluminium foams and polymeric
foams can be made like this. Another possibility is to mix air into the liquid by beating
or whisking, as in chocolate mousse, which is solidified by cooling, or meringue, which
is solidified by cooking (§1.2).
Polymeric foams typically contain 60% to 90% gas, and are found in artificial
sponges, filters, membranes, and bottle closures (synthetic cork). Those which have an
open-cell structure are generally obtained from thermosetting polymers, which poly-
merize and consequently harden as the foam expands. Polyurethane foam is probably
the example most commonly used in industry, as it provides the filling for seats, cush-
ions, and mattresses, is used in buildings for thermal insulation, and is injected into
bicycle or car tyres to stop them deflating. Polymeric foams with a closed-cell structure
are obtained by dispersing gas bubbles in a molten thermoplastic polymer (polystyrene,
Solid foams and other cellular systems 11
polyethylene, polyvinylchloride), which is already polymerized and hardens on cooling.

As they are more rigid, they are used to make various light plastic objects. Expanded
polystyrene is a good thermal insulator, so is used to make cups for hot drinks. There
are also foams made from biodegradable polymers which are now used for packaging.
Metal foams have experienced a recent surge in popularity. Aluminium foam is
lightweight yet strong, making it desirable as a structural material for walls, racing
cars, and aircraft. It is very good at absorbing shocks (although it must be replaced
after each impact), so is often used for bumpers on cars, trams, and even aeroplane
nose-cones (to protect against bird strikes). Aluminium foam has a high surface area
per unit volume, making it useful as a heat exchanger, for example in electronics,
or as a catalyst and porous electrode in chemistry. Its diverse applications include
uses in the military and aerospace industry, for example as weapon silencers and to
stabilize (laminerize) flows in wind tunnels. Lead foam is obtained more easily than
aluminium foam, because lead becomes molten at a lower temperature, but it is too
soft for practical applications. On the other hand, steel foam is made at much higher
temperatures, which makes it unprofitable, despite its strength and low density which
would be valuable in making unsinkable ships. The large surface area, which facilitates
more chemical reactions, in a small volume is put to good use in mobile phone batteries.
In many industrial processes, foam is an undesirable by-product; in the case of
glass, it is particularly awkward because a homogeneous and transparent glass must
not contain visible bubbles. However, manufacturers have recently turned this to their
advantage and are now producing aerated glasses with around 20% gas. Foamed glass
is light, cheap, watertight, does not pollute, is stable at temperatures up to 80 ◦ C, and
is also incredibly robust: in fact, a propagating fracture is stopped when it meets a
bubble. It is already used for the foundations of houses and roads, and could become
common in building for the same reasons as metal foams. It provides excellent thermal
and acoustic insulation and doesn’t contain fibres (unlike glass wool, rock wool, and
asbestos). It exemplifies the importance to industry of producing aerated materials.
Another foamed construction material is cellular concrete, which offers a compromise
between mechanical strength and weight, with good thermal insulation.
4.2 Other cellular structures

Materials made of cells filled with a liquid or a solid, rather than gas, are visually
similar to foams and in particular they tessellate space with non-overlapping cells
without gaps between them.
Some of these systems are so similar to foams that they age in the same way (we
tackle these aspects in chap. 3). This is true of concentrated emulsions (dispersions
of one liquid in another, immiscible, one), of polycrystals (assemblies of monocrys-
tals with different orientations in contact with each other) and magnetic garnets
(assemblies of magnetic domains). We discuss this further on p. 77.
Vesicles are small water-filled sacs formed from lipids, typically around ten microns
in size, which form a cellular pattern (fig. 1.6a) when they assemble. As the lipid
interfaces do not have the same tension, the contact angles are not necessarily 120◦ ,
unlike foams (see §2.2, chap. 2).
12 Uses of foams
(a) (b)
Fig. 1.6: (a) Vesicles are tiny droplets, typically of the order of 10 μm in size, en-
closed by a lipid membrane. As they assemble, the competition between adhesion and
curvature energies causes them to form a structure that resembles a foam. Photograph
courtesy of D. Milioni. (b) Outlines of cells during the formation of a fruit fly
(Drosophila) wing. Size of the image: 160 μm. Photograph courtesy of P.L. Bardet.
A biological tissue (epithelium) is a cellular arrangement in two dimensions. In a

fly’s eye, the facets are arranged in a hexagonal network, possibly even more perfectly
than that of honeycomb. During the formation of the wings (fig. 1.6b) or the thorax,
the initially disordered pattern evolves towards a hexagonal network as the adult
stage is approached, with the cells undergoing structural rearrangements similar to
those seen in foams, which will be discussed in §1.2, chap. 3 and §3.2, chap. 4. These
cell rearrangements are important in skin regeneration and wound healing. The cell
assemblies are yield-stress materials, with energy barriers which can prevent them
from relaxing to the global minimum of energy. In this sense they are closer to foams
than liquids.
Finally, we give some more esoteric examples.
Carbon atoms in graphite make flat hexagonal arrangements which can roll up to
form cylinders (“carbon nanotubes”, fig. 1.7a) or spheres (“buckyballs”). A liquid
which solidifies changes density and cracks, forming patterns which are often cellular.
Examples include salt lakes, basalt (like the Giant’s Causeway in Northern Ireland),
or simply a muddy puddle.
In cosmology, the universe consists of galaxies which enclose large regions of empty
space between them in a structure which is often compared to a foam, although it is
difficult to establish any link. Even more speculative is a recent measurement of cos-
mological radiation that suggests that the universe is finite but without boundary. One
possible interpretation of this is that the universe is a dodecahedral cell with periodic
boundary conditions, obtained by identifying the opposite faces pairwise (fig. 1.7b).
In other words, someone who sets off in a straight line comes back to his or her point
of departure without ever having crossed the boundary (on earth, the analogy is of a
journey that is always to the west, so that you finally return to your starting point
from the east). This cell would have angles of 120◦ between faces and so to be able to
Experiments 13
Graphite
Diamant Fullerène Nanotube

(a) (b)
Fig. 1.7: Foams of all sizes. (a) Carbon atoms arranged in a regular hexagonal
network, at a scale of the order of 10−10 m, which is the basis for fullerenes and
nanotubes. (b) Some observations at the scale of the universe (1023 m) can be in-
terpreted by assuming that the universe is an assembly of dodecahedra, with faces of
constant curvature that meet at angles of 120◦ to form a periodic structure similar to a
foam (cf. fig. 2.21a). Image obtained using J. Weeks’ Curved Spaces software [7]. See
Plate 2.
tessellate space with these cells, the space could not be Euclidean but instead would
have positive curvature [3].
5 Experiments
5.1 Three ways to make a foam
1) Whisking
difficulty level: cost of materials:
preparation time: 5 min experimental time: 15 min
Phenomena demonstrated: making a foam by whisking, foamability, liquid fraction,

stability, loss of transparency, bubble size, appearance of elasticity.
Materials:
• a bowl
• 3 eggs
• a whisk
1. Separate off the yolks and put the egg whites in the bowl.
2. Note the volume of egg white, its colour, and consistency.
3. Whisk the whites, stopping every 10 seconds to repeat the observations and, in
addition, noting the bubble density and size.
14 Uses of foams
2) Shaking
preparation time: 5 min experimental time: 1 hr
Phenomena demonstrated: making a foam by shaking, foam disorder, loss of transpar-

ency, bubble size, appearance of elasticity.
Materials:
• 9 plastic bottles
• water
• washing-up liquid
1. Mix air, water, and washing-up liquid, and shake.
2. For each of the three ingredients, put in a large amount, a little, or none at all.
3. Compare the foam produced by these nine combinations, noting the bubble density,
the volume of foam, its transparency, colour, consistency, and stability.
3) Blowing
preparation time: 5 min experimental time: 1 hr
Phenomena demonstrated: making a foam by beating and blowing, foam disorder,

bubble size and geometry, appearance of elasticity.
Materials:
• a large transparent bowl
• water
• straws
1. Put the washing-up liquid in the water.
2. Make a foam either by agitating with a finger, blowing directly, or blowing through
a straw.
3. Compare the foam produced by these three methods, noting the bubble density,
the volume of foam, its transparency, colour, consistency, stability, and disorder
(variety of bubble sizes).
4. When the bubbles are very large, observe their geometry, in particular that their
faces meet in threes with an angle of 120◦ (see §2.2, chap. 2).
5.2 Chocolate mousse

preparation time: experimental time:
5 min 20–40 min (then 2–3 hrs to cool)
References 15
Phenomena demonstrated: making a foam by whisking, foamability, liquid fraction,

stability, loss of transparency, bubble size, appearance of elasticity.
Reference in the text: §1.2
Materials (for 6 people):
• 6 eggs
• 200 g chocolate
• three bowls
• a whisk
1. Separate the yolks from the egg whites.
2. Whisk the whites until they are very stiff: see §5.1.
3. Break the chocolate into small pieces and melt it in a microwave or in a saucepan
containing a little hot water. Remove it from the heat, and allow to cool.
4. Mix the chocolate with the egg yolks, then mix all of this with the egg whites (try
to be thorough, even if the foam almost completely collapses, as the mixture should
be homogeneous).
5. Leave in the refrigerator for at least 2 hours (and not more than 24 hours).
Comments: The variations are countless. You could:

1. Put in more chocolate.
2. Add strong coffee or alcohol (to the melted chocolate), sugar (to the whites after
whisking), crème fraı̂che (to the yolks), almonds (for decoration at the end), etc.
3. Replace two of the yolks with 150 g butter (mix thoroughly with the melted
chocolate before adding to the yolks).
4. Or don’t even separate the whites and the yolks! Whisk them together with icing
sugar and rum, and add to the melted chocolate.
References
[1] J. Aubert, A. Kraynik, and P. Rand, Sci. Am., 254, 74, 1986.
[2] J. Banhart and D. Weaire, Phys. Today, July, 37, 2002.
[3] J.-P. Luminet, J.R. Weeks, A. Riazuelo, R. Lehoucq, and J.-P. Uzan, Nature,
425, 593, 2003.
[4] La Hulotte, n◦ 28-29, 1er semestre 1997, p. 67 (inside cover).
[5] S. Hildebrandt and A. Tromba, Mathematics and Optimal Form, Scientific
American Library, W.H. Freeman, New York, 1985.
[6] D. Wender (trans.) Hesiod and Theognis, Penguin Classics, Harmondsworth,
England, 1973.
[7] www.geometrygames.org/CurvedSpaces/index.html.fr
2
Foams at equilibrium
In this chapter we will discuss the structure and stability of foams at equilibrium. We start
with a qualitative description of foam physics before turning to very dry foam structures. We
will show how equilibrium properties are affected by the presence of a non-negligible amount
of liquid in the foam, and we will describe the spatial distribution of this liquid.
1 Description at all length-scales

There are at least four length-scales at which we must consider the properties of a
foam (fig. 2.1):
• the observer’s scale, of the order of a metre, at which a foam has the appearance
of a soft, opaque solid (fig. 2.1a).
• the millimetre scale (or even smaller, depending on the foam), at which bubbles
can be distinguished. There are a small number of local geometric rules, called
Plateau’s laws, which describe how the bubbles pack together to form the foam
skeleton (fig. 2.1b).
• the micron scale, or slightly less, which shows how liquid is distributed between the
bubbles (fig. 2.1c).
• the nanometre scale, at which the molecular structure of the interfaces appears.
The presence of particular molecules, like soap, which position themselves at the
air/water interface, is essential to our understanding of the formation and stability
of the interfaces (fig. 2.1d).
In this introductory section we go through fig. 2.1 in reverse, starting at the smallest
scale and working our way up to the largest, to qualitatively describe the elementary
physical properties that allow a foam to exist. In §2 we shall return to each of them
to give a quantitative description.
1.1 At the scale of a gas/liquid interface

Surface tension and the Young–Laplace law Creating an interface between a
liquid and a gas requires energy. This additional energy is γ, the energy per unit
area of the interface, multiplied by the area of the interface created. It means that an
interface tends to reduce its area, and it is therefore under tension. This is why γ is
called surface tension or superficial tension. The word “superficial” refers here to the
surface and not to something of little importance! On the contrary, this tension will
be essential in what follows, because of the vast quantity of interface present in a foam
(up to thousands of m2 per litre, as discussed in exercise 7.1).
18 Foams at equilibrium
(a) (b)
Dry foam
Bubble
Wet foam
Bubbly liquid
(c) (d)
Liquid
Film
Edge
Gas
Fig. 2.1: Foam structure at different length-scales, with each image magnifying the
previous one (scales are very approximate). (a) The whole foam, appearing wet at
the bottom and dry at the top, typical scale ∼10−2 −1 m. (b) Bubbles in a dry part
of a foam, ∼10−4 −10−2 m. (c) A Plateau border (liquid channel), ∼10−6 −10−4 m,
and a soap film (bubble wall), ∼10−8 −10−6 m. (d) A gas/liquid interface containing
molecules with long hydrophobic tails (in the gas) and hydrophilic heads (in the liquid)
∼10−10 −10−8 m.
The shape of a liquid/liquid or gas/liquid interface is determined by the existence of

this tension (see p. 35). For example, if the geometric constraints allow it, an interface
is flat, while if an interface completely surrounds some fluid then (if the effect of
gravity is negligible) the surface is spherical. In general, we determine the shape from
the Young–Laplace law : the pressure difference between the two sides of an interface
is equal to the mean curvature of the interface multiplied by its surface tension (see
§2.1.3). The pressure is higher on the concave side. Surface tension acts to flatten the
surface whilst the pressure difference tends to curve it.
Surfactants The additives used to make foams consist of molecules with a polar
head and a long carbon chain for a tail. The head lowers this molecule’s energy when
it is surrounded by water molecules, so it is called hydrophilic, literally “water-loving”.
The tail increases the molecule’s energy when it is surrounded by water molecules,
so it is called hydrophobic. So the molecule both “likes” and “dislikes” water: it is
amphiphilic. If such a molecule is dissolved in water, it therefore tends to adsorb at
the air/water interface, with its head in the water and its tail in the air (fig. 2.1d).
It forms a layer which is one molecule thick, called a monolayer. Due to their effect
Description at all length-scales 19
on the surface tension of the interfaces to which they adsorb (see §2.1), amphiphilic
molecules are also called surface-active molecules, or surfactants (“which act on the
surface”).
These molecules will adsorb at the surface until they cover it completely. If you
then add more amphiphilic molecules, they remain in solution and don’t play any
significant role in the interfacial properties of a foam.
From interface to film: the role of surfactants Whether it is a foam in a river

at the foot of a waterfall or a foam in a bottle of fizzy water that is shaken, there
are many examples of bubbles which burst in less than a second. Each bubble which
climbs to the surface is a small volume of gas enclosed in a film of water of several
microns in thickness. In general, this type of thin film is unstable: under the effect of
Van der Waals forces [22], the distance between the two air/water interfaces tends to
diminish spontaneously. The film thins and then breaks.
If we instead add several drops of washing-up liquid to a bottle of water before shak-
ing it, a similar foam forms, but this time it lasts. Why? Because certain amphiphilic
molecules carry a small charge. For example, sodium dodecyl sulphate (commonly
abbreviated to SDS, and known commercially as “lauryl sulphate”, which is how it
appears on the labels of washing-up liquids) is an anionic surfactant, meaning that it
carries a negative electrostatic charge. The two air/water surfaces in a film are covered
by charged monolayers that repel each other, stabilizing the film at the thickness at
which electrostatic repulsion and van der Waals attraction cancel each other out (see
§4.1.3). The repulsion between the interfaces is called the disjoining pressure (see
fig. 2.2a). We will see in §2.3, chap. 3 that this repulsion is not sufficient to guarantee
the stability of a film, but that it contributes to it.
1.2 At the scale of a film

The familiar expression “soap film” is a two-fold abuse of language. On the one hand,
rather than soap, which does not foam well, we often use washing-up liquid. Should
we say “washing-up liquid film”? No! Because the film consists mainly of water, and
there is only a small amount of surfactant.
Minimal surfaces It is easy to make a single film, for example by dipping a wire
loop in a soapy liquid. When you take the wire out of the solution, the liquid is trapped
between the wire and two liquid/gas interfaces. This system reduces its surface area
as much as possible, under the constraints that the two interfaces mustn’t touch each
other (the surfactants prevent the formation of a hole) but must touch every part
of the frame. The excess solution rapidly flows out, the two interfaces move closer
together and adopt the same shape to leave a thin film of roughly constant thickness
(much smaller than the dimensions of the frame or of a bubble). The film can therefore
be described as a surface without thickness. Its shape obeys the Young–Laplace law
for a single interface, but its surface tension is twice as large because of the two
interfaces. Since it has the smallest possible surface area, it is called a minimal surface
(see fig. 2.2b). Since the two interfaces are open to the atmosphere, the pressure
Π p
(a) (b)
Fig. 2.2: (a) The liquid in the thin films is at lower pressure than the gas, p < P .
Mechanical equilibrium is determined by the surface tension where the interface is
curved (Plateau borders) and by repulsion between two interfaces where the interfaces
are very close (thin film). This repulsive force is fixed by the disjoining pressure Π,
represented by arrows. (b) A minimal surface obtained with a soap film. The irides-
cence produced by the interference between the two interfaces of the film allow its
thickness (of the order of microns) to be estimated. Photograph courtesy of F. Elias,
see Plate 3.
difference across the film is zero and the mean curvature of the film is also zero, a
defining characteristic of minimal surfaces.
Film thickness A newly created film is not generally at equilibrium. It is a few

microns thick and shows beautiful iridescence. If the film doesn’t break, it thins due
to gravity or capillary suction (see §1.3) until it reaches a thickness h of the order of a
few tens of nanometres. The disjoining pressure then prevents further thinning of the
film and ensures that it reaches mechanical equilibrium (see fig. 2.2a).
From film to bubble A bubble is a soapy film of water which encloses a certain
amount of gas. The presence of the gas is an additional constraint on the shape of the
film, so that the film is no longer a minimal surface. A bubble assumes the smallest
possible surface area to contain the gas, and if it is isolated it is spherical. The pressure
in the bubble is slightly greater than atmospheric pressure, but not enough to compress
the gas appreciably, and so its volume remains fixed. An isolated bubble ceases to be
spherical when it is large enough to be sensitive to external forces, for example wind
or gravity if it is several centimetres in size.
When two bubbles suspended in air come into contact, they spontaneously change
their shape in order to share an interface and thus reduce the total interfacial area
(see fig. 2.8). The final shape results from the competition between the two bubbles
each trying to minimize their surface area, so that they cannot both remain spherical.
More generally, when a group of bubbles come together the conservation of volumes
and the minimization of surface area leads to some simple laws which govern the local
shape of the bubbles known as Plateau’s laws, described in §2.2.
Description at all length-scales 21
If the bubbles are suspended in water, putting two bubbles into contact leads to
an increase in total surface area and so this does not occur spontaneously. The surface
area of the film created between the two bubbles increases with the pressure with
which they are squeezed together, known as the osmotic pressure (see §4.2).
1.3 At the scale of a bubble

Bubbles, films, and Plateau borders A bubble in the centre of a dry foam
is polyhedral in shape because of its neighbours. Its faces are thin films that are
gently curved either because of the pressure differences between the bubbles, or simply
because its perimeter does not lie in one plane. The films intersect in threes along the
edges (see fig. 2.3a), which are liquid-carrying channels known as Plateau borders
after Joseph Plateau (1801–1883). The curvature of the liquid/gas interfaces must
remain finite (by the Young–Laplace law), which imposes a non-zero thickness on the
Plateau borders (see fig. 2.3b). The cross-section of each border is a small triangle
with concave sides. Four Plateau borders intersect at the vertices (or nodes) of each
polyhedral bubble.
From bubble to foam The amount of liquid contained in a foam is defined by
the liquid volume fraction, φl = Vl /Vfoam , the ratio of the volume of liquid to the
total volume of the foam. This quantity is linked to the density ρ of the foam by the
relationship ρ = ρl φl + ρg (1 − φl ) ≈ ρl φl , with ρl and ρg the densities of the solution
and the gas respectively.
Different types of structure are obtained depending on the liquid fraction
(fig. 2.1a):
(1) for φl > φ∗l the bubbles are spherical and do not touch: this is a bubbly liquid;
(2) for 0.05 φl < φ∗l the bubbles touch and take the shape of a squashed sphere at
each bubble/bubble contact: this is a wet foam;
(3) for φl 0.05 the bubbles are polyhedral and the Plateau borders have a negligible
cross-section: this is a dry foam.
Bubble
Face
(soap film)
Edge
(Plateau border)
Vertex
(or node)
(a) (b)
Fig. 2.3: Bubble-scale structure of a foam. (a) Photo of a dry foam and definition of
structural elements. Photograph courtesy of F. Elias. (b) The network of Plateau bor-
ders in a wet foam, obtained by X ray tomography at the ESRF (European Synchrotron
Radiation Facility, see §1.2.4, chap. 5). The films are too thin to be detected. Image
courtesy of R. Mokso.
The transition from bubbly liquid to wet foam occurs when the osmotic pressure
vanishes (§4.2) [34], which corresponds to a precise value of liquid fraction, φ∗l , close
to 0.3 in 3D (see fig. 4.17). Its value corresponds to the unfilled volume within a
packing of hard spheres, as in a pile of oranges. It is a bit lower for an ordered foam
structure than for a disordered structure, since if spheres are stacked in a regular
fashion, rather than at random, more of them can be fitted in a given volume. The
transition between wet foam and dry foam is less well defined. We speak of dry foam
when the presence of the liquid doesn’t play any role in the observed phenomenon,
usually when the liquid fraction is less than 0.05.
The liquid phase is continuous: liquid circulates freely between the films and
Plateau borders. At equilibrium, in the absence of gravity (see §1.4.1), the pressure
in the liquid is thus constant. The lower the liquid fraction, the smaller the cross-
section of the Plateau borders and the greater their curvature: they are therefore at
lower pressure than the gas. In contrast, the curvatures of the films are small and of
opposite sign for each interface. In the absence of disjoining pressure, i.e. as long as
the film is at least a hundred nanometres thick, the pressure which exists between
the interfaces is of the order of the gas pressure in the bubbles. That is why the li-
quid is drawn into the Plateau borders, a process known as capillary suction, which
comes to a halt when the thickness of the film is sufficiently low that the disjoining
pressure becomes non-negligible and balances it. At equilibrium the film is therefore
thin and most of the liquid is found in the Plateau borders and in the vertices where
they meet.
1.4 At the scale of a foam

1.4.1 Foam under gravity Below the millimetric scale, bubble pressures and
surface tension are the only forces present. Above it, gravity causes the liquid to
flow through the network of Plateau borders (see §4, chap. 3). At equilibrium, the
pressure in the liquid is fixed by hydrostatic equilibrium, and so decreases with
increasing height. The bubbles are not subject to gravity and the gas is, on aver-
age, at constant pressure. The pressure difference between the liquid and gas thus
grows linearly with vertical position, allowing the radius of curvature of the Plateau
borders and the liquid fraction to be deduced. A consequence of this is that at equi-
librium there is more water at the bottom of the foam than at the top, as illustrated
in fig. 2.1a.
1.4.2 Quantity of interface At the scale of the bubble, we have seen that the
quantity of liquid/gas interface is the smallest possible, subject to conserving the
volume in the bubbles and, depending on the situation, the liquid volume or pressure.
Yet at the scale of a foam the bubbles are never arranged in the optimal configuration.
There are too many possible arrangements and too few bubble movements for it to
be found. Instead, the foam is in a local minimum of energy: by moving a very small
amount, it will increase its energy, but if it deforms further it may find a lower energy
state.
Local equilibrium laws 23
2 Local equilibrium laws

Minimization of interfacial area, which governs the structure of a foam at equilibrium,
is determined by simple geometric rules at the scale of a film and of a few bubbles.
These laws, and the extent to which they are valid, are discussed in this section.
2.1 Equilibrium of fluid interfaces

2.1.1 Surface tension An increase in interfacial area by an amount dS induces a
change in the energy (or the free energy, which we will make precise later) by an amount
dEsurf = γdS, (2.1)
where γ is defined as the surface (or superficial) tension [22].

The tension γ can be considered an energy per unit area (J.m−2 ) or a force per
unit length (N.m−1 ). One side of a line segment dl (see fig. 2.4a) exerts a force γ dl n̂
on the other, with n̂ the unit normal to the segment in the plane of the film. For pure
water, in air at room temperature, γ0 = 0.0727 J.m−2 = 72.7 mN.m−1 .
A surface tension is defined for any interface that separates a liquid from a gas and,
more generally, two immiscible fluids or a fluid and a solid. Unless indicated otherwise,
we take the surface tension of a solution to be its surface tension in contact with air.
The surface tension is always positive, since otherwise the interface would be unstable.
It is only defined at a scale that is large compared to the size of the molecules, i.e.
at a scale at which the interface can be treated as continuous. In this book we shall
always assume that the size of the molecules at an interface is small compared to its
characteristic size, given, for example, by its radii of curvature (see the definition in
fig. 2.6). So the energy depends only on the area of the interface and not on its shape:
it is equal to the energy of a planar surface of the same area (although if the thickness
h of the film is sufficiently low it may play a role; see §4.1.4).
At a microscopic level, the molecules of a liquid, which are attracted to each other,
find an interface less favourable than when they are in the bulk [22]. This difference,
γ0
Gas
dl Πl
γ
Interface
^
n
Liquid log c
cmc
Fig. 2.4: Surface tension. (a) The surface tension is a force per unit length dl exerted
along n̂ in the plane of the interface. (b) The change in surface tension of a solution
with increasing surfactant concentration. The plateau marks the saturation of the
interface by surfactant molecules and the corresponding concentration is called the
critical micelle concentration (cmc).
which is a characteristic of the two fluids present, is the origin of surface tension. The
surface tension depends on the temperature and what the liquid contains: in general,
it is reduced when the temperature rises or when there are impurities. In particular, its
value is significantly lowered by the presence of so-called surface-active or amphiphilic
molecules (see §6.1, chap. 3). The more amphiphilic molecules are added, the more the
surface tension is lowered, until a concentration, the critical micelle concentration, is
reached at which the surface is saturated with amphiphilic molecules and the surface
tension no longer varies. An example of the change in γ with surfactant concentration
is shown in fig. 2.4b. The difference between the surface tension of a pure liquid and
the surface tension of the liquid in the presence of solutes is called the surface pressure
or Langmuir pressure, denoted Πl (see §2.2.1, chap. 3).
2.1.2 Film tension Fig. 2.5 shows a thin film produced by drawing a wire frame
from a bowl of soapy liquid. A horizontal slice through this film would have a cross-
section of length L and thickness h ∼ 10−7 m L, with perimeter 2L + 2h ∼ 2L
and area hL. The force exerted by the lower part of the film on the upper part
has two contributions, one from the surface tension and the other from the pressure:
Fγ = −γ 2L uz and Fp = p hL uz , with uz denoting the upwards vertical direction and
p the pressure in the liquid.
The force exerted on this slice by the reference pressure pref (in general equal
to Patm ) is Fref = pref hLuz and so the excess force due to the presence of the
film is F = [−2γL + (p − pref )hL]uz (see exercise 7.2). With γ ∼ 5 × 10−2 N/m,
p − pref < 104 Pa (see p. 48) and h ∼ 10−7 m, we find that the second term is
negligible. The excess force per unit length of film sliced, called the film tension, is
therefore very well approximated by
γf ≈ 2 γ. (2.2)
Frame
Soap film Fγ
Fγ
Meniscus L
Soap solution
Fig. 2.5: Experimental measurement of surface tension. A soap film is formed by

lifting a frame out of a soap solution. The film meets the solution in a liquid meniscus.
Measuring the force Fγ necessary to keep the frame in equilibrium allows the surface
tension to be determined (see §1.1.1, chap. 5).
Taking into account both the pressure term and a slight change in surface tension with
thickness leads to variations of the order of 0.1%, which will be discussed in §4.1.4.
If the frame is moved vertically upwards by a distance dx, the associated energy
change is, by eq. (2.1), 2γLdx. This quantity also represents the work done, Fdx.
2.1.3 The Young–Laplace law If an interface is not flat then the surface tension
induces normal forces which are compensated, at equilibrium, by the pressures on each
side (see fig. 2.6b). There is a relationship between the pressures, the tension, and the
shape of the interface, which we derive for an isolated bubble in the shaded section on
p. 27, and state below for any curved surface.
We denote by S the area of an interface of tension γ between a domain A and a
domain B; M denotes points on the interface, and n(M ) is the normal at M oriented
from A towards B. It is characterized locally by its principal radii of curvature R1 and
(a)
n (M)
(b)
A
PB n (M)
R1 2γ 2γ
PA
R2
Fig. 2.6: The Young–Laplace law. (a) The interface separates space into two half-
spaces A and B, with pressure PA and PB respectively. A plane parallel to the normal
to the surface cuts it along a line with radius of curvature R, the radius of curvature of
the surface in the direction given by the plane of the cut. By convention it is considered
positive if the centre of curvature is in A. R takes different values as the plane of the cut
rotates around the normal, and takes its extreme values in two particular directions,
called the principal radii of curvature, R1 and R2 , which turn out to be perpendicular.
(b) Viewing a cross-section through the film illustrates the component of the force
due to surface tension on an element of the surface in the direction of its normal n. A
pressure difference PA > PB ensures that equilibrium is attained.
R2 , defined in fig. 2.6a. By convention, the radius is counted positive if the centre
of curvature is in A and negative if the centre of curvature is in B. We describe the
changes in surface area and volume during a characteristic deformation in which each
point M of the interface is moved a distance dx in the direction of the local normal
n(M ). The change in volume of A is thus dV = Sdx and the change in surface area is
dS = HSdx, with H = 1/R1 + 1/R2 the mean curvature of the surface.
At equilibrium, the energy change associated with this virtual transformation is
zero (to first order in dx). Then −ΔP dV + γdS = 0 where ΔP = PA − PB is the
difference in pressure between domain A and domain B. Finally, the curvature of
the interface and the difference in pressure between the two regions separated by the
interface balance:

1 1
ΔP = PA − PB = γH = γ + . (2.3)
R1 R2
This equation is called the Young–Laplace law, after Pierre-Simon de Laplace (1749–
1827) and Thomas Young (1773–1829), or sometimes just the Laplace law. It indicates
that surface tension tends to reduce the curvature of an interface, making it more
planar, and that it is counterbalanced by a pressure difference, which tends to bend
the interface.
2.2 Plateau’s laws

Plateau’s laws were formulated in 1873 by Joseph Plateau [52], but only fully proved
in 1976 by Jean Taylor [2, 59]. They are based on an ideal foam model, defined below,
which does not take into account all the details of the physics and chemistry of a foam.
The laws stated in §2.2.2 remain valid for real foams as long as the hypotheses of the
model are approximately satisfied (see §4).
2.2.1 Ideal foam We define an ideal foam by making the following simplifications:
(1) The foam is very dry.
The liquid occupies a negligible part of the total volume of the foam and we
assume that the liquid volume fraction φl is zero (see its definition on p. 21).
Experimentally, liquid fractions of φl ∼ 10−4 to 10−2 are attainable.
(2) The foam is at mechanical equilibrium and is thus static.
The forces within the foam are equilibrated and the foam is at rest, in a local energy
minimum. At the scale of the bubbles, thermal fluctuations are negligible; other
phenomena can alter the foam, however (see §1, chap. 3), and approximation (2)
is only correct if the characteristic time of these changes is much greater than the
time taken for the foam to return to mechanical equilibrium (typically a millisecond
for a soap foam).
(3) The foam has an energy proportional to the surface area of its bubbles (or,
for a monolayer of bubbles, proportional to the total perimeter; see §5 on 2D
foams).
The thin films separating the bubbles are “ideal”: they have zero thickness and film
tension 2γ (the factor 2 comes from the fact that each film has two interfaces, see
§2.1.2). This implies that each air/water interface has the same surface tension, as
is the case for a foam at equilibrium if the monolayers of surfactant are themselves
at equilibrium (see §2.2, chap. 3); equivalently, all other contributions to the energy
(gravity, for example) are negligible, which is the case for very dry foams.
(4) The foam is incompressible.
The volume V of each bubble is fixed (or the area A for a monolayer of bubbles,
see §5 on 2D foams). This approximation is valid over sufficiently short times
that the transfer of gas between bubbles can be neglected (see §3, chap. 3), and
remains true if we apply an overpressure to the foam, provided it remains small
compared to atmospheric pressure (see exercise 7.6, chap. 4 and the shaded section
on p. 27).
2.2.2 Local structure of an ideal foam The Young–Laplace law for a single
interface (eq. (2.3)) applies to an ideal film of film tension 2γ (see §2.1.2 and exer-
cise 7.2). We consider two bubbles i and j, separated by a face Sij with principal radii
of curvature R1,ij (M ) and R2,ij (M ) counted as positive if the centre of curvature
is in bubble i, M being any point of the surface. The mean curvature of the film
Hij = 1/R1,ij + 1/R2,ij satisfies eq. (2.3) and is consequently independent of M :
Pi − Pj = 2γHij . (2.4)
The films are thus surfaces of constant mean curvature: this is Plateau’s first law.
For example, an isolated bubble has a spherical shape (see shaded section below).
We deduce from this that the sum of the curvatures of three adjacent faces is zero.
To see this, note that each time a film is crossed, we add the pressure differences.
Traversing a loop from bubble i to j to k and then back to i, the sum of the pressure
differences is zero: (Pi − Pj ) + (Pj − Pk ) + (Pk − Pi ) = 0. In other words,
Hij + Hjk + Hki = 0. (2.5)
This is a strong constraint on the equilibrium structure of a foam.
Equilibrium of a bubble
The Young–Laplace law predicts that the overpressure in a bubble is a function of
its radius (see eq. (2.4)). We demonstrate this directly in the following two examples.
a – Force balance on half a bubble

Imagine a spherical bubble with radius R that has been cut in half, as shown in
fig. 2.7. The gas is confined by a film of negligible thickness h (h R) and of
tension 2γ. Mechanical equilibrium is a result of the balance between the pressure
forces acting on the internal and external surfaces FΔP = πR2 ΔP and the surface
tension force Fγ = 4πRγ acting on the cut perimeter of the bubble (see fig. 2.7b).
(a) (b)
FΔP
h
Pint Pext
R Pint
Fγ
Fig. 2.7: (a) Cut through a soap bubble of radius R and film thickness h. (b) The
forces acting on the upper half of the bubble.
These two forces are oriented perpendicular to the cut. This leads to the expected
overpressure:
γ
Pint − Pext = 4 · (2.6)
R
b – Free energy balance over the whole bubble

In order to minimize the free energy E of the system consisting of the bubble and
the external and internal gas, we can express how it changes with R, at constant
temperature T , as:

4 4
dE = −Pint d πR3 − Pext d − πR3 + 2γ d 4πR2
3 3
= 4πR [R(Pext − Pint ) + 4γ] dR. (2.7)
The ideal gas law allows us to express Pint as a function of the radius R and of
nRGP T , with RGP the ideal gas constant. Then integrating gives
4
E = −nRGP T ln R3 + Pext πR3 + 2γ4πR2 + cst.
3
The free energy increases when R becomes very large, because the external pres-
sure and the surface tension both act to oppose growth. E also increases if R
becomes very small, because the internal pressure prevents the bubble from col-
lapsing. The curve E(R) does indeed have a minimum when dE/dR = 0, given by
eq. (2.6).
Up to bubble radii of the order of tens of microns, this overpressure is very
small compared to atmospheric pressure. For a bubble of radius 2 mm and tension
50 × 10−3 N/m, 4γ/R ≈ 100 Pa = 10−3 atm. The volume of gas may be considered
fixed in most cases.
Plateau’s second law concerns the way in which three films meet along an
edge. Consider, for example, two bubbles of the same volume V touching each other
(fig. 2.8a). The system reduces its energy by sharing part of the interface; by sym-
metry the shared film is a circular disk, with radius denoted by a. We assume that
each bubble is spherical, with a truncated region (which is the case in practice). The
system is thus characterized by the radius R of the sphere and the radius a of the
shared film. These two parameters are linked by conservation of volume: if a increases,
R must likewise increase to be able to contain the volume driven out of the truncated
region. Minimization of the surface area thus implies a compromise between the gain
associated with the sharing of a part of the interface and the increase in radius of
the sphere. Owing to the volume constraint, a single parameter must describe the
structure and, rather than R or a, the appropriate parameter is the angle α defined in
fig. 2.8a, from which R and a can be deduced. An expression for the area, and thus
the total energy of the system, as a function of α can be obtained analytically. The
result is shown in fig. 2.8b, and the minimum is obtained when α = 60◦ .
The angle at which the three films meet is β = 180 − α = 120◦ . This value can also
be found from a different line of reasoning: the (vectorial) forces of surface tension
exerted by each of the three films cancel out if and only if the angle between each pair
of adjacent films is 120◦ . When a third bubble is added, it minimizes the surface area
of the system by sharing a film with each of the first two bubbles. These three films
again intersect at 120◦ .
Plateau’s third law is obtained by adding a fourth bubble to the system: at the
centre there are four edges connected at a vertex with tetrahedral symmetry (fig. 2.9).
This angle of 109.47◦ between the edges is the only possible choice that allows angles
of 120◦ between the films. The angles found in these configurations are identical for
more complex arrangements of bubbles and constitute a rule of assembly for films in
a foam. In particular, a vertex at which more than four edges meet can always be
separated into a number of four-fold vertices, reducing the energy (see exercise 7.3
in 2D).
(a) (b)
9.7
Surface energy (a.u.)
9.6
β=π–α 9.5
α 9.4
9.3
9.2
9.1
0 15 30 45 60 75 90
α (degrees)
Fig. 2.8: The double bubble. (a) Two bubbles share a common film to minimize their
surface area. (b) The surface energy of the double bubble as a function of the angle α.
The only equilibrium solution is given by the energy minimum.
120°
Film Vertex
109.5°
120°
Edge or
Plateau border
Fig. 2.9: Meeting of 4 bubbles: there are 6 films, 4 edges, and 1 vertex oriented at
characteristic angles to each other, in a configuration that is found throughout a foam.
Summary of Plateau’s laws

(1) Equilibrium of faces: the soap films are smooth and have a constant mean curvature
which is determined by the Young–Laplace law (eq. (2.4)).
(2) Equilibrium of edges: the films always meet in threes along edges, forming angles
of 120◦ = arccos(−1/2).
(3) Equilibrium of vertices: the edges meet four-fold at vertices, forming angles of
θa = arccos(−1/3) ≈ 109.5◦ ≈ 1.91 rad.
These three laws are a necessary and sufficient condition to ensure mechanical equi-
librium of an ideal foam. If an infinitesimal perturbation (that preserves the bubble
volumes) is imposed on the films in a foam structure that obeys these laws, the amount
of interface, and thus the energy, is increased. Plateau’s laws have two-dimensional
equivalents, given on p. 59.
2.2.3 Condition at solid walls It is necessary to add to these laws a condition

for contact with a solid wall: at equilibrium, the angle between a solid wall and a film
in an ideal foam is 90◦ . In practice, this result is valid if the radius of curvature of the
solid wall is large compared to the size of the Plateau border.
3 Dry foams
Despite their varied properties (chemical composition, surface tension, bubble size),
all foams, and also highly concentrated emulsions, have a very similar appearance.
In effect, the bubbles fill space in a disordered way. This leads to rules for the local
arrangement of bubbles with their neighbours (topology), their shapes and sizes (geo-
metry), and the relationship between topology and geometry. In order to understand
Dry foams 31
the properties that all foams share, we consider again the simplified ideal foam, defined
on p. 26, that is very dry, at mechanical equilibrium, and for which the energy is pro-
portional to the interfacial area. Whenever it makes things simpler, we will consider a
two-dimensional foam (see §5).
3.1 Number of neighbours: topology

3.1.1 Euler constant To a mathematician, a dry foam is a partition of space,
that is, a division of space into many regions without overlaps or gaps. A 2D foam is a
partition of the plane. Another example of a 2D partition is the division of a country
into counties: each town belongs to exactly one county (no town belongs to two counties
or to none). On a sheet of squared paper, the small squares partition the sheet.
We define a face to be an idealized surface separating two bubbles. There are as
many faces as soap films in a foam. There is a relationship between the total numbers
of bubbles, faces, edges, and vertices in space:
−N + Nfaces − Nedges + Nvertices = χEuler (= constant). (2.8)
This formula is due to the mathematician Leonhard Euler (1707–1783); it is a topolo-
gical relationship that is independent of the shape of the bubbles and is true even for
a foam which is not at equilibrium.
The constant χEuler in eq. (2.8) is a small integer (check it for some simple
examples!). It is typically equal to one, but may be zero or two if the foam is
contained in a closed box or has a free surface (see shaded section on p. 43). The
important thing is that it doesn’t change if bubbles or faces are added or removed.
For example, the transformation shown in fig. 3.3 conserves the number of bubbles,
increases the number of faces by one, the number of edges by two (one disappears and
three appear), and the number of vertices by one (two disappear and three appear):
the change in eq. (2.8) is thus − 0 + 1 − 2 + 1 = 0.
In two dimensions, Euler’s formula is

N − Nedges + Nvertices = χEuler . (2.9)
For example, in fig. 3.8 one bubble is lost, three edges are lost (five lost and
two gained), and two vertices are lost, so −1 − (−3) + (−2) = 0. This can also be
seen in figs. 3.4 and 3.7.
3.1.2 Average number of neighbours In the shaded section below we show a

remarkable consequence of eq. (2.9) [30, 62]: in an ideal 2D foam with a large number
of bubbles, the average1 n of the number n of neighbours of the bubbles is always
close to 6. More precisely, n is equal to 6, less a small correction that depends on
the periphery of the foam (see shaded section on p. 43) and on N :
cst
n = 6 − . (2.10)
N
1 By convention in the whole of this chapter, angle brackets indicate an average over all bubbles in
the foam, or, if given with a subscript, over a certain collection of bubbles.
In a real foam there are other, generally small, corrections, due for example to the
non-zero liquid fraction. We emphasize that this rule does not fix the number of sides
of each bubble, it is only the average which is fixed at 6. Indeed, the proportion of
5-sided bubbles is often greater than the proportion of 6-sided bubbles. If there are
bubbles with 3, 4, or 5 sides, there must also be some with 7 or 8 sides, or even more.
To describe the difference from a hexagon, we define the quantity qt = 6 − n, called
the topological charge of a bubble. The average over the foam of the topological charge
is zero for a large number of bubbles (its value is of the order of 1/N ) [28].
Consequences of Euler’s formula and Plateau’s laws in 2D [30, 62].

In two dimensions, n refers to the number of sides (or edges) of a bubble or, equi-
valently, its number of vertices. Thus n is the average number of sides per bubble,
the average being taken over all N bubbles in the foam.
If N is sufficiently large that the effects of the periphery of the foam can be
neglected, each edge is shared between two bubbles, n ≈ 2Nedges /N . Further, each
edge links two vertices and each vertex is shared between the three edges which meet
there (see Plateau’s laws in 2D on p. 59), so Nvertices /2 = Nedges /3. Putting this
together, we can write
2 3
N= Nedges = Nvertices .
n n
Then Euler’s formula in 2D (eq. (2.9)) gives
n n χEuler

1− + = · (2.11)
2 3 N
The right-hand side tends towards 0 as the foam becomes large and we thus obtain
n = 6. This reasoning applies to any partition in which z̄, the average number of
edges which meet at a vertex, is known. Then 1/n = 1/2 − 1/z̄. If all the vertices
are four-fold, for example the squares on a sheet of squared paper, n and z̄ are
each equal to four and we verify that 1/4 = 1/2 − 1/4.
Returning to three dimensions, we can show (see exercise 7.4) that for each in-
dividual bubble there is a link between its number of faces f = Nfaces and its average
number of edges per face n̄. A simple counting argument using Plateau’s laws implies
that
2 3
f= Nedges = Nvertices . (2.12)
n̄ n̄
Inserting this into Euler’s formula (eq. (2.8)) gives
12
6 − n̄ = · (2.13)
f
Dry foams 33
Whether the number of faces is very small or very large, n̄ is thus always strictly less
than 6. Put another way, no bubble has all hexagonal faces, and there must be some
faces with 5, 4, or 3 edges.
Eq. (2.13) is easily tested on simple polyhedra such as those of fig. 2.13, where
n̄ = 3, 4, or 5, respectively. A familiar example is a football, which has 20 hexagonal
and 12 pentagonal faces: here n̄ = (6 × 20 + 5 × 12)/32, and 6 − n̄ = 12/32, which
obeys eq. (2.13). Another example is the Kelvin cell (see the shaded section on p. 50),
a truncated octahedron which tessellates 3D space (just as hexagons tessellate the
plane). It has 8 hexagonal and 6 square faces, so f = 14, n̄ = (8 × 6 + 6 × 4)/14, and
then 6f − n̄f = 6 × 14 − 8 × 6 − 6 × 4 = 12, as required.
If eq. (2.13) is averaged over all the bubbles of a foam (cf. exercise 7.4) we find

12
f = · (2.14)
(6 − n̄)
In 3D, unlike 2D, the average number of faces in a foam is not fixed. If the foam is
disordered, n̄ is often quite close to 5 (fig. 2.10), and f is typically between 12 and
15 (and often between 13 and 14). On the other hand, the variance μf2 ≡ f 2 − f 2 of
the distribution of the number of faces in a foam is very variable: this is one measure
of the disorder of a foam (see §3.2.1, chap. 5). An example of the distribution of f is
given in fig. 3.24.
3.1.3 Rivier’s theorem For each bubble in a 3D foam, the number of faces that
have an odd number of edges is even [57, 58]. To see this, assign to each face the
number (−1)n , where n is its number of edges. For any bubble, the product of (−1)n
over all faces is (−1)2Nedges , because each edge is counted twice, which is equal to 1.
Hence the result.
Applying the same argument to any closed surface in a 3D foam, i.e. the outside
of a group of connected bubbles, shows that the surface also has an even number
of faces that have an odd number of edges. A corollary is that if a line is drawn
(a) (b)
Fig. 2.10: Examples of simulated 3D bubbles. (a) A bubble with f = 13 faces that is
commonly observed in foams [48]: it has one face with 4 edges, 10 with 5 edges and 2
with 6 edges. (b) A bubble with f = 26 faces, all either pentagonal or hexagonal;
although it is at equilibrium and unstressed, it is irregular and elongated [18].
between the centres of any two bubbles that share a face with an odd number of edges,
there is an even number of lines entering the closed surface. These “Rivier lines” can
form loops, or end at the surface of the foam, but they never terminate in dangling
ends.
3.1.4 Correlations between neighbouring bubbles In disordered foams we

observe that bubbles with more faces generally have neighbours with few faces.
More specifically, take a 3D bubble with f faces and write f neighbours for the
average number of faces of its f neighbours. Both experiments and simulations suggest
that f and f neighbours are inversely correlated, that is if one is large (compared to
the average value) then the other is small.
It is not a general rule, but Lewis, Aboav, and Weaire have all observed in prac-
tice that f neighbours ∼ A + B/f where A is a constant slightly less than f and B
is another positive constant [1, 56]; see fig. 2.11 and eq. (5.13) for an alternative
expression.
350
300
250
f<f>neighbours
200
150
100
5 10 15 20
Number of faces, f
Fig. 2.11: Example of the relationship between the number of faces f of a bubble and
the number of faces of its neighbours, in a foam obtained by numerical simulation with
the Surface Evolver [38]. The midpoint of each error bar is the average of f f neighbours
over bubbles with the same f , and the length of each error bar represents the standard
deviation. This representation underlines the affine variation of this term with f, as
observed by Aboav and Weaire (solid line, eq. (5.13)).
Dry foams 35
3.2 Bubble geometry

3.2.1 Energy of a foam Since each soap film has two gas/liquid interfaces, the
surface energy (eq. (2.1)) of a foam can be written

N
E = 2γ Sij = γ Sij = γ Si = γSint , (2.15)
0≤i<j≤N 0≤i,j≤N i=0
where the index 0 means the region outside the foam, Sij is the area of the face
separating bubble i and j (Sij = 0 if bubbles i and j are not neighbours, Sij = Sji ,
and Sii = 0), Si = j Sij is the surface area of bubble i, S0 is the surface area of
N
the outside of the whole foam, and Sint = i=0 Si is the total amount of gas/liquid
interface in the foam.
Eq. (2.15) can be non-dimensionalized by the average bubble volume,
N
V = N1 i=1 Vi :
E
N
Si
= · (2.16)
γV 2/3
i=1
V 2/3
In 2D, the expression is similar:
E
ij Lint
= 2 = , (2.17)
λA 1/2
0≤i<j≤N
A 1/2 A1/2
where λ is the line tension, ij and Lint are the 2D equivalents to Sij and Sint in
eq. (2.15), and A is the average bubble area (see §5.2). At equilibrium a foam is at
a minimum of energy. Eq. (2.16) shows that the existence of surface tension is crucial
in this minimization; its value, however, as well as the value of the average bubble
size, which vary from one foam to another, don’t play an important role. Instead,
it is the shape of the bubbles that determines the minimum, and this is a purely
geometric problem. That is, the equilibrium structures of foams are alike, irrespective
of their average bubble size and chemical composition (although they do depend on
the polydispersity, that is the width, or variance, of the bubble size distribution, see
p. 252). Because of this universality, an expert can generally recognize from a photo
if the foam is at equilibrium.
Energy minimization tends to reduce the surface area of the bubbles and would
also reduce their volume if the bubble pressures did not oppose this. These pressures
adapt in order to keep the volumes constant, and are slightly higher than the pressure
of the outside atmosphere (see shaded section on p. 27). Foams which have a free
surface, like the head on a glass of beer, obey (see exercise 7.6)
N
E= (Pi − Patm ) Vi . (2.18)
2 i=1
Here Patm is the gas pressure outside the foam, Pi is the pressure of bubble i, and E
is the energy given by eq. (2.15). This expression, written here in three dimensions,
generalizes to all dimensions (see exercise 7.6).
3.2.2 Global and local minimization Energy minimization causes frustration:

an almost spherical bubble minimizes its surface area, but increases that of its
neighbours, so that a foam must rapidly find a compromise. While it is out of equilib-
rium, after some external perturbation for example, everything that may be adjusted
(curvature of the faces, position of the vertices, length of the edges) is adjusted while
respecting the constraints (the number of bubbles and the bubble volumes). The total
surface area thus decreases until Plateau’s laws are satisfied throughout.
During this approach to equilibrium, it may happen that an edge shrinks without
managing to satisfy Plateau’s laws. Its length decreases and reaches zero, and its two
ends meet. A vertex at which 6 edges meet is created (in 3D). This situation is unstable
(see exercise 7.3 in 2D) and gives rise to the formation of a new face (a T1 process, see
§1.2.2, chap. 3). The reverse process is equally likely. The number of faces and edges
is consequently not fixed.
With or without T1 processes, when the surface area (or energy) of a foam reaches
a minimum, it is at equilibrium, and stops moving. A foam cannot spontaneously over-
come the energy barrier which separates one equilibrium configuration from another.
In particular, thermal fluctuations with amplitude kB T ∼ 4 × 10−21 J, are negligible
in comparison to the energy barriers between equilibrium states, which are much lar-
ger (see p. 80) and the foam is unable to change its configuration. The equilibrium
configuration of a foam thus depends on its history, including, for example, the way
in which the sample was prepared.
If the bubbles are generated randomly, there is consequently almost no chance of
them spontaneously assuming, among all possible equilibrium configurations, the one
with the smallest possible surface area or energy (fig. 2.12). We say that a foam is
Fig. 2.12: Each bubble of this 2D foam has a similarly sized neighbour. Therefore
each bubble is almost regular. It is a rare (and artificial) example where the foam
might be expected to approach its minimum energy configuration [28]. Photograph
courtesy of E. Janiaud [28]. Reproduced by kind permission of The American Physical
Society.
Dry foams 37
in a local, not a global, minimum of energy. In fact, it is not proven (for a foam of
more than three bubbles) which configuration corresponds to the global minimum of
energy, even in the simple case where all the bubbles are exactly the same size (see
the shaded section below).
Global minimization and the Weaire–Phelan structure

What is the global minimum of surface area, and therefore the global energy min-
imum, of a foam in which the bubbles are all exactly the same size? This is an
elegant mathematical problem, made difficult because all candidates have compar-
able surface area (see eq. (2.19)). The theoretical lower limit is a sphere, which has
S/V 2/3 ≈ 4.836.
In three dimensions, if all the bubbles have exactly the same shape, they have the
same pressure and consequently every face has zero mean curvature. An idealized
bubble with zero mean curvature (flat faces), defined on p. 44, which cannot be
realized in practice, has f = 13.4 and n̄ = 5.1; its surface area is S/V 2/3 = 5.254,
a value which is probably impossible to beat. The best real bubble is conjectured
to be the Kelvin cell (see shaded section on p. 50), with 14 faces and an average of
5.14 edges per face; its surface area is S/V 2/3 ≈ 5.306, believed to be the smallest
possible value for a bubble which tiles space on its own.
Denis Weaire suggested that it might be possible to further lower the surface
area by introducing as many pentagonal faces as possible (there are none in the
Kelvin cell!), so that each face has close to n = 5.1 edges (and not just on aver-
age) and is consequently not very curved. As bubbles with only pentagonal faces
are unable to tessellate space on their own, it is necessary to mix bubbles of
at least two different shapes (but, in this case, with the same volume). Weaire
proposed candidates inspired by known crystalline structures, and his student
Phelan numerically simulated them using the Surface Evolver software. Their res-
ult, known as the Weaire–Phelan structure (fig. 5.13), has S/V 2/3 ≈ 5.288. This
structure, which inspired the architect of the swimming pool for the 2008 Beijing
Olympics, is the best result known for bubbles of the same volume. Does any-
thing better exist? We don’t know how much time will be required to answer this
question . . .
In two dimensions the ideal bubble can be realized. It is possible to tile the plane
with a bubble without curved edges: a tiling of regular hexagons, the hexagonal
honeycomb, is the least perimeter division of the plane into bubbles with equal
area (fig. 4.10A). This was known to the Romans, more than two thousand years
ago, as the solution found by bees to fill a beehive with cells using the smal-
lest possible amount of wax (fig. 1.5b), but was only rigorously demonstrated in
2001 [31].
3.2.3 Bubble shape Plateau’s laws (see p. 30) are a necessary and sufficient con-
dition for a collection of bubbles to form an equilibrium foam structure, but there are
many different possible structures. In theory, bubbles of the same volume can have
(a) (b) (c)
Fig. 2.13: In 3D, regular bubbles are possible only for a few values of f . For example:
(a) f = 4, tetrahedron, (b) f = 6, cube, (c) f = 12, dodecahedron. The value of
S/V 2/3 for these bubbles decreases with f and lies between about 5.3 (typical bubble,
eq. (2.19)) and 4.83 (a sphere) [17].
very different surface areas, but simulations suggest a more limited range of shapes,
which can be expressed in three equivalent ways [41]:
S 3.3
S ≈ 5.3 V 2/3 ± a few %, ≈ , RS ≈ 1.05RV , (2.19)
V RV
with RV = (3V /4π)1/3 , the radius of a sphere with the same volume as the bubble,
and RS = (S/4π)1/2 , the radius of a sphere with the same surface area as the bubble.
Regular bubbles, in which the faces are all identical (fig. 2.13), have a slightly lower
value of S/V 2/3 .
The same can be said of 2D bubbles, where
P ≈ 3.72A1/2 ± a few % ≈ 6.6 RA , (2.20)
with RA = (A/π)1/2 and P the perimeter of the bubble. Highly deformed bubbles,
which would have a very high ratio of perimeter to square-root of area, are almost
never observed.
For regular 2D bubbles (those in which all the sides are identical, fig. 2.14), the
relationship (2.20) is true to within less than 1% (see fig. 2.14b and exercise 7.5).
This result is specific to foams, and is not true, for example, for polygons with straight
edges and variable angles, for which the ratio P/A1/2 varies greatly with n (it is equal
to 4 for a square, 3.72 for a hexagon, and 3.5 for a polygon with very large n, that
closely resembles a circle).
3.3 Topology and geometry

3.3.1 Geometry–topology correlations Experimentally, we observe that bubb-
les which have more neighbours are generally the largest. That is, edge lengths , face
areas S, and bubble volumes V, each normalized by the average bubble size to the
appropriate power, are correlated with the number of faces f . In 2D, this correlation
is often characterized by the Desch–Feltham relationship, which gives the bubble
perimeter in terms of its number of neighbours [56], or Lewis’ relationship, which
gives the surface area of the bubbles in terms of its number of neighbours [54]. In
3D, the curves V f (f ), Af (f ), and f (f ), obtained from numerical simulations
where measurement is easier than in experiments (fig. 2.15), are not quite linear and
Dry foams 39
(a) (b) 3.78
3.77
3.76
3.75
/A1/2
3.74
3.73
3.72
3.71
3 5 7 9 20 infinity
n
Fig. 2.14: Regular 2D bubbles with sides of equal lengths. (a) Bubbles with the same
edge length superimposed one on top of another to compare their areas and curvatures
(n = 3, 4, 5, 6, 7, and 8). Simulations by S. J. Cox. (b) Perimeter P/A1/2 for regular
2D bubbles with n sides [28], see exercise 7.5.
different fits are proposed in the literature. Nonetheless, it is significant that these
curves are strictly monotonic.
This result is true for the whole foam on average (fig. 2.15), but not for each
individual bubble. It is currently empirical: we are unable to accurately predict the
statistical correlation between bubble size and number of faces, although several mod-
els have been proposed [25, 49]. That the structure minimizes its energy is not sufficient
to understand this correlation: recall that a foam is in a local, not a global, energy
minimum, so to predict this correlation requires that we determine which “type” of
local minimum is adopted most often, and this turns out not to be the global min-
imum. Rivier [56] suggested an explanation for this correlation by assuming that the
value of the energy in a local minimum has no influence on the probability of being
adopted by the foam, and that the foam can explore many different configurations.
More recently, de Almeida [21] reconciled the minimization of surface energy with
the maximizing of disorder during mixing by defining the analogue of a free energy
for foams. The minimization of this energy functional enables the distribution of sizes
and numbers of neighbours of bubbles, as well as their correlations, to be predicted
statistically (fig. 3.24).
There are different definitions of “disorder” in a foam. We can use a topological
definition, for example μf2 (see p. 250), or a geometric one, for example μV2 (defined in
a similar way by replacing the number of faces with the bubble volumes). These two
measures of disorder usually vary in the same way from one foam to another [25, 54]:
they both measure the difference between a given structure and a very organized
structure. In a regular pattern in which all the bubbles are identical—the honeycomb
in 2D and the Kelvin foam in 3D—both measures are identically zero. Nevertheless,
it is possible to construct examples in which the two measures are very different. For
(a) (b)
2.0
4
<S>f /<S>
1.5
3
<V>f /<V>
1.0
2 0.5
1 0
6 8 10 12 14 16 18 20
0 Number of faces
5 10 15 20 25
Number of faces
(c)
2.0
1.5
< >f /< >
1.0
0.5
5 10 15 20
Number of faces
Fig. 2.15: Correlations between topology and bubble size in 3D [38] (Desch–Feltham
and Lewis correlations). (a) Relationship between bubble volume and number of faces;
(b) the same for surface area; (c) the same for edge lengths. To improve the statistics,
the curves are averages taken from several images of a coarsening foam (during which
these curves change very little, see §3.2.3, chap. 3). The continuous line is an average
over all bubbles with given f (denoted ·f ), normalized by the average value for all
bubbles; the bars show the standard deviation for each value of f .
example, a 2D foam in which the bubbles are hexagonal but with very different sizes has
a much greater geometric disorder than a topological one. Conversely, a monodisperse
2D foam can be made from heptagons and pentagons, with a much smaller geometric
than topological disorder.
3.3.2 Number of neighbours, curvature, and pressure in 2D We describe

a circuit around the perimeter of the 2D bubble shown in fig. 2.16. The direction of
motion keeps changing: first, from a to b we follow an arc of a circle: if R is its radius
and its length, the angle through which we turn is /R. We then arrive at a vertex,
which has internal angle 2π/3; the angle through which we turn from b to c is called
the external angle, equal to π − 2π/3, or π/3.
Continuing our circuit, we add the rotations c to g and ij /Rij = κij ij along
each edge ij (separating bubble i from bubble j). By convention, the radius Rij and
Dry foams 41
f e
d
a
g c
2π/3
π/3
b
Fig. 2.16: A circuit around a bubble in 2D. We label this bubble i. It has n neigh-
bours, here n = 5, labelled j = 1, . . . , n. The edge separating bubbles i and j has length
ij , radius of curvature Rij , and curvature κij = 1/Rij . In this example, all the Rij
are positive (bubble i is convex). Photograph courtesy of S. Courty [29]. Reproduced
by kind permission of Springer Science + Business Media.
curvature κij of the edge (ij) are counted as positive if the centre of curvature is
towards i and negative otherwise (see fig. 2.16). Therefore Rij = −Rji and κij = −κji .
Finally, we come back to our point of departure, a, after making a complete turn of
2π. Then

n
π
κij ij + n = 2π, (2.21)
j=1
3
since we added π/3 at each of the n vertices. The quantity
n
q= κij ij (2.22)
j=1
is an angle. More precisely, it is the angle at which it is necessary to curve the sides
of a polygon so that they meet at an angle of 120◦ and it can be closed. It is also
called the geometric charge, and is proportional to the topological charge qt = 6 − n
(see §3.1.2) which also describes the difference from an ordered foam. It is also used in
other contexts, for example to describe the defects (disclinations) in a crystal or the
curvature of space.
The definition in eq. (2.22) allows us to rewrite the relationship (2.21) in a form
that links the geometry, on the left, and the topology, on the right:

n
nπ π π
q= κij ij = 2π − = (6 − n) = qt · (2.23)
j=1
3 3 3
This describes the link between the number of sides of a bubble and its shape. A
bubble with 2, 3, 4, or 5 sides has, on average, positive curvatures and consequently
convex sides (fig. 2.14a). Bubbles with two sides are rare: they generally arise when
a bubble with three sides loses one through film rupture, and we will disregard them
in the following. Bubbles with one vertex are forbidden by Plateau’s laws, since this
would require its single side to curve back on itself with an angle of 120◦ , but that’s
impossible for an arc of a circle.
For bubbles with n > 6 sides, on the other hand, the geometric charge is negative,
and on average the sides are concave (fig. 2.14a). There is no upper limit for n, which
may reach 40 or more.
A bubble may have all sides flat, i.e. without curvature (κij = 0), and then it must
be a hexagon (n = 6, q = 0), whether or not it is regular. The converse is not true: a
hexagon may have both convex and concave sides (as long as the sum of κ remains
zero). n
If the geometric charges of all the bubbles are added, the sum i q = i,j=1 κij ij
counts each side twice (in the form (ij) and (ji)). Since ij = ji and κij = −κji , the
terms cancel each other out pairwise (except for the bubbles situated at the edge of a
foam, see the shaded section on p. 43) and we recover the result (see p. 32) that in a
(sufficiently large) dry 2D foam the average of the geometric (or topological) charge
is zero.
In other words, there must be enough bubbles with 3, 4, or 5 sides, in which the
charge is positive, to compensate for the negative charge of bubbles with more than
6 sides. We observe that in a disordered foam, the value most represented is n = 5,
even if the average number of sides is n = 6, because there can be faces with much
more than n = 6 sides, but not much less.
Moreover, the pressure difference Pi − Pj across an edge balances its curvature
κij (the Young–Laplace law, eq. (2.50)) and we obtain a fundamental relationship
for 2D foams, discovered in 1952 by the mathematician John von Neumann (1903–
1957) [56, 60]. It links the pressure in a bubble to its shape and to its number of
neighbours:

n
e
n
π
(Pi − Pj ) ij = κij ij = (6 − n) = q, (2.24)
j=1
2λ j=1
3
where e is the thickness of the 2D foam and λ the line tension (see p. 59). Bubbles with
positive geometric charge (3, 4, or 5 sides) consequently have a higher pressure than
the average (weighted by perimeter) of their neighbours. On the other hand, bubbles
which have many sides have lower pressure compared to their neighbours. This plays
a fundamental role in foam coarsening (see §3, chap. 3). In a way, the pressure is
analogous to an electrostatic potential: it is maximal on bubbles which have a positive
charge, and minimal on those with a negative charge. It is the curvature of the edges
which allows us to visualize pressure differences between bubbles [28].
It is difficult to say which of the pressure, the geometric charge, and the curvature
is the cause of the other two. Instead, we express the condition of equilibrium, and
therefore the balance between all three, as the configuration of a foam which (locally)
minimizes the total perimeter at fixed bubble areas.
Dry foams 43
All of the above is based on the fact that you turn an angle 2π when traversing
a closed circuit. This is a very simple two-dimensional statement of a more general
theorem in electrostatics, due to the mathematician Gauss, which has numerous im-
plications for foams (2D and 3D). Rivier showed [3] that if you take any region of a
2D foam containing several bubbles, the sum of κ around the periphery of this region
is always equal to the sum of the geometric charges contained within it [28]. The sum
of the geometric charge over the whole foam is 6 (see shaded section below).
Geometric charge of a 2D bubble touching a wall

The argument associated with fig. 2.16 is equally applicable to a bubble in contact
with a wall. In this case, two of the edges meet the wall at a right angle (see
p. 30). We recover eq. (2.23) if the wall is counted as two sides [28], i.e. a bubble
is considered to have n sides if n − 2 of them are free. It therefore has n − 3 free
vertices with external angle π/3 (see p. 40) and two vertices in contact with the
wall with external angle π/2, and eq. (2.21) becomes
n−2
ij π π
+ (n − 3) + 2 = 2π. (2.25)
Rij 3 2
j=1
Thus, a (pentagonal) bubble which has 4 free sides and one which touches a straight
wall has zero geometric charge and all its sides can be straight. A (hexagonal) bubble
which has 5 free sides and one which touches a straight wall has negative geometric
charge and is concave.
With this convention, the sum of the geometric charges (eq. (2.22)) over the
whole foam is 2π, both for a foam enclosed in a box (fig. 5.12b) and for a free
foam (fig. 2.12) [28]. Equivalently, the total topological charge of a foam is 6, so
the average topological charge is 6/N and the constant in eq. (2.8) is χEuler = 1.
3.3.3 Number of neighbours, curvature, and pressure in 3D The topo-

logical charge can be generalized to three dimensions [58], as can the geometric
charge:

n
Hij Sij
q= 1/3
, (2.26)
j=1 Vi
using the fact that each face is a surface of constant mean curvature (H = 1/R1 + 1/R2 ,
see p. 27). Conventionally, curvatures 1/Rij,1 and 1/Rij,2 are counted as positive if
the centre of curvature is closer to bubble i (see p. 27). The product Hij Sij has the
dimension of a length, so is normalized by bubble volume Vi to the power 1/3.
When defined in this way, the geometric charge doesn’t play exactly the same role
as in 2D. In particular, the von Neumann relationship between the curvature of the
faces of a bubble and its number of neighbours (eq. (2.24)) has no equivalent in 3D,
since:
• Hij = −Hji and Sij = Sji but Vi = Vj , so even in the case of a very dry foam
without walls, the terms of eq. (2.26) do not cancel each other out pairwise
on neighbouring bubbles (as it does in 2D, see p. 42), except in the case of
monodisperse foams (where all the Vi are equal).
• There are no bubbles with all faces planar, which would provide a reference shape,
as the hexagon does in 2D. Since in 3D the edges meet at an angle θa ≈ 109.5◦ ≈
1.91 radians (see p. 30), this would impose on a flat face with n straight edges the
condition n(π − θa ) = 2π (cf. §3.3.2), so that n = 2π/(π − θa ) ≈ 5.104 . . . [4, 35].
That is, this face would have a non-integer number of edges! Nonetheless, we
can determine some of the properties of this idealized bubble with exactly zero
curvature. Since the average number of edges per face n and the number of faces
per bubble f are related (eq. (2.13)), it would have f = 12/(6 − n) ≈ 13.4 faces.
This is also not a whole number, but it is very close to the average number of
faces observed in real foams (see p. 33) [4].
• The geometric charge of a bubble does not depend only on its number of faces, i.e.
two bubbles with the same f do not necessarily have the same q. This observation
is particularly significant for coarsening (see §3, chap. 3). We observe in practice
that the differences are small: on average, bubbles which have less than 12 faces
are rather convex (q > 0), those which have more than 15 faces are rather concave
(q < 0), and the geometric charge of bubbles with f faces varies little (fig. 2.17).
Nonetheless, if a bubble has faces which all have approximately the same num-
ber of edges and the same shape, we observe that it has a lower q (it is more
concave) than a more asymmetric bubble with the same f (cf. the dashed line in
fig. 2.17).
• The function q(f ) is not linear (fig. 2.17), unlike q(n) in 2D. By assuming that
bubbles are almost regular and that the faces are not very curved, Mullins was
able to derive analytically the approximate formula [51]:
1/3
3
qMull (f ) ≈ − G1 (f ) G2 (f ),
4π

π (f − 1)1/2
with G1 (f ) = − 2arctan 1.86 ,
3 f −2
−1/3
f − 2 3
and G2 (f ) = 5.35 f 2/3 1/2
− G1 (f ) . (2.27)
2 (f − 1) 8
This expression changes sign at f = 13.3; for large f it increases (in absolute value)
as the square-root of f . The central role of q in foam coarsening (see fig. 3.22
and eq. (3.30)) has motivated an intensive search for a simple expression for q(f )
(see [38]). Theoretical calculations focus on regular bubbles [33], while numerical
simulations give detailed information about the equilibrium configuration of a
foam [42] (or even a single bubble, in order to reduce the calculation time [17]).
Wet foams 45
1
<q>f
–1
4 6 8 10 12 14 16 18 20 22 24 26
number of faces, f
Fig. 2.17: Average geometric charge of 500 bubbles in a dry 3D foam as a function
of the number of faces, obtained with Surface Evolver [38]. The geometric charge q of
a bubble is the mean curvature integrated over the surface, normalized by V 1/3 (see
eq. (2.26)), and qf is its average for bubbles with f faces. The error bars represent
the standard deviation for each f and the dashed line shows the theory for regular
bubbles, eq. (2.27).
These different approaches confirm Mullins’ law for 3D dry bubbles (eq. (2.27)
and fig. 2.17). More recently, MacPherson and Srolovitz [45] showed that
n
A
Hij Sij = 2πL − π , (2.28)
j=1
3
where L is a (complicated but exact) function of the shape of a bubble, with a

value close to twice its diameter, and A the sum of the lengths of its edges. The
geometric charge can be deduced from this expression (see eq. (2.26)).
4 Wet foams
We now describe the equilibrium configuration of a foam which contains a non-
negligible quantity of liquid, so that the parts of an ideal foam (films, edges, and
vertices) must all inflate to accommodate foaming solution. We determine the distri-
bution of liquid in such a foam from the equilibrium laws established in §2 and we
show that this equilibrium can be established even in the presence of gravity.
4.1 Modification of the structure

The quantity of liquid in a foam is characterized by the liquid fraction φl = Vl /Vfoam
(see p. 21). A foam which has just been made usually has φl
1−10%, and its liquid
fraction decreases over time because of gravity-driven drainage (see §4, chap. 3). To
understand the structure of a wet foam, we do the opposite: we take a dry foam
structure and add solution.
Fig. 2.18 shows how the liquid modifies the structure of a foam. As long as φl is
small, typically φl < 5%, we can describe a wet foam in terms of:
• bubbles, with gas volumes equal to those in the reference dry structure;
• soap films, the surface area of which decreases as φl increases, so that their
thickness h, and therefore their volume, remains negligible;
• Plateau borders, the liquid channels which form between the films, replacing the
edges in a dry foam (see figs. 2.3 and 2.19), with characteristic length and
cross-sectional area proportional to rP2 B , which will be determined below;
• vertices, the junctions of four Plateau borders, of characteristic volume rP3 B , which
replace the vertices of a dry foam.
At higher liquid fractions, when rP B ∼ /2, neighbouring vertices meet and the
Plateau borders are no longer clearly defined. The foam then becomes a collection
of almost-spherical bubbles which are slightly deformed where they press against one
another.
Above a critical liquid fraction, denoted φ∗l , the bubbles no longer touch each other.
When φl > φ∗l , we no longer speak of a foam, but rather of a suspension of bubbles,
or a bubbly liquid. The value of φ∗l depends on the structure (see fig. 4.17).
4.1.1 Validity of Plateau’s laws In the rest of this section we consider a foam
with low but non-vanishing liquid fraction, i.e. with h rP B /2 (the film thickness,
Plateau border size, and Plateau border length respectively). Plateau’s laws should
dry foam wet foam
Fig. 2.18: When liquid is added to a dry foam (left, initial state) it accumulates
mainly in the vertices and Plateau borders (right, final state). Photographs courtesy
of F. Elias.
Wet foams 47
(a) 120° (b)
Plateau
borders
Film
Vertex Plateau
Axis
border
r1 = r r2
rPB = r
109, 5°
Fig. 2.19: (a) A vertex in a moderately wet foam. (b) A Plateau border, indicating
the notation used. The lengths r1 and r2 are the radii of curvature of the liquid/gas
interfaces and rP B is the length of one side of the cross-section of the Plateau border.
We show in the text that rP B = |r1 |, denoted r in the following.
remain valid, at least approximately (see fig. 2.19a). If the three thin films attached
to each Plateau border are extrapolated inwards they almost meet, at an angle close to
120◦ , along the Plateau border axis (see fig. 2.19b). Similarly, the axes of four Plateau
borders almost meet at a point that we call the centre of the vertex, and intersect
at close to 109.5◦ . The structure of this foam of non-negligible liquid fraction is thus
very close to that of an ideal dry foam, with a simple thickening of the edges. In two
dimensions these results are exact (see the decoration theorem on p. 61). In three
dimensions we could be more precise about the angles by taking into account a line
tension which acts along the axis of the Plateau borders, and whose value depends on
the liquid fraction [26, 40].
4.1.2 Plateau border geometry The liquid in a wet foam is confined between
gas/liquid interfaces, so it is the Young–Laplace law (eq. (2.3)) which imposes the local
curvature and determines the shape of the Plateau borders and vertices. If we ignore
gravity (§4.3), at equilibrium the pressure p of the liquid is uniform (the liquid phase
is continuous) and the Young–Laplace law is

1 1
P −p=γ + , (2.29)
r1 r2
where P is the bubble (gas) pressure and r1 and r2 are the radii of curvature defined in
planes perpendicular and parallel to the axis of the Plateau border. r1 is of the order
of the Plateau border size rP B , and r2 is of the order of the radii of curvature of the
soap films. The difference in pressure between the two phases is the capillary pressure,
denoted Pc . For the values of φl considered here, the Plateau borders are sufficiently
long that r1 |r2 |. Writing r = r1 , eq. (2.29) thus simplifies to
γ
Pc = P − p
· (2.30)
r
The film curvatures are at most of the order of 1/ and the pressure variations from
one bubble to the next are consequently at most of the order of γ/ γ/r. We can
deduce from this that in the absence of gravity, for foams of moderate liquid fraction,
the radius of curvature r of the Plateau borders is almost uniform within the foam.
Moreover, the pressure difference between the liquid and the gas is much greater than
the pressure differences between bubbles, and the lower the liquid fraction, the lower
the pressure of the liquid phase. For r = 0.1 mm and γ = 50 × 10−3 N.m−1 , we find
P − p = 500 Pa.
To a good approximation (see §4.1.4), the corners of the Plateau borders meet
the films tangentially, the angles between the films are 120◦ , and the three radii of
curvature r1 are equal. So the triangle in which the cross-section of a Plateau border
is inscribed is equilateral with side length rP B = r (see fig. 2.19b). The area of the
Plateau border cross-section is
√ π 2
s= 3− r . (2.31)
2
4.1.3 Film geometry The curvature of the films is smaller than that of the
Plateau borders, which are therefore at a lower pressure than the films. Then ca-
pillary suction acts to suck much of the water from the films into the Plateau borders.
Nevertheless, equilibrium must be reached, and in order to understand this we need
to look at a smaller scale, the thickness of a film, of several tens of nanometers, at
which the interfaces of the film begin to interact (see §2.3.1, chap. 3). This interaction
is associated with a new constraint, oriented perpendicular to the interfaces, known as
the disjoining pressure [8, 36]. Its value, and especially its sign, provide information
about the interaction of the two interfaces: a positive disjoining pressure is a result of
interfaces which repel each other. As discussed in §1.2, this is a necessary condition
for films to be stable (and therefore for foams to exist).
If we neglect film curvature, the pressure in the gas, P , is the same on each side
of the film. In the liquid, both the disjoining pressure Πd (see fig. 2.2a) and the bulk
liquid pressure p act. The balance of normal forces on an interface is then
Πd (h) = P − p, (2.32)
where h is the film thickness. Equilibrium is reached when h is such that the jump
in pressure at the interface, P − p, is the equal to the jump Pc at the Plateau border
surface (eq. (2.30)), thus
γ
Πd (h)
. (2.33)
r
The variation of Πd with h will be presented in §2.3, chap. 3. The equilibrium thick-
nesses are very small, of the order of tens of nanometres. Most of the liquid phase is
Wet foams 49
therefore situated in the network of Plateau borders. The liquid fraction then fixes r
(see shaded section on p. 50) and we deduce h from eq. (2.33).
4.1.4 Transition from a film to a Plateau border The surface tension γf of

a film is the energy necessary to increase its surface by one unit of area. We have
accepted up until now that this tension has a value of 2γ (see §2.1.2). A careful
thermodynamic calculation (see [8, 36]) shows however that γf depends slightly on
the disjoining pressure through the relationship
∞
γf (γ, h) = 2γ + Πd (h ) dh + hΠd (h). (2.34)
h
Since Πd decreases quickly when h becomes large, both correction terms vanish for
thick films. But for a thin film the surface tension γf is slightly lower than 2γ. The
magnitude of the correction can be estimated by assuming that Πd ∼ γ/r for lengths of
the order of h ∼ 10−7 m. For r
10−4 m, we find (2γ − γf )/γ
h/r
10−3 . The rel-
ative variation in γ is consequently very small and, in this book, we do not distinguish
γf from 2γ.
Nevertheless, the difference between these two values is large enough to create an
observable angle θmacro between the film and the Plateau border where they meet.
The mechanical equilibrium of this “transition region” is illustrated in fig. 2.20. We
determine the balance of forces along a line tangent to the film and perpendicular to
the axis of the Plateau border (horizontal direction on the figure). To the left of the
transition region the film pulls with a tension γf (h); this is quite far from the Plateau
border so the film thickness is constant and equal to h. On the right, the thickness
of the film is quite large so that Πd is negligible and the Plateau border pulls with
a tension 2γ cos θmacro . We deduce that cos θmacro = γf (h)/(2γ). With the previous
numerical values, this suggests cos(θmacro ) = 0.999 and θmacro ∼ 3◦ .
In the transition region neither the curvatures nor the interactions between the
interfaces can be neglected. Looking more closely, we observe that a film and a Plateau
border join tangentially. We can estimate the length of the transition region, ξ, by
dimensional analysis. In this region, the order of magnitude of the curvature is h/ξ 2 ,
θmacro
Plateau
Film Transition
border
region
r
10 nm 1 μm 100 μm
Fig. 2.20: Transition from a film to a Plateau border. From a distance, the Plateau
border appears to intersect the film at an angle θmacro along the contact line. A more
precise examination of this line reveals a “transition region” in which the Plateau
border curvature varies rapidly, owing to the disjoining pressure. When viewed at the
film scale, the Plateau border thus meets the film tangentially.
√
which must equal the curvature of the Plateau border 1/r. Then ξ
hr, which
corresponds to a length of the order of a micron, i.e. much greater than h but smaller
than r.
The Kelvin cell: an example of an ordered foam

An ordered monodisperse foam with a liquid fraction less than 6.3% forms the
Kelvin structure (body-centred cubic, bcc).
In the dry case, the Kelvin cell has 8 hexagonal faces, 6 square faces, and
36 edges, each of length (see fig. 2.21 and experiment (6.4)). Its surface area
and volume are
√
3 3 2 √
SK = 8 + 62 , VK = 8 23 . (2.35)
2
The radius of a sphere of the same volume is
√
RV = (6 2/π)1/3 . (2.36)
The characteristics of a Kelvin cell with low but finite liquid fraction are obtained
from the dry case and from the size r of the Plateau borders. The volume of a
(a) (b)
Fig. 2.21: The Kelvin cell is the unit cell of an ordered monodisperse foam with
low liquid fraction. (a) Dry Kelvin cells (shown schematically with straight edges
and flat faces). The arrows represent the three basis vectors of the body-centred
cubic (bcc) lattice. Each bubble is inscribed in a cube. It shares one six-sided face
with its 8 nearest neighbours, placed at the vertices of the cube (vector u), and one
four-sided face (within each face of the cube) with its 6 nearest neighbours, placed
at the centre of the neighbouring cubes (vectors v and w).
(b) Kelvin cells for liquid
fractions of 1% (top) and 3% (bottom), obtained with the Surface Evolver.
Wet foams 51
√
Plateau border is roughly s, where s = ( 3 − π/2)r2 is the cross-section of the
Plateau borders (see fig. 2.19b). There are 12 Plateau borders per bubble (36 edges
shared by 3 bubbles), with a total volume VP B = 12s ≈ 1.94r2 .
The volume of a vertex can be approximated by considering four spheres of ra-
dius r packed tetrahedrally. The√ tetrahedron connecting the sphere centres has
side length 2r and volume 2 2r3 /3; it overlaps each sphere over 1/24 of its
3
volume,
√ 3 that is3 πr /18. So the volume of the cavity between the four spheres is
2 2r /3 − 4πr /18. For all 6 vertices of a Kelvin cell, the total volume is
Vn = 1.47 r3 . A more precise numerical calculation in Surface Evolver (see §2.1.2,
chap. 5) suggests that the volume of the Plateau borders and vertices is equal
to VP B + Vn = 1.935 r2 + 2.2627 r3 . The result is a little higher than our ap-
proximation because the surface of a vertex is not really part of a sphere of
radius r.
A rough estimate of the volume of liquid in the films, not taking into account the
reduction in surface associated
√ with the non-zero size of the Plateau borders, gives
Vf = hSK /2. If r h then this volume is negligible compared to the volume of
the Plateau borders. The liquid fraction is then approximately
r2 r3
φl
(VP B + Vn )/VK
0.171 + 0.2 · (2.37)
2 3
For φl < 1%, we have r and it is reasonable to neglect the second term of this
expression:
r2 r2
φl
0.171 2

0.33 2 ; so r
RV 3φl . (2.38)
RV
These results are quite close to those predicted by numerical simulations of

disordered foams, and they serve as a reference throughout the book.
4.2 Osmotic pressure

If a solution (of sugar, for example) is separated from its solvent by a semi-permeable
membrane, a flow of solvent is established towards the solution. The pressure that
must be applied to arrest this flow is called the osmotic pressure of the solution. As
the foam is a “solution” of bubbles in water, we can ask if a similar phenomenon occurs
in foams. We show here that it does, but for a very different reason.
4.2.1 Physical origin Princen first demonstrated the existence of an osmotic pres-
sure in concentrated emulsions and foams [53]. In contrast to the case of solutions, for
which this pressure is due to the entropy of the solute, in foams it arises because a wet
foam has lower surface energy than a dry foam.
We consider a foam in contact with a reservoir of soapy water at a semi-permeable
membrane through which liquid but not bubbles can pass (fig. 2.22a). In order to
(a) (b) 10
Patm
–1/2
Foam 1
Π0/(γ/RV)
Membrane
p 0.1
Foaming p h
solution
0.01
0.01 0.1 ϕ
ϕ
Fig. 2.22: (a) Diagram of an experiment measuring the osmotic pressure Πo of a

foam using a U-tube connected to the solution reservoir (Πo = ρl gh). (b) The change
in Πo with liquid fraction for an ordered monodisperse foam with bubble radius
RV = 150 μm, γ = 37 mN.m−1 , and φ∗l = 0.26 (see fig. 4.17). The continuous
line is eq. (2.46) and the line of slope − 12 shows the limit for dry foams (eq. (2.45)).
Data from [34].
prevent the liquid being sucked up by the foam, we must impose a pressure difference
Πo between the reservoir (at pressure p) and the atmosphere above the foam (at Patm ):2
Πo = Patm − p. (2.39)
The osmotic pressure Πo is deduced from the work required to extract from the foam
a volume dVl of liquid. Neglecting the gravitational potential energy and the com-
pressibility of the gas, this work is determined by the change in interfacial, or surface,
energy:
−Πo dVl = γ dSint , (2.40)
where Sint is the total area of the interfaces. However, a variation in dVl changes the
liquid fraction dφl according to

Vl (1 − φl )2
dφl = d = dVl , (2.41)
Vl + V g Vg
with Vg the volume of gas in the foam. Then Πo varies with φl according to

2 d Sint
Πo = −γ (1 − φl ) . (2.42)
dφl Vg
2 Note that the difference between the osmotic pressure Π and the capillary pressure P (defined
o c
on p. 48) is small compared to atmospheric pressure Patm .
Wet foams 53
At the volume fraction corresponding to close packing (denoted φ∗l , see fig. 4.17)
the bubbles are spherical, the interfacial energy density γSint /Vg is a minimum, and
consequently Πo (φ∗l ) = 0. As the liquid fraction is lowered, the bubbles deform (they
become polyhedral as φl → 0) and their surface area is therefore greater than for
spheres of the same volume. In this case, the foam sucks up liquid from the reser-
voir, in order to enlarge the Plateau borders and make the bubbles rounder, with
a pressure Πo . The osmotic pressure becomes larger as the foam gets drier (see
fig. 2.22b).
4.2.2 Variation of osmotic pressure with liquid fraction We now consider an

ordered monodisperse foam at equilibrium. If its liquid fraction doesn’t exceed a few
percent, it forms the body-centred cubic Kelvin structure (see shaded section on p. 50).
At higher liquid fractions, it takes a face-centred cubic (fcc) or hexagonal close-packed
(hcp) structure. Given the structure of a foam we can predict the osmotic pressure for
a given bubble size and liquid fraction.
Dry foam In the dry limit, where the Plateau borders are long and slender, the
jump in pressure between the gas in a bubble and the liquid around it depends simply
on the radius of curvature r of the Plateau borders (see §4.1.2):
γ
P −p= · (2.43)
r
The overpressure in a bubble is of the order of 4γ/RV Patm . Consequently, P ≈ Patm

and the osmotic pressure is
γ
Πo = · (2.44)
r
By expressing r as a function of φl using eq. (2.38), we find
γ 1
(bcc) Πo
√ . (2.45)
RV 3φl
A similar result holds for disordered foams. Thus, for a surface tension of 30 mN.m−1 ,
bubbles of radius RV = 150 μm, and a liquid fraction φl = 0.01, we predict
Πo = 1150 Pa, which corresponds to a height of water h = Πo /ρl g of the order of
11 cm, easily measurable using the device shown in fig. 2.22a. Note that eq. (2.45)
agrees with experimental data in the dry limit, φl 1 % (fig. 2.22b) and indicates that
the smaller the bubbles, the more significant the role of the interfaces, and therefore
the larger the value of Πo .
Wet foam In order to predict the osmotic pressure of a wet foam, we first need to
know how the interfacial energy density varies with φl (eq. (2.42)), e.g. from numer-
ical simulations (Surface Evolver, §2.1.2, chap. 5). As φ∗l is approached, the osmotic
pressure falls rapidly: fig. 2.22b shows that Πo decreases by an order of magnitude
when φl doubles (around φl ∼ 0.1). This happens because the contact area between
adjacent bubbles tends to zero at φl = φ∗l (cf. the shaded section on p. 197). At
equilibrium, an ordered wet monodisperse foam has a face-centred cubic configuration
and so φ∗l = 0.26. For φl ≤ φ∗l , the measurements and simulations of osmotic pressure
are well-described by the empirical expression
γ (φl − φ∗l )2
(fcc or hcp) Πo (φl ) = 7.3 √ , (2.46)
RV φl
shown in fig. 2.22b. This expression coincides with eq. (2.45) in the dry limit.
Thus given the bubble size and the surface tension, measuring Πo indicates the
liquid fraction of the foam. This expression also describes the osmotic pressure of
concentrated emulsions, with φl corresponding to the volume fraction of the continuous
phase [47, 53].
4.3 Role of gravity

We have seen that the geometry of the liquid-carrying parts of a foam is dictated by
the liquid pressure, and have until now assumed that the liquid pressure is uniform
throughout the foam. In fact, at equilibrium the pressure in the liquid is determ-
ined by hydrostatics: as a function of height z, we have p(z) = p(0) − ρl gz. On the
other hand, the film curvatures are not modified by the presence of the liquid, and
so the bubble pressures are not sensitive to liquid fraction or gravity, and can be
assumed to be equal, on average, to a constant P , with weak variations compared
to P − p.
Eq. (2.30) links liquid pressure to Plateau border curvature and therefore radius
of curvature r to height:
γ γ
= P − p(0) + ρl gz = + ρl gz. (2.47)
r(z) r(0)
Since r is connected to liquid fraction (through eq. (2.38)), we have

√
−1/2 −1/2 3z RV zd
φl (z) − φl (0) =
2, (2.48)
λ2c λc

where d ∼ 2RV is a characteristic bubble diameter, λc = γ/ρl g is the capillary
length, and φl (0) is the liquid fraction at the bottom of the foam. The latter is equal
to φ∗l , the liquid fraction of a packing of spherical bubbles (see fig. 4.17). Eq. (2.30)
is valid for dry foams and should not be applied at the bottom of a foam, although
any error in liquid fraction is restricted to this part of the profile. Eq. (2.48) is known
as the equilibrium liquid fraction profile. This profile is shown in fig. 2.23. The max-
imum equilibrium foam height is fixed by the disjoining pressure of the films and the
pressure in the Plateau borders.
2D and quasi-2D foams 55
0.10
0.08
Liquid fraction
0.06
0.04
0.02
0
0 0.02 0.04 0.06 0.08 0.10
Height, z(m)
Fig. 2.23: The equilibrium profile of liquid fraction (see fig. 0.1). The liquid frac-
tion, obtained from eq. (2.48) (with γ = 40 mN.m−1 and d = 2 mm), changes from
φ∗l at the bottom of the foam (z = 0) to a value close to zero at the top of the
foam.
5 2D and quasi-2D foams

In this section we present detailed information on bubble monolayers (quasi-2D foams).
We describe their structure, including the 3D aspects (§5.1), a strictly 2D model of their
equilibrium structure (§5.2 and §5.3), and aspects of rheology which are specifically
2D (§5.4).
A foam usually extends in three directions in space. In a foam that is not too
polydisperse, it is possible for the bubbles to be arranged in a single layer (fig. 2.24).
These bubble monolayers are obtained by forcing bubbles to spread out over a (gen-
erally flat and horizontal) surface, or between two surfaces; the three geometries used
are represented in fig. 2.25a–c. We talk about a bubble monolayer or a quasi-2D foam
when each bubble touches both the upper and lower surfaces.
A horizontal quasi-2D foam is not significantly affected by drainage (see §4,
chap. 3), and when protected from evaporation (fig. 2.25a and 2.25b) shows little
coalescence (see §5, chap. 3). As quasi-2D foams are much more stable than 3D foams,
and also much easier to visualize, they are widely used as model systems for research.
Numerous results obtained in 2D are equally valid in 3D, at least qualitatively.
Viewed from above (see fig. 2.24), only the boundaries between two bubbles are
visible: they form a continuous network of curved lines, easily found by image ana-
lysis (see §3.1, chap. 5). Structural information is purely two-dimensional, and this is
(a) (b)
Fig. 2.24: 2D foams. (a) Langmuir foam (pentanoic acid on the surface of water)
observed in polarized light at the Brewster angle (bubbles of around 50 μm). The
reflected intensity cancels out on a bare interface, and 2D bubbles, which are not
very dense, then appear darker than the dense continuous phase (see shaded section
on p. 58) [16]. Photograph courtesy of S. Courty. (b) Dry foam between two plates,
viewed from above (bubbles of several mm). Photograph courtesy of O. Lordereau [44],
reproduced by kind permission of Springer Science + Business Media.
(a)
(b)
(c)
(d)
Fig. 2.25: Three possible configurations for a 2D foam (after [14]): (a) between two
solid flat plates (Hele–Shaw cell or glass–glass), (b) between the foaming solution and
a solid flat plate (liquid–glass), (c) above the foaming solution (Bragg bubble raft).
(d) The shape of the liquid/gas interface in a foam of type (b), obtained by numerical
simulation [55].
sufficient to interpret most of the observed phenomena. Nevertheless, analysing these

2D images does require knowledge of the complete, three-dimensional, structure of a
2D foam. For example, the shape of a bubble between liquid and plate is shown in
fig. 2.25d.
5.1 3D structure of a monolayer of bubbles between two plates

We next describe the 3D structure of the simplest quasi-2D foam, i.e. a foam between
two plates. It consists of two types of films: wetting films covering the surface of the
plates, and films separating bubbles (simply referred to as films in the following). The
latter appear clearly in fig. 2.26 as elongated rectangles. A more precise examination
shows that the films, when viewed from above, are gently curved. For an isolated dry
bubble between two plates, the cylindrical shape represents a minimum of surface
energy and, by analogy, we conjecture that the films between the bubbles are parts of
a cylinder with axis perpendicular to the plates. That is, the curvature viewed from
the side is zero, but the curvature viewed from above adapts itself to the pressure
difference between two neighbouring bubbles. This property is essential, for it enables
us to relate the bubble pressures to the curvature of the films, as measured on the
images (see eq. (2.50)).
There are also two types of Plateau borders (fig. 2.26). Firstly, the intersections
between three inter-bubble films, which have the same cross-sectional shape as in a 3D
foam and are oriented perpendicular to the plates. In addition, surface Plateau borders
are the intersections between the wetting films and the inter-bubble films that appear
on an image when viewed from above. In a 2D foam between two plates there are two
identical sets of surface Plateau borders: the network that contacts the top plate is
superimposed on the network on the bottom plate (see fig. 2.24b). The cross-section
of these surface Plateau borders is visible in fig. 2.25a. The thickness of the lines
formed by the surface Plateau borders on the images gives a qualitative idea of their
radius of curvature, although the precise relationship is difficult to determine because
the curved walls of a surface Plateau border deflect light (see §1.2.3, chap. 5).
The monolayer is stable if the plates are not too far apart. If the liquid fraction
is significant, or if the plates are very close together, the surface Plateau borders at
the top and at the bottom touch each other, forming a liquid bridge, and the films
disappear. If, on the other hand, the distance, e, between the plates increases, the
(a) (b)
Fig. 2.26: (a) Oblique view of a foam between two plates, illustrating both types of
Plateau border described in the text. Photograph courtesy of I. Cantat. (b) A ferrofluid
foam with slightly higher liquid content, illuminated from below. Photograph courtesy
of E. Janiaud.
bubbles are slightly stretched at first, then finally, beyond a certain distance emax ,
the smallest bubbles become detached from one of the plates and the 2D structure is
lost. For a single bubble of volume V between two plates, we have, in the dry limit,
emax = (πV )1/3 . Beyond that, the cylindrical form becomes unstable, leading to a
hemisphere in contact with one of the two plates [20]. This stability limit decreases
with the size of the surface Plateau borders.
The structure of a fast-flowing foam between two plates is more complicated. In
fact, when surface Plateau borders move along the plates, they experience friction
(figs. 4.13 and 5.16), and so the film curvature viewed perpendicular to the plates is
no longer zero. Then the pressure difference between two neighbouring bubbles is no
longer directly related to the curvature measured on an image.
Truly 2D foams: Langmuir foams

Langmuir foams deserve more than any others to be called two-dimensional since
they are only one molecule thick [43].
Water-insoluble surfactants are dissolved in chloroform and deposited on a liquid
surface to form a Langmuir monolayer. They can exist in different phases: 2D gas
(molecules well-separated and disordered), 2D liquid (molecules in close proxim-
ity and disordered), 2D liquid crystal (molecules with a certain order; also called a
mesophase), 2D solid (resists shearing in the plane), or a 2D crystal (molecules com-
pletely ordered). Under very high pressures, there is an irreversible return to three
dimensions. The surface pressure (or Langmuir pressure) Πl ≡ γwater − γ depends
on the area per molecule am . The isotherm Πl (am ), measured by compressing the
surface with barriers, gives rise to coexistence plateaux between phases, although
less markedly than in 3D.
In the liquid/gas coexistence domain, the monolayer spontaneously organizes
in the form of a 2D foam composed of bubbles of gaseous phase, separated by a
continuous liquid phase (or the reverse, depending on the confinement). A Langmuir
foam coarsens [7], and its structure is directly visible with a microscope at the
Brewster angle (see fig. 2.24a), or by adding fluorescent markers to increase the
contrast between the liquid phase and the gas phase.
One particular property of this foam is that it is not necessary to add a supple-
mentary chemical species to stabilize the liquid interfaces: the electrostatic dipolar
interactions between the surfactants are sufficient. We also find this property in
ferrofluid foams, which are stabilized by magnetic interactions (see §6.1.4, chap. 3).
5.2 A model for a dry 2D foam

The energy of a 2D foam at equilibrium depends only on the shape of the network of
lines viewed from above, a property that enables a model of bubble monolayers (or
Langmuir foams) to be developed.
Energy and line tension In the dry limit, the structure of a 2D foam between two
plates is unaffected by translation in the vertical direction over the distance e which
separates the two plates.
This translational invariance is lost in geometries (b) and (c) of fig. 2.25, and
also in geometry (a) as soon as the surface Plateau borders are no longer negligible.
However, as long as the vertices can be ignored, the curvature r of the surface Plateau
borders (in the plane of their cross-section) can be considered as constant everywhere.
The quantity of interface per unit length of surface Plateau border is thus a well-
defined constant dependent upon the precise shape of the cross-section of the surface
Plateau borders. In this case the energy is

E = 2λ ij = λLint , (2.49)
where λ is the line tension, defined for a single interface analogously to surface tension,
and Lint is the total length of interface when the foam is viewed from above. The value
of λ is approximately γ(e + r(π − 2)) and is slightly greater than γe. In the very dry
limit for a foam between two plates, λ ≈ γe.
Plateau’s laws and the structure of dry 2D foams The behaviour of a 2D

foam is best described in terms of a cross-section through a horizontal cut, as shown
in fig. 2.27. The films are lines, the Plateau borders are small concave triangles and
the surface Plateau borders are no longer taken into account. We can now use a 2D
vocabulary: the line tension 2λ (in Newtons) replaces the surface tension 2γ of a film,
the energy is proportional to the edge lengths, and we refer to bubble area rather than
volume.
We consider first the case of an ideal dry foam (see definition on p. 26). The
principle of energy minimization gives the 2D version of Plateau’s laws (cf. p. 30):
• edges are arcs of circles;
• edges meet in threes at vertices;
• angles at vertices are 120◦ ;
• edges meet solid walls at 90◦ .
Fig. 2.27: 2D representation of a bubble monolayer at different liquid fractions [19].

This mathematical idealization of a 2D foam has given rise to many theoretical predic-
tions and numerical simulations. The three images illustrate the decoration theorem
(§5.3): when the liquid fraction increases, the structure of the foam remains unchanged.
In particular, the film curvatures are not modified as long as the Plateau borders do
not touch each other.
The first property results directly from the equilibrium of pressures along an edge sep-
arating two bubbles. In fact, with Pi and Pj the pressures in two neighbouring bubbles i
and j, e the thickness of the foam, and Rij the radius of curvature of the edge, counted
as positive if the centre of curvature is towards bubble i, the pressure difference is
2λ
Pi − P j = · (2.50)
eRij
Since Pi − Pj is the same everywhere along the edge, its radius of curvature is
constant: it is an arc of a circle.
The third property expresses the equilibrium of a vertex under the effect of three
forces (line tensions) of the same magnitude, tangent to the edges. The instability of
vertices at which four edges meet is dealt with in exercise 7.3. As in 3D (eq. (2.5)),
the algebraic sum of the curvatures of three edges which meet at a vertex is zero [46].
In addition, the three corresponding centres of curvatures are aligned [50].
The fourth property arises from symmetry: it is independent of the actual contact
angle of the foaming solution, which affects only the matching between the Plateau
border and the wall.
5.3 Two-dimensional liquid fraction

The liquid fraction is the ratio of the volume of liquid to the total volume of the
foam: it is well-defined for a foam in a Hele Shaw cell (fig. 2.25a), but no longer
makes sense when the foam consists of a monolayer of bubbles floating on the foaming
solution (figs. 2.25b and c). The lower boundary of the latter systems is not well-
defined and therefore the liquid fraction is no longer a useful quantity (although the
osmotic pressure, §4.2, is still well-defined).
Instead, we define the two-dimensional liquid fraction φl,2D as the area fraction of
liquid in a horizontal cut through the monolayer (fig. 2.27). Unfortunately, this value
depends on the position of the cut, as the cross-section of the vertical Plateau borders
varies with height, more so as the foam becomes wetter. The section that is chosen is
based on the following rule: the minimum edge length at which a T1 is triggered (see
p. 78) must be the same in the real foam as in its idealized 2D representation [55].
This simple rule gives an unambiguous definition of the liquid fraction in a 2D foam
and it enables comparisons between experimental, theoretical, and numerical results.
The actual (3D) liquid fraction of a foam between two plates is always greater than
the two-dimensional liquid fraction, as it takes into account the presence of liquid in
the surface Plateau borders. In the following, the liquid fraction of a monolayer of
bubbles will refer to its two-dimensional liquid fraction and φl,2D will be written φl .
The smallest accessible liquid fractions vary greatly from one geometry to an-
other. All the accessible liquid fractions in 3D are also accessible in a Hele–Shaw cell
(fig. 2.25a): it is sufficient to take a 3D foam with the required liquid fraction and
confine it between two plates. It is possible to reach liquid fractions as low as φl ∼ 10−4
in this geometry. In the liquid glass set-up (fig. 2.25b), the liquid fraction can be re-
duced by increasing the distance between the surface of the solution and the plate,
until bubbles start to detach from the plate around φl ∼ 10−2 . For the bubble raft
(fig. 2.25c), the liquid fraction decreases when the lateral pressure exerted on the
bubbles increases. However, if the pressure increases too much, the bubbles slide on
top of one another and the structure becomes three-dimensional. Low liquid fractions
are inaccessible and φl ≥ 10−1 .
Decoration theorem The decoration theorem states that the 2D structure of a wet
foam is the structure of a dry foam in which each vertex is decorated by a Plateau
border [63]. This implies that if the three edges meeting at a Plateau border are
extended inwards (while respecting their curvature), they intersect at a single point
(the centre of the Plateau border) at 120◦ . In 3D, this is a good approximation (see
§4.1.1) but no longer strictly true.
A 2D Plateau border is the cross-section of a 3D Plateau border, the shape of
which is discussed in §4.1.2. They are small triangles with concave sides, whose radius
of curvature r is fixed by the pressure p of the liquid phase and the pressure P of
the neighbouring bubble according to P − p = γ/r. If all three neighbouring bubbles
have the same pressure, the area s of the Plateau border is given by eq. (2.31). In a
hexagonal network there are two Plateau borders per hexagon, of area Ah , and so the
liquid fraction is
2s √ r2
φl,hex = = 2 3−π · (2.51)
Ah Ah
In this case the Plateau borders come into contact at a critical liquid fraction of 9%
(cf. fig. 4.17).
For a large number of bubbles, the average number of Plateau borders is always
equal to two in a 2D foam (see eq. (2.10)). In a disordered foam, the area s of a Plateau
border remains close to the value given by eq. (2.31), and eq. (2.51) is therefore a good
approximation:
√ r2
φl ≈ 2 3 − π , (2.52)
A
with A the total area of the foam divided by the number of bubbles.
5.4 2D foam flows

Both 2D and 3D quasi-static flows are a priori very similar, and a 2D study is likely to
highlight in a simple way the 3D response. On the other hand, rapid 2D flows are not
a model for 3D because friction with the bounding plates (in a glass–glass or liquid–
glass geometry) is a form of dissipation with no 3D equivalent (see p. 190). In a bubble
raft, where friction is absent, rapid flows are unstable and the bubbles do not stay in
a monolayer. Nonetheless, confined foam flows are interesting, for example for their
applications in microfluidics or in the oil industry (flow in porous and fractured rocks).
Thanks to their geometry, direct visualization of 2D foams gives access to many
flow properties. With the help of image analysis (see §3.2, chap. 5), we can identify
the outline of individual bubbles and follow their movement. We can then find the
stress and strain fields (tensors), strain rates, and time averages of velocity, pressure
(see §3.3.3, chap. 5), and T1 density (see definition on p. 78), as described below. This
information is more difficult to obtain in 3D.
Quasi-static flow around an obstacle Fig. 2.28 illustrates quantities that can
be found through image analysis [24]. Here, a foam confined between liquid and glass
flows continuously along a channel containing a circular obstacle (fig. 2.28a; for the
3D case see fig. 4.34). The bubbles are deformed as they pass around the obstacle
(elongated parallel to the flow downstream and stretched perpendicular to the flow
upstream) and store elastic energy in a way that can be quantified with a texture
tensor (see §3.2.3, chap. 5) derived from the images. The velocity and pressure fields
are shown in figs. 2.28b and 2.28c. The map of T1 density in fig. 2.28d illustrates
the number of T1 events per unit time and unit area and their average orientation.
This measure of the plasticity field shows upstream–downstream asymmetry. Taken
together, these observations enable rigorous testing of models of foams [11].
Shear between two walls The quasi-static flow of a 2D foam between coaxial
cylinders, one fixed and the other rotating (cylindrical Couette flow [23]), is not always
homogeneous (cf. §4.3.2, chap. 4) since the stress is heterogeneous [13], and localized
shear bands are often observed near the moving wall. If a foam is sheared between two
straight parallel walls [61], the stress is a priori homogeneous, and the way in which
(a) (b)
(c) (d)
Fig. 2.28: Quasi-static flow of a monodisperse 2D liquid–glass foam around a cyl-

indrical obstacle [24]. (a) Photograph of the experiment (courtesy of B. Dollet). Maps
of the (b) velocity field, (c) pressure field (light grey represents high pressure, dark
grey represents low pressure), and (d) T1 density. The latter is represented by an
ellipse in which the axis shown represents the link between neighbouring bubbles be-
fore the T1 (the other axis, not shown, is for the link between bubbles that become
neighbours after the T1), and the axis length is proportional to the frequency of T1s
oriented in that direction.
Experiments 63
(a) (b)
1.0
0.4 Shear rate = 0.0014s–1
Shear rate = 0.0014s–1 Shear rate = 0.0028s–1
Shear rate = 0.0028s–1 0.5 Shear rate = 0.014s–1
0.2 Shear rate = 0.014s–1
Vbubbles\V0
Vbubbles\V0
0.0
–V0 0.0 –V0
–0.2 V0 –0.5 V0
–0.4
–1.0
–10 –5 0 5 10 –10 –5 0 5 10
y/d y/d
Fig. 2.29: Velocity profile of a monodisperse 2D foam with bubble diameter d in

simple shear between parallel walls travelling at velocity −Vo and Vo [61]. (a) Liquid–
glass experiment; (b) bubble raft. In the first case, only the bubbles near the walls
move, whereas in the second case the velocity profile is almost linear.
the foam is confined becomes significant [13]: if the foam is confined between two glass
plates or between liquid and glass, the flow is localized close to the walls; on the other
hand, the flow of a bubble raft is homogeneous, as illustrated in fig. 2.29 [61], although
structural disorder can also play a role [39], as well as the way in which the foam is
prepared [6, 12].
Migration of a large bubble Consider the flow of a foam through a Hele-Shaw
cell that consists of one large bubble surrounded by small ones. At low flow rate all
the bubbles move at the same speed but at high flow rate the large bubble migrates
more quickly than the others, without rupture of the films, through a succession of
T1s in front and behind the large bubble [10]. This migration is due to the lower wall
friction felt by the large bubble.
6 Experiments
6.1 Surface tension and surfactants
1) Pepper displaced by soap
Phenomenon demonstrated: spreading of surfactant molecules at an interface.

References in the text: §2.1.1 and §2, chap. 3.
Materials:
• a bowl
• water
(a) (b)
Fig. 2.30: Pepper displaced by soap: (a) the pepper is sprinkled uniformly on the
surface of the water; (b) a drop of washing-up liquid is placed at the centre and
pushes the pepper to the edges. Photographs courtesy of F. Elias and F. Rouyer.
• a small amount of washing-up liquid diluted in water (about 5%)

• black pepper
1. Fill the bowl with water.

2. Sprinkle the pepper evenly to form a thin layer at the surface of the water
(fig. 2.30a).
3. Carefully place a drop of the detergent solution at the centre.
The pepper is pushed to the edges of the bowl (fig. 2.30b) by the surfactants spreading
out on the surface of the water. Repeat the experiment with a drop of pure water to
emphasize the role of the surfactants.
2) My cup runneth over

Phenomenon demonstrated: reduction of surface tension by the addition of surfactant.
References in the text: §2.1.1 and §2, chap. 3.
Materials:
• a transparent, plastic cup
• water
• a small amount of washing-up liquid diluted in water (about 5%)
1. Fill the cup to the brim, until the liquid surface is higher than the top of the glass
(fig. 2.31a).
2. Carefully place a drop of detergent solution right at the centre of the surface.
The water runs over the sides of the glass. The surfactant lowers the surface tension
and changes the contact angle between the glass and the liquid. Then the interface
becomes flatter and the excess is lost (figs. 2.31b and 2.31c).
Experiments 65
(a) (b) (c)
Fig. 2.31: One more drop of liquid makes the cup overflow: (a) the interface is raised
above the rim of the cup because of a high contact angle; (b) adding a little surfactant
lowers the contact angle and causes the liquid to overflow; (c) the excess water is lost.
Photographs courtesy of F. Elias and F. Rouyer.
6.2 Creation and observation of 2D and quasi-2D foams

preparation time: 10 mins experimental time: 30 mins
Phenomena demonstrated: Structure of a 2D foam: hexagonal arrangement, Plateau’s

laws.
References in the text: §2.2, §3 and §5.
Materials:
• a transparent container with a large open surface (for example a large plastic food
container)
• water
• a straw
• a transparent flat plate, for example a flat-bottomed glass dish
• an empty transparent CD case (i.e. without the plastic CD support)
1. Dilute a small amount of detergent in the container filled almost to the top with
water. The three steps below describe three ways of forming a 2D foam (see
fig. 2.25).
2. Bubble raft: with one end below the surface, blow air through the straw. Bubbles
rise, spread out, and arrange themselves on the surface.
3. Liquid–glass: cover this foam with the pre-wetted (i.e. dipped in the detergent
solution) plate. The bubbles now appear polygonal from above.
4. Hele–Shaw (glass–glass): lower the closed CD case vertically into the soapy
water, until the hole for the CD support is underwater. Put one end of the straw
into this hole and blow air in. You should create a 2D foam inside the case, with
polygonal bubbles and a well-defined liquid fraction.
In each case, the harder you blow, the larger and more polydisperse the bubbles.
Observe the curvature of the films separating neighbouring bubbles. If you blow more
gently and smoothly, the foam will be less polydisperse, with ordered regions. It
is sometimes possible to obtain a honeycomb pattern, with perhaps a few defects
(dislocations, grain boundaries, etc.).
Comments:
1. You can also create a 2D foam by squeezing a small amount of shaving foam between
two transparent plates. In this case, the bubbles are smaller and polydisperse.
2. Once the 2D foam has been made, it is possible, with good lighting and a good
camera, to illustrate the properties presented here: Plateau’s laws (§2.2), topology
(§3.1, §3.3), coarsening and von Neumann’s law (§3.1.1, chap. 3), and 2D flow (§5.4).
6.3 Giant soap films

preparation time: 30 min to 2 hrs experimental time: 30 s
Phenomena demonstrated: liquid flow, film tension, iridescence, robustness and rup-
ture of soap films.
References in the text: §1.2 and §1.1.3, chap. 5.
Materials:
• patience
• a fairly high point of attachment (e.g. a ceiling hook) above a cleanable floor
(inside a building)
• nylon fishing line (or cotton thread)
• a plastic bottle (e.g. a 2 litre soft-drink bottle)
• a bucket
• several litres of water
• good quality washing-up liquid (see comments below)
• a weight (100–200 g)
• a paper clip
• a bradawl (or a cork-screw or large needle)
1. Pierce a small hole in the cap of the bottle with the bradawl, small enough to allow
a liquid flow rate of only a few drops per second.
2. Thread two lengths of fishing line (better is a single length folded in half) through
the hole and attach them to the paper clip (which then rests in the neck of the
bottle).
Experiments 67
(a) (b)
Reservoir of solution
(bottle)
Nylon fishing line

(frame for film)
"Pulling"
line
Weight
Bucket
Fig. 2.32: Giant soap film: (a) At the start the two lengths of fishing line are touching.
(b) When they are pulled apart they enclose a giant film.
3. Make two large holes (a few cm in diameter) at the base of the bottle, thread a
second piece of fishing line through it and hang the bottle from the ceiling, neck
downwards.
4. Attach the other ends of the two pieces of fishing line to the weight and
check that they are firmly attached (fig. 2.32a); place the bucket beneath the
apparatus.
5. Attach to each of these vertical pieces of line a horizontal “pulling” line (the other
end of which can be fixed so that it does not get tangled).
6. Pour into the bottle (through a hole in its base) a small amount of soapy water
(5% washing-up liquid).
7. When the liquid has descended to completely cover the lines, pull them apart
(fig. 2.32b).
Most films last only a few seconds or perhaps a minute, but if they rupture they can
be regenerated many times while there is still soap-solution in the bottle. The vis-
ible iridescence of the soap film indicates its micron thicknesses (see §1.1.3, chap. 5),
so a film of 10 m in height and a metre in width contains only a few centilitres
of liquid! Blow on the film, and observe how deformable it is. Objects that have
been dipped in the soap solution can be passed through the film. Moving an object
within the film allows you to observe its wake, showing the 2D flow of liquid in the
film. It is even possible to use objects with different shapes (for example an aero-
foil) in order to visualize the influence of the shape of the object on the flow of the
fluid [15].
Fig. 2.33: With a good washing-up liquid, this giant film reached 18 m in height and
showed spectacular colours. Photograph courtesy of F. Mondot [5]. See Plate 4.
Comments:
1. For more details see [5].
2. For this activity the best washing-up liquid seems to be sold by Procter and Gamble
under the label Dreft, Dawn, or Fairy. It enables the generation of films greater than
10 m in height which may last for more than ten minutes.
3. By feeding liquid into the film at several points, you can improve its longevity. In
the laboratory, some last for hours.
6.4 Kelvin cell

preparation time: 15 min to 2 hrs experimental time: infinite
Phenomenon demonstrated: 3D visualization of a Kelvin cell.

Reference in the text: shaded section on p. 50 on the Kelvin cell.
Materials:
• scissors
• glue
• photocopier (or tracing paper)
1. Copy fig. 2.34 using a photocopier or tracing paper.
2. Cut around the edge.
3. Fold.
4. Glue the tabs around the faces and stick the structure together.
Exercises 69
Fig. 2.34: Kelvin cell, to cut, fold, assemble, and stick, then to keep on your desk
(diagram taken from the Ph.D. thesis of E. Janiaud) [37].
Comments:
1. Increasing the size of fig. 2.34 makes the task easier.
2. Why stop at one? Make several cells to see how they fit together and amuse yourself
indefinitely!
7 Exercises
Solutions to the exercises are available on request through the following webpage:
http://www.oup.co.uk/academic/physics/admin/solutions.
7.1 Interfacial area of a foam
(a) Find, from the data on p. 50, the total interfacial area per unit volume of a dry
Kelvin foam, as a function of the radius RV of the bubbles. Compare this with a
packing of spherical bubbles of the same radius. We will ignore this difference in
the following, which deals only with orders of magnitude.
(b) A monolayer of SDS has a thickness of about 1.5 nm, and the surface area per
molecule at the surface is about 0.20 nm2 . Consider a litre of foam made with
a solution of SDS (see p. 19), with bubbles of 1 mm in diameter. Calculate the
volume of SDS required, if it is assumed to be pure and of the same density as in
the monolayer.
(c) If the concentration of the SDS solution, measured before foaming, is
8 × 10−3 mol.L−1 and the liquid fraction of the foam is 5%, what is the volume of
solution required to produce the litre of foam? How are the surfactants distributed
between the liquid phase of the foam and the surface?
7.2 Film tension and the Young–Laplace law

This exercise illustrates, for a cylinder, relationships (2.3) and (2.4).
Fig. 2.35: Determination of the film tension.
(a) Write down the balance of forces exerted on the shaded film element in fig. 2.35,
including surface tension, the thickness h of the film, a constant pressure p in the
liquid phase, and pressures Patm and P on either side. The film is part of a cylinder
of internal radius R and external radius R + h.
(b) Show that you recover the Young–Laplace law (eq. (2.3)) by defining the film ten-
sion as γfilm = 2γ − (p − Patm )h, which justifies the expression for the film tension
in §2.1.2. By estimating the order of magnitude of each term, show that γfilm ∼ 2γ
to recover relationship (2.4).
7.3 Plateau’s laws in 2D

This exercise illustrates Plateau’s laws, listed on p. 30. It is continued in exercise 7.5,
chap. 4.
Consider a system consisting of two parallel plates held apart by four pins posi-
tioned at the corners of a rectangle a × b. A soap film is attached to each pin, and is
surrounded by the pin, the two plates (along which it slides freely), and a central film,
as illustrated in fig. 2.36. The pressure is Patm on both sides of each film, which is
thus straight (when viewed from above).
Fix the lengths a and b and allow the system a single degree of freedom: the length d
of the central film, which can be oriented in the x or y direction.
Fig. 2.36: Plateau’s laws. The four black dots are the pins viewed from above.
Exercises 71
(a) Calculate the total edge length, as a function of d, for the two directions of the
central film.
(b) Calculate the energy of the configuration in which all four pinned films meet at the
centre (d = 0). Deduce from this that, depending on the value of a/b, the four-fold
vertex is unstable in one or both directions, and thus always unstable.
(c) From the energy, show that as a function of a/b, there are only one or two equilib-
rium positions (in the latter case, one is stable and the other metastable). Confirm
that the angles between the films are 120◦ for these equilibrium structures.
7.4 Euler’s formula
(a) A 3D bubble is analogous to a 2D foam which is wrapped onto a sphere; in this

case eq. (2.9) has χEuler = 2. By replacing the number of bubbles by the number
of faces in eq. (2.9), and using eq. (2.12), demonstrate eq. (2.13).
(b) Starting from eq. (2.9), repeat the argument presented in 2D in the shaded section
on p. 32 for a 3D foam. Then deduce the relationship between the average number
of faces per bubble and the average number of edges per face in a 3D foam.
7.5 Perimeter of a regular 2D bubble

Consider a 2D bubble which has n identical sides, that is the same length b, the same
curvature κ = 1/ρ, and which meet each other at an angle of 120◦ , as in fig. 2.37.
Calculate the area of a sector of the bubble as a function of b and n, i.e. calculate
the area of the triangle OBB joining the centre of the bubble to two adjacent vertices,
and then calculate the correction due to the curvature of the side (concave or convex)
by determining the area of the circular segment between the straight line BB and
the arc BB . Deduce from this the perimeter P and the area A of the bubble, as a
δ b
B B′
β a 60°
α
O
ρ
Fig. 2.37: A 2D bubble which has n = 5 identical sides, adapted from [27]. It is like
a regular polygon except that its sides are curved.
√
function of b and n. Tabulate the dimensionless quantity P/ A as a function of n,
and show that its numerical value depends only weakly on n.
7.6 Energy and pressure

Consider an equilibrium foam with free boundary (i.e. which doesn’t touch the walls
of a box). Suppose that we expand all the lengths by multiplying them by a certain
factor k (whilst conserving the pressure and the shape of each bubble). How do the
surface areas and volumes of each bubble change?
Since the foam is at equilibrium (i.e. it minimizes some judiciously chosen thermo-
dynamic potential) before expansion, demonstrate eq. (2.18), linking the surface energy
of the entire foam with the pressures in each of the bubbles. Deduce from it the equa-
tion of state of the foam. Show that these equations also apply to a two-dimensional
foam.
References
[1] D.A. Aboav, Metallography, 3, 383, 1970.
[2] F.J. Almgren Jr and J.E. Taylor, Sci. Am., 235, 82, 1976.
[3] T. Aste, D. Boosé, and N. Rivier, Phys. Rev. E, 53, 6181, 1996.
[4] J.E. Avron and D. Levine, Phys. Rev. Lett., 69, 208, 1992.
[5] P. Ballet and F. Graner, Eur. J. Phys., 27, 951, 2006.
[6] S. Bénito, F. Molino, C.-H. Bruneau, T. Colin and C. Gay, Eur. Phys.
J. E, 35, 51, 2012.
[7] B. Berge, A.J. Simon, and A. Libchaber, Phys. Rev. A, 41, 6893, 1990.
[8] V. Bergeron, J. Phys.: Cond. Matter, 11, R215, 1999.
[9] B.P. Binks, Curr. Opin. Colloid Interface Sci., 7, 21, 2002.
[10] I. Cantat, C. Poloni, and R. Delannay, Phys. Rev. E, 73, 011505, 2006.
[11] I. Cheddadi, P. Saramito, B. Dollet, C. Raufaste, and F. Graner, Eur.
Phys. J. E, 34, 1, 2011.
[12] I. Cheddadi, P. Saramito, and F. Graner, J. Rheol., 56, 213, 2012.
[13] I. Cheddadi, P. Saramito, C. Raufaste, P. Marmottant, and F. Graner,
Eur. Phys. J. E, 27, 123, 2008.
[14] R.J. Clancy, E. Janiaud, D. Weaire, and S. Hutzler, Eur. Phys. J. E, 21,
123, 2006.
[15] Y. Couder, J. Phys. Lett., 45, 353, 1984.
[16] S. Courty, B. Dollet, F. Elias, P. Heinig, and F. Graner, Europhys. Lett.,
64, 709, 2003. See erratum: B. Dollet, F. Elias, and F. Graner, Europhys.
Lett., 88, 69901, 2009.
[17] S.J. Cox and M.A. Fortes, Phil. Mag. Lett., 83, 281, 2003.
[18] S.J. Cox and F. Graner, Phys. Rev. E, 69, 031409, 2004.
[19] S.J. Cox, J. Non-Newton. Fluid Mech., 137, 39, 2006.
[20] S.J. Cox, D. Weaire, and M.F. Vaz, Eur. Phys. J. E, 7, 311, 2002.
[21] R.M.C. de Almeida and J.C.M. Mombach, Physica A, 236, 268, 1997.
References 73
[22] P.-G. de Gennes, F. Brochard-Wyart, and D. Quéré, Capillarity and

Wetting Phenomena: Drops, Bubbles, Pearls, Waves, Springer, New York, 2004.
[23] G. Debrégeas, H. Tabuteau, and J.-M. di Meglio, Phys. Rev. Lett., 87,
178305, 2001.
[24] B. Dollet and F. Graner, J. Fluid Mech., 585, 181, 2007.
[25] M. Durand, J. Käfer, C. Quilliet, S. Cox, S.A. Talebi, and F. Graner,
Phys. Rev. Lett., 107, 168304, 2011.
[26] J.-C. Géminard, F. Caillier, and P. Oswald, Phil. Mag. Lett., 84, 199, 2004.
[27] J.A. Glazier, Dynamics of cellular patterns, Ph.D. thesis, University of Chicago,
1989. http://biocomplexity.indiana.edu/jglazier/docs/dissertation/Glazier-
Dissertation.pdf
[28] F. Graner, Y. Jiang, E. Janiaud, and C. Flament, Phys. Rev. E, 63, 011402,
2001.
[29] F. Graner,“Two-dimensional fluid foams at equilibrium”, in Morphology
of Condensed Matter—Physics and Geometry of Spatially Complex Systems,
K. Mecke and D. Stoyan (eds.), pp. 187–214, Lecture Notes in Physics 600,
Springer, Heidelberg, 2002.
[30] W.C. Graustein, Ann. Math., 32, 149, 1931.
[31] T.C. Hales, Disc. Comp. Geom., 25, 1, 2001.
[32] S. Hilgenfeldt, A.M. Kraynik, S.A. Koehler, and H.A. Stone, Phys. Rev.
Lett., 86, 2685, 2001.
[33] S. Hilgenfeldt, A.M. Kraynik, D.A. Reinelt, and J.M. Sullivan, Europhys.
Lett., 67, 484, 2004.
[34] R. Höhler, Y. Yip Cheung Sang, E. Lorenceau, and S. Cohen-Addad,
Langmuir, 24, 418, 2008.
[35] C. Isenberg, The Science of Soap Films and Soap Bubbles, Dover, New York,
1992.
[36] I.B. Ivanov, Thin Liquid Films: Fundamentals and Applications, Surfactant
Sciences Series Vol. 29, CRC Press, Boca Raton, 1988.
[37] E. Janiaud, Élasticité, Morphologie et Drainage Magnétique dans les Mousses
Liquide, Thèse de l’Université Paris 7, 2004. http://tel.archives-ouvertes.fr/tel-
00007740.
[38] S. Jurine, S. Cox and F. Graner, Colloid Surf. A., 263, 18, 2005.
[39] G. Katgert, M.E. Möbius, and M. van Hecke, Phys. Rev. Lett., 101, 058301,
2008.
[40] N. Kern and D. Weaire, Phil. Mag., 83, 2973, 2003.
[41] A.M. Kraynik, D.A. Reinelt, and F. van Swol, Phys. Rev. E, 67, 031403,
2003.
[42] A.M. Kraynik, D.A. Reinelt, and F. van Swol, Phys. Rev. Lett., 93, 208301,
2004.
[43] M. Lösche, E. Sackmann, and H. Möhwald, Ber. Bunsen.-Ges. Phys. Chem.,
87, 848, 1983.
[44] O. Lordereau, Les mousses bidimensionnelles: de la caractérisation à la
rhéologie des matériaux hétérogènes, Ph.D. thesis, Université de Rennes, 2002.
[45] R. MacPherson and D. Srolovitz, Nature, 446, 1053, 2007.
[46] M. Mancini and C. Oguey, Eur. Phys. J. E, 22, 181, 2007.

[47] T.G. Mason, M.-D. Lacasse, G.S. Grest, D. Levine, J. Bibette, and D. A.
Weitz, Phys. Rev. E, 56, 3150, 1997.
[48] E.B. Matzke, Am. J. Bot., 33, 288, 1939.
[49] M.P. Miklius and S. Hilgenfeldt, Phys. Rev. Lett., 108, 015502, 2012.
[50] C. Moukarzel, Physica A, 199, 19, 1993.
[51] W.W. Mullins, Acta Metall., 37, 2979, 1989.
[52] J.A.F. Plateau, Statique expérimentale et théorique des liquides soumis aux
seules forces moléculaires, Gauthier-Villard, Paris, 1873.
[53] H.M. Princen and A.D. Kiss, Langmuir, 3, 36, 1987.
[54] C. Quilliet, S. Ataei Talebi, D. Rabaud, J. Käfer, S.J. Cox, and
F. Graner, Phil. Mag. Lett., 88, 651, 2008.
[55] C. Raufaste, B. Dollet, S. Cox, Y. Jiang, and F. Graner, Eur. Phys. J. E,
23, 217, 2007.
[56] N. Rivier, in Disorder and Granular Media, edited by D. Bideau and A. Hansen,
Elsevier, Amsterdam, p. 55, 1993.
[57] N. Rivier, Phil. Mag. A, 40, 859, 1979.
[58] J.-F. Sadoc and R. Mosseri, Geometrical Frustration, Cambridge University
Press, Collection Alea-Saclay, 1999, chap. 4.
[59] J.E. Taylor, Ann. Math, 103, 489, 1976.
[60] J. von Neumann, in Metal Interfaces, American Society for Metals, Cleveland,
p. 108, 1952.
[61] Y. Wang, K. Krishan, and M. Dennin, Phys. Rev. E, 73, 031401, 2006.
[62] D. Weaire and N. Rivier, Contemp. Phys., 25, 59, 1984.
[63] D. Weaire and S. Hutzler, The Physics of Foams, Clarendon Press, Oxford,
1999.
3
Birth, life, and death
1 Foam evolution
We first introduce qualitatively the different mechanisms by which a foam forms,
ages, rearranges, and then collapses, which collectively determine foam stability and
lifetime. Later on in the chapter each of these stabilizing and destabilizing mechanisms
is described in detail.
1.1 The competition between different processes

Aqueous solutions don’t all foam in the same way. When you try to produce a foam—
for example by shaking a closed flask containing liquid and gas—you may observe
different behaviour: a few short-lived bubbles, a fragile foam which rapidly disappears,
or a more stable one lasting several hours. The fact that a foam forms at all is the result
of several different mechanisms which tend either to produce and stabilize it (§1.1.1)
or to destroy it (§§1.1.2, 1.1.3 and 1.1.4). Each of these effects has its own charac-
teristic time- and length-scales and a particular dependence on the liquid fraction
and the bubble size. The competition between them determines the foam’s lifetime,
its macroscopic behaviour, and how it evolves in time, or ages. The complexity of
the phenomena that are observed comes in part from the combined effects of these
different mechanisms.
1.1.1 Stabilizing mechanisms A foam doesn’t form spontaneously. Energy is

required to disperse the gas in the liquid, but energy alone is evidently not sufficient.
The foamability or foaming capacity of a solution is a qualitative measure of its capacity
to produce a foam when it is shaken, or when bubbles are injected into it. Surfactants
have a strong influence on the foamability of a solution, as described in §2. They
adsorb at liquid interfaces and films and modify their properties. Both the surface
concentration at equilibrium and the speed with which they cover an interface are
significant. In addition to the static properties of a surfactant-laden interface, we will
see that its viscoelasticity is equally important.
1.1.2 Drainage In response to gravity the liquid flows through the foam (fig. 3.1)
and the foam is said to drain. In §4 we describe a theoretical approach to drainage that
is similar to the description of flow in porous media, in which the foam is treated as a
continuous medium with an effective permeability. Nevertheless, a foam has character-
istics which distinguish it from a porous solid: the size of the pores (the cross-sectional
76 Birth, life, and death
H(t) g
L(t)
Time
Fig. 3.1: Liquid accumulates beneath a draining foam over time, and it becomes drier
and drier (see experiment 7.2).
area of the Plateau borders) depends on the liquid fraction, and thus on the flow;
moreover, the walls of these pores (the gas/liquid interfaces) are viscoelastic (see §2.2).
The drainage rate depends on the difference in density between the gas and the
liquid. In emulsions—that is, a dispersion of one liquid (oil) in another (water), or
vice versa—drainage is less spectacular and markedly slower, because the densities are
similar, and the process is called creaming.
Note that drainage is important in the early stages of solid foam formation, when
the liquid is often still molten. It must be avoided if a homogeneous product is required.
1.1.3 Coarsening Any gas is to some extent soluble in any liquid. As a con-
sequence, the liquid films do not act as absolute barriers to the gas. As a result of
the pressure difference between bubbles, the gas may diffuse through the liquid phase
from one bubble to another. As we explain on p. 77, this results in a reduction in
the number of bubbles and an increase in their average size, and hence this process,
called coarsening, influences the life of a foam. In §3 we describe how the volume of
a bubble varies as a function of its topology and its size, and discuss the effects of
physico-chemical parameters and the liquid fraction.
The laws of equilibrium, established in §2.2.2, chap. 2, assume that the volume of
each bubble is fixed. Coarsening modifies the bubble volumes, but sufficiently slowly
that the conclusions of that chapter are correct: coarsening induces just a slow variation
in the control parameter (the volume of the bubbles) and the foam structure changes
little by little. In other words, as a foam slowly coarsens, it is (at almost every instant)
very close to mechanical equilibrium.
It is important to distinguish bubbly liquids from foams. In the first case, the
bubbles are well separated from each other; the diffusion of the gas is between the
bubbles and the solution, which acts as a reservoir of gas. The pressure of a bubble
is proportional to its curvature (eq. (2.4)). Then the pressure of a spherical bubble
is inversely proportional to its radius (eq. (2.6)), so that the smallest bubbles have a
greater overpressure relative to the liquid and they lose more gas than they receive
from it. On the other hand, the largest bubbles receive more gas from the liquid than
they lose to it, so the small bubbles shrink and the largest continue to grow. Small
Foam evolution 77
bubbles eventually disappear, and since no new bubbles are created, the total number
of bubbles decreases, and so the average size increases. This process, which shows
an increase in average bubble radius with t1/3 (both in 2D and 3D), is known as
Ostwald ripening (or Lifshitz–Slyozov–Wagner ripening), and we will consider it no
further.
In the case of dry foams, gas exchange occurs directly, from bubble to bubble.
Here too, a bubble’s pressure is proportional to its curvature (eq. (2.4)), but now the
curvature depends on the bubble shape (eqs. (2.26) and (2.27)). For a dry foam, the
average bubble radius increases with t1/2 (both in 2D and 3D). This process, called
von Neumann–Mullins coarsening, is described in detail in §3. In a dry 2D foam, the
curvature of an edge depends directly on a bubble’s number of neighbours (eqs. (2.21)–
(2.24)). This simplifies the theoretical study of coarsening: the change in area is thus
determined by the number of neighbours (and no longer by size as in Ostwald ripening).
It is generally the case that for a given liquid fraction the behaviour of the foam lies
somewhere between the two extremes.
There are other structured materials which, like foams, are organized into cellular
domains of a discrete phase dispersed in a second continuous phase. If the domains can
exchange matter (generally by diffusion), then the material coarsens in a similar fash-
ion: some domains disappear and the average size of the remaining domains increases.
This coarsening reduces the amount of energetically-costly interfaces.
This principle leads to various physical processes that drive coarsening. Foams
coarsen due to pressure-driven diffusion. Emulsions, in which the structure resembles
a foam, coarsen in the same way (it is this which causes mayonnaise to separate, for
example). A crystal consists of small monocrystalline domains, called grains, in which
the crystalline orientations are different (fig. 3.2). Their walls, or grain boundaries,
have an energy-cost proportional to their surface area. The crystals coarsen in a similar
fashion to dry foams. In fact it was the metallurgist Cyril Stanley Smith who launched
the modern study of foam physics in 1952: after one of Smith’s lectures on crystal
structure, von Neumann established the 2D growth law described in §3. Under certain
conditions the thin films of a magnetic garnet arrange themselves into domains of the
same magnetization, typically with dimensions of about ten micrometers, separated
by walls (Bloch walls) and this cellular structure coarsens. Finally, during a first-order
Fig. 3.2: The grains of a polycrystal of ice formed in the Antarctic, observed here
between crossed polarizers. Each colour corresponds to an orientation of the crystal lat-
tice. As in a dry foam, the grains coarsen and deform. Image courtesy of G. Durand [19];
the size is 5 × 2.5 cm. See Plate 5.
phase transition, for example gas/liquid or liquid/solid, the domains of two phases of
the same material coexist (fig. 2.24a), and coarsening occurs.
1.1.4 Film rupture The study of the rupture of a single film (which depends on
the dimensions of the film and the liquid fraction) and the interactions between films
(highly significant for rupture) will be presented in §5. We have already noted that, in
comparison with drainage and coarsening, it is the mechanism of foam evolution that
is least understood.
Any process which induces the rupture of a film between two bubbles (see §5.1)
tends to destroy a foam. Since the thickness of a film (<1 μm) is small compared
to its surface area (of the order of mm2 ), the film between two bubbles is a fragile
thing. It is easily ruptured, and then the two bubbles are said to coalesce (fig. 3.8).
The result of the coalescence is a reduction in the total number of bubbles (which
is the same effect as with coarsening but brought about by a completely different
mechanism).
A film breaks if the stabilizing mechanisms are absent or too weak. In the first
place there are dynamic effects, which can rupture them, in particular as films ap-
proach their equilibrium thickness (see §5). But even if equilibrium is reached and a
film is stabilized by repulsive forces between its surfaces, it is only metastable. The
spontaneous fluctuations in thickness or density close to the equilibrium state can in-
duce rupture. These fluctuations are all the more critical because the film is thin yet
of large extent. Films on the outside of the foam (as in the bath tub or the head on
a glass of beer) are even more fragile, due to the presence of external perturbations
such as grease, dust, and evaporation of liquid. Film rupture causes a foam to lose gas
and its volume decreases until it completely disappears. As we will describe in §5.3,
it is sometimes desirable to get rid of a foam, or prevent one from forming, and this
stimulated rupture is possible with antifoaming agents.
1.2 Elementary topological processes

1.2.1 Topological changes in a dry foam A topological change refers to any
process which modifies the number of faces (or neighbours) of the bubbles in a foam.
For example, as the size of the bubbles changes during coarsening, or as a foam is
made to flow by applying a strain to it, the bubbles change shape. Some Plateau
borders grow longer while others shorten and vanish. In order to continue to re-
spect Plateau’s laws (see p. 30), some topological changes must therefore take place.
All rearrangements other than coalescence or division are topologically equivalent to
a combination of two elementary processes called T1s and T2s [84]. T1s (§1.2.2)
are caused by both coarsening and flow, while T2s (§1.2.3) are due to coarsening
alone.
1.2.2 T1 topological changes When a Plateau border in a 3D foam shrinks to

the point of disappearance, more than four Plateau borders find themselves in con-
tact at a single vertex. Since this configuration is unstable, the bubbles rearrange
themselves, in a transformation called a T1, towards a new configuration of lower
Foam evolution 79
Fig. 3.3: T1 topological changes in 3D. The black lines represent the Plateau borders
(solid lines for those in the plane of the figure, and dashed lines for those behind it);
the grey ones illustrate the links between nearest neighbours before (dashed line) and
after (continuous line) the T1.
energy, resulting in a change of each bubble’s neighbours. Two possible T1 trans-

formations are illustrated in fig. 3.3: from left to right the vertical Plateau border
disappears and a new triangular face is created; three bubbles move apart and two
move closer together. From right to left the reverse transformation occurs. There is a
third type of T1, which can be viewed as being composed of two of the T1s illustrated,
in which a quadrilateral film separating two bubbles retracts not to a point but to a
Plateau border, and then a new quadrilateral face is created, perpendicular to the first
one [65]. In all cases, a T1 can be characterized by the size and directions of the links
which appear and disappear (fig. 3.3).
In 2D, a T1 affects four bubbles (fig. 3.4): two bubbles move apart and two
move closer together, a link between two bubbles disappears (they are no longer
nearest neighbours), and another is created (see exercise 7.3, chap. 2). The reverse
transformation is identical and there is therefore only a single type of T1 in 2D.
A T1 is a topological process in which the movement of six connected Plateau
borders (four films in 2D) induces neighbour-swapping. It results in a non-equilibrium
state, and is followed by a relaxation, the duration of which is controlled by dissipative
processes (see §3.2.3, chap. 4).
The whole process (a T1 and its associated relaxation) induces motion of bubbles
beyond the ones which swap neighbours. Superimposing the position of the bubbles
Fig. 3.4: T1 topological changes in 2D. The links (or first nearest neighbours) are
represented by the thick lines (dashed for the link which disappears; continuous for
the link which appears). Images courtesy of C. Raufaste and P. Marmottant [35],
reproduced by kind permission of The European Physical Journal.
Fig. 3.5: Range of a T1 in 2D. The geometry of a foam before and after the T1
are superimposed. Only a few bubbles around the “flipped” film are affected by the
transformation. The foam skeleton shown here is from fig. 3.6 (see shaded section) [26].
before and after a T1 (fig. 3.5) illustrates its range, which remains poorly understood.
The effect of liquid fraction is also unclear: in a wet foam, we can still define links
which appear and disappear, but they do not necessarily do so simultaneously, or in
any correlated fashion, so that even the definition of a T1 is ambiguous.
Spontaneous and induced T1s

We have up until now considered the T1s that occur spontaneously in a foam, as
a result of coarsening or flow. In these cases, the structure adopted by the foam
immediately before the T1 is an equilibrium structure (subject to the deformation
being slow), with one or several very short edges. Due to the deformation, or to
changes in bubble volumes, the equilibrium associated with the initial topology is
no longer stable: an edge shrinks to zero and a T1 therefore occurs.
In a similar way, it is possible to force a T1 to occur by applying a local force on
a vertex (fig. 3.6). In particular, this is possible in a 2D magnetic foam (see §6.1.4):
two vertices can be brought together by focusing the magnetic field with a metal pin
until a four-fold vertex is produced. When the pin is withdrawn, two equilibrium
states are accessible to the foam: the initial one, and a second one obtained by
changing the topology. This second state does not necessarily exist, cf. exercise 7.3,
chap. 2.
The energy barrier between two states separated by a T1 (i.e. the difference in en-
ergy between the initial structure and the one with a four-fold vertex) is sufficiently
high in this “induced” case that a T1 never occurs due to thermal fluctuations. In
effect, this barrier is of the order of γ times the surface area of a face in 3D, which
corresponds to about 10−5 J for a soap foam in which the bubbles are of the order
of centimetres. If the bubbles are tiny, of the order of microns, the barrier can be
much lower, perhaps 10−14 J. Langmuir foams have the smallest energy barrier, of
the order of pN × μm, which is 10−18 J, yet this is still orders of magnitude greater
than thermal fluctuations (kB T ≈ 4 × 10−21 J, where kB is Boltzmann’s constant).
Foam evolution 81
Fig. 3.6: Induced T1 and the reverse process induced in a 2D magnetic foam (see
§6.1.4). On the left: initial state; in the middle: state obtained after the T1, with
lower energy; on the right: return to the initial state after reversing the T1. Image
courtesy of F. Elias, C. Flament, F. Graner [26], reproduced by kind permission of
Taylor & Francis Ltd.
1.2.3 Topological changes that change the number of bubbles The second
elementary process, called a T2, is the disappearance of a bubble: in 3D this is usually a
bubble with four faces and in 2D it is a bubble with three sides (fig. 3.7, left to right).
In a foam, a T2 can only occur if the bubble volumes are changing: it is thus a process
typical of coarsening. The reverse process is nucleation (fig. 3.7, right to left), the
appearance of an additional bubble, which occurs when a foam is made by bubbling.
A third topological process is coalescence (fig. 3.8, left to right), in which a film
between two bubbles disappears. Topologically, coalescence can be formally broken
down into a T2 and T1s, although physically these processes are totally unconnected.
The reverse process is division, where one bubble splits into two (fig. 3.8, right to left).
As with nucleation, this occurs during foam formation. T2s and nucleation are
both, a priori, isotropic, while coalescence and division have a well-defined direction
defined by the links which appear and disappear [35].
Fig. 3.7: From left to right: T2 topological change in 2D. From right to left: nucleation
of a bubble in 2D. The links which disappear are represented by dotted lines. No links
appear.
Fig. 3.8: From left to right, coalescence of two bubbles in 2D. The film separating
the two bubbles ruptures. From right to left, division of one bubble into two. The
links which disappear and appear are represented by dotted and continuous lines,
respectively.
Finally, from the topological point of view, all these processes leave Euler’s formula
invariant, as expected, i.e. they do not change the sum N − Nfaces + Nedges − Nvertices
in 3D or the sum N − Nedges + Nvertices in 2D (see §3.1.1, chap. 2). In 2D, they also
leave the topological charge of the foam unchanged, creating as much new charge as
they destroy.
2 Birth of a foam
In this section, we explain why it is possible to form a stable foam. We first consider
the effect that surfactants have on the static and dynamic properties of an interface
and on a liquid film. We then discuss the relationship between the foaming capacity of
a solution and the properties of the resulting foam’s interfaces and films. While certain
correlations become clear, numerous questions remain open.
2.1 Foamability: introduction to the role of surfactants

Not all liquids foam in the same way; each possesses a unique foaming capacity and
therefore a different foamability (as defined on p. 75). When we try to make a solution
foam, two extremes are apparent. If the foamability is poor, a small number of bubbles
form and then disappear within a few seconds, like the foam under a waterfall. On the
other hand, if the foamability is very good, the foam lasts for at least a few minutes: it
appears stable, the bubbles don’t burst instantly, and all the gas used to produce the
foam is contained within it. This is the case for a solution of washing-up liquid or for
shaving foam, for example. Between these two extremes, there are many intermediate
possibilities: for example bubbles form and produce a few centimetres of strongly
heterogeneous, unstable foam on the surface of the liquid which soon disperses, as on
a fizzy drink or lager.
In order to obtain a stable foam, it is evidently necessary to add to the liquid
several drops of “soap” or washing-up liquid, which contain surfactants (see p. 18).
In this chapter we will describe in detail what it is that the presence of surfactants
alters at the scale of a liquid interface, and then at the scale of a film. Among all the
effects that are due to surfactants, we will determine those that are responsible for the
existence of the foam, including the adsorption time, elastic modulus in compression,
and the interactions between the interfaces.
2.2 Interfacial properties and foamability

2.2.1 Static surface tension Surfactants adsorb to the gas/liquid interfaces and
thus modify the surface tension γ (§2.1.1, chap. 2). It is therefore important to know
how γ depends on the surfactant concentrations in the bulk, c, and at the surface, Γ
(fig. 3.9).
By analogy with bulk systems, the problem can be tackled in terms of equations of
state, which determine the relationships between different thermodynamic quantities
at equilibrium. We first express the thermodynamic balance for the interface using the
Gibbs–Duhem equation (in which volume is replaced by area and bulk pressure by
surface pressure) to show that at constant temperature [1, 13]
Birth of a foam 83
Γ [mol/m2] Air
cs [mol/m3]
Water
c [mol/m3]
Fig. 3.9: Distribution of surfactants at the surface and in the bulk at equilibrium:
c is the surfactant concentration in the bulk far from the interface, Γ the surface
concentration, and cs the concentration just below the interface. Note the micelles,
the interfacial layer, and the free surfactants.

dγ = − Γi dμi , (3.1)
where dγ and dμi indicate small variations in γ and in μi , the chemical potential
of component i. In an ideal solution (without interactions), the chemical potential
is μ = μ0 + kB T ln c (kB is Boltzmann’s constant), and then for a single species in
solution we have
dγ = −ΓkB T d(ln c). (3.2)
This relationship (Gibbs’ equation) describes the adsorption isotherm at the surface,
and is the basis of all physical models of interfaces.
Irving Langmuir proposed a dynamical model (see eq. (3.7)) which predicts the
surface concentration Γ at equilibrium:
Γ/Γ∞ c
= (3.3)
1 − Γ/Γ∞ ca
where Γ∞ is the saturated surface concentration and ca is Szykowski’s concentration,
of order 10−1 kg.m−3 , discussed in §2.2.2. This allows us to calculate dc/dΓ and then
use Gibbs’ equation (3.2) to obtain a value for the surface pressure Πl :

Γ∞
Πl = γ0 − γ = kB T Γ∞ ln (3.4)
Γ∞ − Γ
where γ0 is the surface tension of the pure liquid. This can be rewritten in terms of
the concentration c using eq. (3.3):

c
Πl = γ0 − γ = kB T Γ∞ ln 1 + . (3.5)
ca
Eqs. (3.3–3.5) show how the surface tension (or, equivalently, the surface pressure),
bulk concentration, and surface concentration are related. Experimentally, if we know
two of these quantities (often the bulk concentration and the surface tension) it is
possible to deduce the third. Although quite simple, Langmuir’s equations of state are
accurate to within a few percent.
There are other, more complex, equations of state for an interface [41] that include
the interactions between molecules at the surface. In effect, the adsorbed molecules
on the surface can interact in numerous ways, for example electrostatically, by steric
repulsion, or by excluded volume (see exercise 8.2).
2.2.2 Dynamics of adsorption If we introduce surfactants into a pure solution

we observe a decrease in surface tension until equilibrium is reached, but this is not
instantaneous (fig. 3.10). Likewise, if we rapidly increase the size of an interface
between a gas and a surfactant solution, the tension increases, then it relaxes back
towards its initial value. The temporal evolution of the surface tension sometimes
differs greatly from one solution to another, even if the equilibrium values are close.
In particular, it varies with bulk concentration, as illustrated in fig. 3.10. Several
processes are important in the temporal evolution of γ(t): a molecule must diffuse to
the surface (traversing a region where the concentration is lower than in the bulk) and
it must then adsorb to the interface, overcoming the associated energy barrier. Either
of these processes may control the dynamics.
70
Surface tension (mN/m)
60
50
c
40
1 10 100 1,000
Time (s)
Fig. 3.10: At time t = 0 an interface is created rapidly in a cationic solution of sur-

factants. Surface tension is measured by the pendant-drop method (see §1.1, chap. 5)
as a function of time for different bulk concentrations c (from c = cmc/100 to c = cmc,
where cmc stands for critical micelle concentration, defined on p. 146). As the con-
centration is increased, as indicated by the arrow, there is a gradual shift from the
diffusion-limited to the adsorption–desorption limited case and the adsorption times
are reduced until equilibrium is reached. Reproduced from [70] by permission of The
Royal Society of Chemistry.
Birth of a foam 85
In either case we assume that the equilibrium equation of state linking Γ to γ is

known (Langmuir’s equation (3.4) or similar, depending on the system), and so it is
sufficient to determine the expression for Γ(t) to find γ(t).
Diffusion-limited case During the dynamic process of populating an interface,
there is more adsorption than desorption. In the diffusion-limited case, a region close
to the interface is depleted of surfactant molecules compared to the bulk, and the
local surfactant concentration, cs , is smaller than c; the thickness ζ of this region is
typically of the order of Γ/c. To zeroth order, disregarding cs , the typical time-scale
for adsorption can be determined quantitatively as tdiff = Γ2 /(Dv c2 ), where Dv is
the diffusion coefficient of the surfactants in the bulk liquid, which is just the time
required to diffuse a distance ζ. It therefore appears to be very sensitive to the bulk
liquid concentration. For example, for common surfactants of low molecular mass (see
§6.1.1), Γ is of the order of mg.m−2 and Dv = 10−10 m2 .s−1 . With c = 10−2 g.L−1 we
find tdiff = 20 s; with c = 1 g.L−1 we find tdiff = 0.002 s.
A quantitative description of this process is achieved by solving the diffusion equa-
tion for the bulk liquid. We assume that the process of adsorption/desorption is
very rapid compared to typical diffusion times, and that Γ is therefore in equilib-
rium with the subsurface layer. That is, the concentration below the interface, cs , is
given by the equation of state at the surface (eq. (3.3)). One thus obtains the surface
concentration Γ(t) as a function of time [81]:
t
Dv √ c (t)
Γ(t) = Γ0 + 2c t − √s dτ . (3.6)
π 0 t−τ
This equation
√ is not explicit, as cs (t) itself depends on Γ. However, it simplifies to
Γ(t) ∼ t at short times (while cs c), and at long times there is an asymptotic √
solution showing that the difference from the equilibrium value decreases as 1/ t.
Case limited by the adsorption–desorption barrier In this second case, ob-
served for example at high concentrations where diffusion is rapid, the situation is
different: the concentration in the bulk, c, is uniform, but Γ is no longer in equilibrium
with the bulk. Energy barriers to adsorption are generally caused by electrostatic or
steric forces. Conversely, leaving the interface also has an energy cost. For example,
certain molecules such as proteins modify their spatial configurations when they are
adsorbed, as a way of optimizing their amphiphilic character, making it more difficult
for them to return to the bulk.
Langmuir assumed that the adsorption sites are localized at the surface, that the
surfactants can adsorb and desorb, and that the interactions between molecules are
negligible. With these assumptions, the flux of matter towards the interface is
dΓ(t)
= ac(Γ∞ − Γ(t)) − bΓ(t). (3.7)
dt
The saturated surface concentration Γ∞ is the concentration that would be found if
there was only adsorption, without equilibrium with desorption. In eq. (3.7), the first
term on the right corresponds to adsorption and the second to desorption, with a
and b the adsorption and desorption coefficients. Their ratio ca = b/a is Szykowski’s
concentration (eq. (3.3)), which expresses the competition between adsorption and
desorption. Eq. (3.7) predicts the equilibrium value of Γ (eq. (3.3)), and also how the
system converges to it; the difference from the equilibrium value varies as exp(−t/τb ),
with τb = 1/(ac + b).
Case of complex solutions With a micellar solution (c > cmc; see p. 146), it is
necessary to consider several characteristic times: the diffusion of individual surfactant
molecules, the diffusion of micelles, the adsorption of surfactants or micelles, and
all the effects of micellization and de-micellization (which are rapid—of the order of
ms—for ionic surfactants, but rather slower—of the order of seconds—for non-ionic
ones) [12]. Likewise, for molecules more complex than common surfactants (see §6.1),
it may be necessary to take into account the conformational rearrangements at the
interface (protein unfolding, for example), or the formation of interfacial links between
molecules. However, these effects only become important over long periods of time, as
equilibrium is approached.
In addition to the variations in concentration perpendicular to the surface and the
evolution of surface tension as a function of time, a gradient of γ can also appear along
the surface. The interface reacts to these spatial gradients by creating an interfacial
viscoelasticity via the surfactants.
2.2.3 Interfacial viscoelasticity An interface has a two-dimensional viscoelasti-

city which characterizes the response of the surface to shear and extensional stresses.
Like bulk complex fluids (§2, chap. 4), these interfaces respond linearly to small
deformations, but their response to large deformations is flow-dependent.
The presence of an adsorbed layer of surfactants at the surface, in contact with
the surfactant reservoir in the bulk, gives rise to an interfacial viscoelasticity with
properties that are distinct from those of an uncovered interface. These viscoelastic
properties depend not only on the concentration and the solubility of the surfactant
in water, but also on the time-scale of the mechanical perturbation.
We introduce here the general concepts of interfacial rheology which apply to the
liquid/gas interfaces, and in particular to those interfaces covered by surfactants [21].
In §1.1, chap. 5, we describe several methods that quantify the rheology of interfaces.
We consider an element of length dl in a macroscopic interface. By transposing to
2D the 3D formalism (eq. (4.14)), we obtain an expression for the surface force which
generalizes the surface tension (fig. 2.4a):

s
dfi = σij n̂j dl, (3.8)
j
where n̂ is a unit vector in the tangent plane of the interface, normal to the element
dl. The indices i and j denote the components of vectorial and tensorial quantities.
s
The total surface stress tensor σij includes an isotropic contribution due to surface
s
tension (γ plays the opposite role to 3D pressure), to which the components of σ̄ij
are added, normalized to a force by a length, containing the viscous and possibly the
elastic terms:
Birth of a foam 87
s s
σij = γδij + σ̄ij (ε, ε̇). (3.9)
Here γ is the surface tension at equilibrium, corresponding to the bulk concentration

of surfactant far from the interface.1 The viscous and elastic terms are connected
to, respectively, the rate of deformation ε̇ and the deformation ε of the interface by
constitutive laws.
Response to an expansion or compression Variations in surface area are one

type of deformation that can be applied to a liquid interface by either expanding or
compressing it.
When a purely viscous interface of area A is expanded at a rate A−1 dA/dt it is
subject to surface stresses (with the notation given in fig. 3.11a):
s s 1 dA s
σxx = σyy = γ + ηd ; σxy = 0. (3.10)
A dt
The coefficient ηd , linking the stress to the rate of expansion, is the surface dilatational
viscosity, expressed in kg.s−1 .
In expansion and compression, in addition to a viscous response, the surfactants
also induce a surface dilatational elasticity. When the interface is stretched locally by
an amount δA, the surface concentration of surfactants Γ is modified, which changes
the local value of surface tension. This is taken into account in the dynamic term σ̄ s .
In the case where the total quantity of adsorbed surfactants is not modified (if they
are insoluble for example), the variation in concentration δΓ/Γ induced by the deform-
ation is δΓ/Γ = −δA/A. The elastic response is thus linked to the Gibbs–Marangoni
elastic modulus, EGM , defined in terms of the values at equilibrium (eq. (3.4)):

dγ
EGM = − . (3.11)
d ln Γ eq
(a) s (x + dx)
σxx (b) s (x + dx)
σxy
vy
s (y)
σyy
s (y + dy)
σyy x
y
s (x)
σxx s (x)
σxy
Fig. 3.11: Notation for stress on an interface. (a) Expansion and (b) shear of an
interface.
1 We have chosen this convention here, but the isotropic stress terms can be included in either γ
or σ̄ s without loss of generality.
It is measured in a Langmuir trough, from the isotherm of the insoluble layer in

compression. The interfacial stress in this case is
s s δA s
σxx = σyy = γ + EGM ; σxy = 0. (3.12)
A
On the other hand, with soluble surfactants the response to dilation becomes com-
plicated and depends on the mechanical excitation frequency. For an oscillation of the
surface of the sample at frequency ω, the complex surface dilatational elastic modulus,
Es∗ , can be split into two parts:
Es∗ (ω) = Es (ω) + iEs (ω). (3.13)
The real part (elastic modulus) describes the part of the stress that is in-phase with
the strain. The imaginary part (the loss modulus) describes the stress that is out-of-
phase by π/2 with respect to the strain; the latter modulus is linked to the surface
viscosity by Es = ωηd .
Note that calling this modulus elastic recognizes that this process induces an in-
phase response. Nevertheless, it is not associated with a solid elastic property of the
monolayer. The exact parallel in 3D is the elastic compression or expansion of a fluid,
as in sound propagation for example.
Response to shear We now consider the shear response, in which the shape of
an interface changes but its surface area remains constant. For a simple shear flow
imposed on a flat interface (fig. 3.11b), the surface shear rate is
∂vy
ε̇sxy = , (3.14)
∂x
where vy is the speed of the interface in the y direction, which varies with x. The
surface shear stress is thus linked to the strain rate ε̇sxy by
s
σxy = ηs ε̇sxy ; s
σxx s
= σyy = γ. (3.15)
ηs is the surface shear viscosity and depends strongly on the surfactants (their concen-
tration and molecular type, as discussed in §6.1). In the Newtonian case, this viscosity
is independent of the shear rate, which is usually the case for surfactants of low molecu-
lar mass. Nevertheless, more complex behaviour is found when interfaces are covered
by proteins or polymers: then viscosity can depend on shear rate, and this may be
associated with the appearance of plasticity (we shall return to these ideas in §2.1,
chap. 4), and even fracture.
Surface elasticity due to shear may also be observed in a surfactant monolayer.
Use of the term “elastic” is perfectly justified here, as it concerns behaviour specific
to solids, the in-phase response to shear. In fact, like a 3D phase, a 2D phase can be
solid, liquid, or gas (see §5, chap. 2), and a 2D solid resists shear.
As in the case of expansion and compression (eq. (3.13)), we introduce a 2D complex
modulus, G∗s , for the response to shear, defined as for a 3D fluid (eq. (4.9)):
G∗s (ω) = Gs (ω) + iGs (ω), (3.16)
where the in-phase response to the deformation is given by Gs , and the out-of-phase
stress by Gs = ωηs .
Birth of a foam 89
Values of the viscoelastic moduli The existence and frequency-dependence of

elastic and viscous moduli result from two contributions at the surfactant scale. One,
which is intrinsic to the interface, is due to hydrodynamic interactions and to fric-
tion between the adsorbed molecules, while the other comes from the exchange of
surfactants between the interface and the underlying liquid.
In fact, the response to shear (in which the shape changes but the surface area
remains constant) is dominated by intermolecular interactions at the interface. On
the other hand, the response to expansion and compression is strongly coupled to the
degree of surfactant exchange between the surface and the liquid (fig. 3.12), which is
controlled by the diffusion of surfactants in the bulk liquid and adsorption–desorption
processes at the interface. Nevertheless, this response also contains an intrinsic contri-
bution linked to mechanisms occurring at the interface and so during expansion and
compression it is difficult to distinguish the different contributions.
However, in the response to expansion and compression two limiting cases are easily
established. If the oscillation frequency ω is very low, or the surfactants very soluble,
there is sufficient time for surfactant transfer to occur and for the surface concentration
to adjust. So there is no variation in surface tension and no stress response to a
compression, hence Es = Es = 0. In the limit of high ω, or very soluble surfactants,
the layer no longer has time to react because surface and bulk are unable to exchange
surfactants sufficiently rapidly, hence Es = EGM and Es vanishes.
Between these limits the simplest model, developed by Lucassen and Van der
Tempel [50], assumes that surfactant exchange between the surface and the bulk is
controlled by diffusion. Then
1+Ω
Es = EGM (3.17)
1 + 2Ω + 2Ω2
and
Ω
Es = EGM , (3.18)
1 + 2Ω + 2Ω2
where

Dv dc
Ω= (3.19)
2ω dΓ
combines the diffusion coefficient, Dv , of surfactants in the bulk, with the surface and
bulk concentrations. For common surfactants this model is adequate, implying that
Fig. 3.12: Sketch of an interface subjected to an expansion/compression oscillation.

The black arrows show how the surfactants can move between the interface and the
bulk. Their response to the oscillation determines the viscoelasticity of the interface.
intrinsic effects play only a minor part in expansion and compression. In practice,
there is a maximum in elasticity as the concentration is varied, which often occurs just
below the cmc (see p. 146). This indicates that the presence of micelles complicates
the exchange between surface and bulk. Refinements to this model take into account
surface diffusion and energy barriers to adsorption or desorption.
Quantitatively, shear moduli are always much lower than compression moduli.
Moreover, the shear elastic modulus of soluble surfactants is usually zero (it is only
non-zero if there is a solid layer at the surface, for example a rigid network or a complex
interconnected structure). The dilatational elastic modulus typically varies between a
fraction of a mN.m−1 to hundreds of mN.m−1 .
The shear viscosity, ηs , has a typical value of 10−8 to 10−7 kg.s−1 for surfaces
covered with surfactants of low molar mass, and can reach 10−5 kg.s−1 for surfaces
covered in proteins, for example. In the former case, such interfaces are referred to
as mobile, while in the latter case they are immobile. This distinction is made quant-
itative later, in terms of the Boussinesq number (eq. (3.62)), which recognizes the
contribution of bubble size too. On the other hand, ηd is typically greater than ηs by
one or two orders of magnitude. Different models for the surface dilatational viscosity
of a film in steady extension are given in §6.2.
2.2.4 Interfacial phenomena and foamability As far as surface tension is con-

cerned, a simple rule of thumb is that the lower the value of γ, the less costly in energy
is the creation of an interface. Producing a foam requires a significant increase in the
area, Sint , of gas/liquid interface. That the interfacial energy E is proportional to γ
(eq. (2.15)) suggests that a low surface tension should enhance foamability.
However, it is the adsorption dynamics that matter, rather than the value at equi-
librium. Even when the final value of γ is low, if this is obtained after a long and slow
adsorption, the solution will foam poorly.
To create a stable foam, newly formed bubbles need to be covered with surfactant
as quickly as possible. A high surface concentration is required as soon as the bubbles
touch one another and create films. It is necessary to compare a typical adsorption time
tad (i.e. the time-scale over which the surface tension decreases) with a characteristic
bubble formation time texp , which depends on the method used to produce the foam.
If texp is long it is not necessary for the surfactants to be adsorbed rapidly, and it is
possible to produce a foam even with a low surfactant concentration. On the other
hand, if texp is short, only systems that are able to lower the surface tension very
rapidly will foam.
In order to better understand this correlation between adsorption dynamics and
the way in which the foam is made, it is possible to measure the surface tension of an
interface with continuously increasing area. By repeating the experiment at increasing
rates of expansion, we can determine the rate at which the surfactants become less
efficient at decreasing the surface tension [51].
So it is necessary for the surfactants to be able to adsorb rapidly at the surface to
form a dense layer there. Yet this does not explain how and why the presence of this
layer is indispensable to foam generation.
Birth of a foam 91
We saw in §2.2.3 that the adsorption of surfactants to an interface has a major

effect on its viscoelastic properties (both in extension and under shear), and so sur-
face rheology plays a role in foam production. The viscoelasticity of the interface in
expansion and compression, characterized by Es and Es , is widely recognized as an
important interfacial parameter for foamability and foam stability.
Yet there are no quantitative criteria that predict whether or not a foam will
form. Several effects contribute to the link between interfacial viscoelasticity and
foamability:
• For a given applied stress, a larger interfacial dilatational elastic modulus Es
results in a smaller strain. That is, the film stretches less and is therefore less
likely to rupture.
• Elasticity also acts to restore the interface. If an expansion is not uniform, or if
there is a fluctuation in surfactant density, a depleted region is created (fig. 3.13);
a surface tension gradient appears, which is greater when Es is larger. This induces
an elastic restoring force in the plane of the interface which serves to counteract
the fluctuation by bringing back surfactants and by limiting the stretch of the
interface. The modulus Es relates the size of the fluctuation or local stretch of
the interface to the surface tension gradient, and consequently to the stretching
or restoring force.
• In addition to transport at the interface, this mechanism of minimizing the inter-
facial gradients also induces a coupling between surface tension gradients and the
flow of liquid below the interface, known as the Marangoni effect. This flow in the
subsurface layer enables surfactant to be transported more rapidly to “oppose”
interfacial fluctuations.
• Even more significantly, at the scale of a foam film (fig. 3.13), the Marangoni
effect not only draws surfactants to the surface but also draws liquid into the film.
As a surface tension gradient is often associated with a fluctuation in thickness
(as in fig. 3.13), there is then a secondary benefit: if the interfaces are elastic, the
restoring force will simultaneously correct both the fluctuations in concentration
and thickness via the Marangoni effect, thereby reducing the risk of film rupture.
σ2 σ2
σ1
JW JW
Fig. 3.13: Sketch of a film with elastic interfaces, where a local stretch causes a
gradient in the surface tension coupled with a reduction in thickness. The combination
of flows induced at the surface and in the bulk enable differences in thickness and
concentration to be restored. In particular, the flux of liquid JW in the film brings
surfactant back to the interface and re-inflates the film.
• A low value of the modulus |Es∗ | facilitates the stretching of a surface; however, if
it stretches too quickly the adsorption dynamics of the surfactants may become
significant, as discussed previously. At the other extreme, if Es is high, then the
response of the interface could be solid-like, leading to the possibility of fracture.
• The elasticity of the interfaces during expansion and compression, as for surface
viscosity during shearing, modifies the process of drainage in the films and Plateau
borders. In fact, flows at the surface, governed by interfacial parameters, are
coupled to bulk flows (see §4).
• Interfacial elasticity also appears to control the appearance of bell-shaped liquid
drops (see §5) in films which form rapidly. These dimples have a destabilizing
effect on a film.
Given all these competing effects, how can elasticity be tuned to optimize foam-
ability? It is often found that an intermediate value of the dilatational elasticity—of
the order of tens of mN.m−1 —is optimal, offering a balance between the elasticity-
controlled effects described above [46]. For common surfactants, this often corresponds
to a concentration close to the cmc.
To summarize, it is clear that foamability is enhanced by a viscoelastic interface
which is rapidly covered by surfactants. However, it is not always possible to explain
the existence of a foam by these two criteria alone. As we will now describe, the
interactions between the interfaces at the scale of a film play a crucial role.
2.3 Properties of liquid films and foamability

2.3.1 Interactions between two interfaces During foaming, surfactants adsorb
to the surface of each bubble. When two bubbles meet, what forces are acting between
the two interfaces? This question proves crucial to our understanding of the stability
of the thin liquid film which separates the bubbles of a foam.
We consider first a liquid film consisting of two infinite interfaces separated by a
thickness h of liquid. We use the disjoining pressure Πd (§4.1.3, chap. 2) to characterize
the forces per unit area between these two interfaces, which result from the following
attractive and repulsive contributions [4, 58, 74].
First, the London–van der Waals forces, also known as dispersion forces, arise
because, on a microscopic level, two molecules which have electrostatic dipoles, whether
permanent or oscillating, interact: the correlation between their dipoles induces an
interaction potential V (r) which varies as 1/r6 , where r is the distance between them.
Macroscopically, after integration over all molecules of a thin film, this always results
in an attraction between both interfaces:
Ah
ΠvdW = − , (3.20)
6πh3
where Ah is the Hamaker constant, which is positive for a thin film in air. Since
interfaces tend to get closer, the film tends to thin: van der Waals forces always make a
film unstable. This fundamental fragility must then be balanced by other contributions,
hence the need for surfactants.
Birth of a foam 93
A second contribution is the electrostatic interaction, which is often significant

because the interfaces are generally electrically charged, due to the ionic character of
many surfactants (see §6.1). These charged surfaces repel each other, but this repulsion
is screened by the presence between them of an ionic solution, containing co-ions with
the same charge as the surface and counter-ions with the opposite charge. The effective
range of the electrostatic potential ψ of a charged interface (see fig. 3.14), taking into
account the role of the co-ions, is given by the Debye length,

εkB T
λD = . (3.21)
8πne2
Here ε is the relative permittivity of the fluid, n is the charge density, and e the
elementary charge. λD is also the typical size of the ionic cloud induced by a charged
surface. Therefore, when two charged surfaces approach each other the interaction
becomes significant when the separation h falls below 2λD .
(a)
λD
Liquid
Ψ(x)
Air Air
h > 2λD
Interface
(b)
λD
Ψ(x)
Air
h < 2λD
Fig. 3.14: Electrostatic potential ψ(x) of two charged thin film interfaces a distance h
apart. λD is the effective range of the electrostatic potential ψ. (a) There is no repuls-
ive interaction as long as h > 2λD . At the centre of the film, we find the electrostatic
potential in the bulk (dashed line). (b) On the other hand, for h < 2λD , the electro-
static potential in the film is nowhere equal to that in the bulk. The concentration of
counter-ions is greater than in the bulk, and this excess of counter-ions is responsible
for the repulsive force between the thin film interfaces.
Third, at small separations, the ionic clouds overlap to generate a repulsion. This
interaction is therefore entropic in origin, as suggested by the presence of kB T . With
the usual assumption of a small overlap and a constant potential, the electrostatic
disjoining pressure is approximately
Πel ∼ exp(−h/λD ). (3.22)
Adding ΠvdW and Πel (eqs. (3.20) and (3.22)) results in the DLVO model, named
for Derjaguin, Landau, Vervey, and Overbeck, which is sufficient to describe the
stability of many colloidal systems [28]. However, in foams there are often further
contributions.
At very small separation the surfactant layers adsorbed on the interfaces interact
via steric repulsion, since the surfactant layers cannot inter-penetrate. For films with
surfactants of low molecular mass this is associated with steric hindrance between
water molecules and the hydrophilic head of the surfactants. Fig. 3.15 illustrates the
van der Waals, electrostatic, and steric contributions to the total disjoining pressure
in a film:
Πd = ΠvdW + Πel + Πste . (3.23)
These interactions between bubbles only act over short distances, generally less than
100 nm. In fig. 3.15, we have represented the electrostatic interaction as suffi-
ciently strong compared to dipolar attractions that a maximum appears in the curve.
Similarly, at short distances steric interactions prevail over dipolar interactions and
Πd becomes positive again. This type of non-monotonic curve is usual for common
surfactants and yields one or two possible stable thicknesses for a given pressure
(fig. 3.20) [4].
Fourth, the supramolecular interaction is also significant for surfactant solutions
and foams. In foaming liquids, there are often organized structures in the bulk, on a
much larger scale than that of the molecules, for example micelles or bilayers formed
ste el
vdW
Fig. 3.15: Disjoining pressure between two interfaces as a function of interface sep-
aration h. The different contributions to the disjoining pressure—van der Waals,
electrostatic, steric—are represented by the dashed lines, and the continuous line
represents their sum.
Birth of a foam 95
(b)
(a)
Πd
Fig. 3.16: Stratification. (a) The disjoining pressure Πd oscillates as a function of the
thickness of the film h. (b) These oscillations result from the confinement and self-
organization of supramolecular structures in the film, for example packings of bilayers
or micelles.
by surfactants at high concentration. The confinement of these structures in the films

induces an additional disjoining pressure, which is long-range and depends on distance
(fig. 3.16). Due to the confinement, we may observe stratification in a film, i.e. organ-
ization into parallel layers. The concentration profile therefore oscillates in a direction
perpendicular to the film.
Steric effects due to confinement and non-inter-penetration can appear at a thick-
ness greater than 100 nm. For proteins, adsorbed polymers, and solid particles
adsorbed at an interface, steric repulsion is sometimes observed at thicknesses of sev-
eral hundreds of nanometres. Then the films are not uniform in thickness; instead they
appear bumpy with a congealed gel between the surfaces (fig. 3.17).
The fact that the disjoining pressure is non-monotonic has an important con-
sequence: thermodynamically, the parts of the curve where dΠd /dh is positive are
unstable and corresponds to inaccessible film thicknesses.
Fig. 3.17: Inhomogeneous films stabilized by milk protein (casein, see p. 147), ob-
tained using the film balance technique (§1.1.4, chap. 5). The bulk concentration of
the protein, and the stability of the film, increases from left to right. Interference im-
ages (§1.1.3, chap. 5) show that the thickness is of the order of several hundreds of
nanometres and that it is non-uniform. The surface area of the film in the image plane
is 3 mm2 . Photographs courtesy of A. Saint-Jalmes. See Plate 6.
Fig. 3.18: Thickness discontinuity. View of the top of a single film showing a discrete
jump in thickness from a thicker film (appearing clear) to a thinner film (appearing
darker). The images correspond to different times, and the pressure on the film is
constant. The dimensions and techniques are the same as in fig. 3.17. Photographs
courtesy of A. Saint-Jalmes. See also Plate 7.
For the usual case described in fig. 3.15, if, by departing from Πd = 0, we increase
the disjoining pressure (or if we push the two interfaces together), h first decreases
continuously to a critical value h (at the local maximum), then jumps directly—at
constant pressure—to a smaller value on the second stable branch at h .
This discontinuity in film thickness can be seen using interference (fig. 3.18; see
§1.1.3, chap. 5): if a film with uniform initial thickness greater than h is subjected
to a sufficiently large pressure, then over time some darker domains appear, corres-
ponding to reduced thicknesses. These progressively join together and, at equilibrium,
the film ends up with a lower uniform thickness. Between these two values, a whole
range is inaccessible. It is the same for stratified films: for a curve of oscillating dis-
joining pressure like that in fig. 3.16, film thinning occurs via several discrete jumps
in thickness.
Fifth, a flow of liquid in a film can also induce a dynamic pressure, which stabilizes
the thickest films. For this contribution to play a role the flows must be fast, both
along the films and along the adjacent Plateau borders. This requires large radii of
curvature (see §4.4.3), which is the case for centimetre-sized bubbles or for high liquid
fractions.
2.3.2 Conditions for local stability and foamability Returning to a foam

as a whole, we ask how the film thickness is determined by the disjoining pressure,
and thus by which of the above contributions. In order to find out, we need to again
consider the equilibrium of pressures between a flat film and the adjacent Plateau
border (fig. 3.19), introduced in §4.1.3, chap. 2.
The Plateau border curvature is such that the liquid in a film is sucked into the
adjacent Plateau borders. This is known as capillary suction and is always present
since Plateau borders surround every film.
In order for a flat film to exist at equilibrium, it is necessary that, in addition to
the liquid pressure p, there is a positive disjoining pressure in the film, which balances
the gas pressures and resists the suction from the Plateau borders. At equilibrium, the
Birth of a foam 97
P Bubble 1
Plateau
Film Πd + p p border
Pc Bubble 2
Fig. 3.19: Local equilibrium of pressures: in order to balance the different gas pres-
sures and to oppose capillary suction into the Plateau borders (and so enable a stable
film to exist), repulsive forces must act between the interfaces, inducing a disjoining
pressure Πd .
disjoining pressure is equal to the capillary pressure Pc (eq. (2.30)), Πd = P − p = Pc .

This relationship is fundamental to film stability (and to a foam’s existence!).
The capillary pressure Pc depends on the properties of a foam and is determined
by experimental conditions. It is inversely proportional to the radius of curvature of
the Plateau borders r (eq. (2.30)), which itself depends both on bubble diameter d
and the liquid fraction φl . Very different ranges of capillary pressure may therefore
be found: either low (a few Pa) for wet foams and/or large bubbles, or high (of the
order of several hundreds, even thousands, of Pa) for small bubbles and/or dry foam,
corresponding to small radii of curvature of the Plateau border, r < 0.1 mm.
Consider a foam with uniform and constant bubble size and liquid fraction, which
fixes the capillary pressure and the capillary suction. The equilibrium of pressures
implies that this foam can only exist if there are repulsive forces in the films which give
rise to a disjoining pressure Πd with Πd = Pc , allowing an equilibrium film thickness
h to be found. To evaluate the limit of stability of a foam it is therefore necessary
to compare the possible capillary pressures in the foam with the possible disjoining
pressures in a film. Foamability is enhanced when repulsive interactions are strong and
long-range.
Returning to fig. 3.15, which corresponds to common surfactants of low molecular
mass, there are ranges of thickness where the electrostatic and steric forces are greater
than the dipolar forces. At low capillary pressures (Pc = Pc1 , of the order of several
hundreds of pascals), a film thickness of several tens of nanometres can be stable
(fig. 3.20) [4]. Such a film, stabilized by the equilibrium between electrostatic and
dipolar forces, is called a common black film (CBF), with a typical thickness in the
range 10 to 80 nm.
For much greater capillary pressures (Pc = Pc2 ), there is a single equilibrium state
fixed by the equilibrium between steric and dipolar forces; it corresponds to a nano-
metric liquid film. Although extremely fragile, such a film, known as a Newton black
film (NBF), is nevertheless observable in surfactant-stabilized foams, with a thickness
in the range 2 to 4 nm depending on the surfactant chain lengths. In principle, by
carefully decreasing the pressure from Pc = Pc2 to Pc1 , it is possible that a NBF could
still be observed, at thickness heq1 (fig. 3.20). These films are called black (see §1.1.3,
chap. 5) because at these thicknesses they no longer reflect light (only destructive
interference occurs).
Fig. 3.20: Equilibrium thicknesses heq1 , heq1 , and heq2 for two values of capillary
pressure Pc1 and Pc2 . For large thickness, a common black film is stable while at the
nanometre scale a Newton black film is found.
2.4 Summary of the microscopic origins of foamability

We have seen that macroscopic foamability is associated with certain microscopic
properties peculiar to each interface, as well as to the forces which appear between
two interfaces. Can we now improve and predict the foamability of a solution?
2.4.1 Improving foamability There is a correlation between isolated film sta-

bility and foam stability. A good foaming solution must be able to rapidly adsorb
large quantities of surfactants at the surface, and the surfactants must produce a
positive and significant disjoining pressure, that is, strong and long-range repulsion,
for instance with electrostatic charge. There is also an optimal value of the surface
dilatational elasticity (neither too large, nor too small) to restore the surface after
fluctuations—notably via the Marangoni effect—and limit the dynamic instabilities of
the film interfaces.
We have seen that by increasing the surfactant concentration c, the adsorption
kinetics increase, the equilibrium value of γ decreases, and the viscoelastic moduli and
the range of the repulsive forces both increase. Consequently, all the effects which tend
to stabilize the foam are increased: by increasing c we obtain a better foaming solution
and a more stable foam. Nevertheless, foamability cannot be improved indefinitely by
varying only the concentration: when the concentration reaches the cmc, the surface is
saturated and the surface concentration no longer depends on the bulk concentration;
thus it is pointless to go much beyond the cmc (unless it is to increase stratification
effects in the films). In addition, for a weakly surface-active molecule (which doesn’t
adsorb much at the interface), increasing the concentration does not populate the
interfaces, but instead produces structures in the bulk.
Coarsening 99
2.4.2 Predicting foamability Are measurements at the liquid interfaces (for ad-
sorption dynamics) and on single films (to test their stability) sufficient to understand
and predict the foamability of a solution? The answer is usually yes, at least for simple
surfactant systems of low molecular mass, but there are cases where measurements on
the substructures don’t reflect the actual properties of the 3D foam, in particular
for complex chemical formulations: surfactant mixtures, surfactant–polymer mixtures,
proteins, etc. In certain cases, measurements of the adsorption dynamics, γ(t), on a
free interface are poorly correlated with foamability. Thus certain systems foam at
concentrations for which the γ(t) measurements at the scale of a single gas/liquid
interface suggest that only very little material adsorbs, and that it does so slowly
(suggesting a poor foamability). In this case, a measurement on a single interface does
not reflect the properties of the foam: it doesn’t take into account, for example, rapid,
forced adsorption phenomena, which occur when bubbles come into contact during the
process of foam generation.
Herein lies the main problem: due to experimental limitations, we can only study
certain foam substructures at time-scales, frequencies, amplitudes, or length-scales
different from those that occur during foaming and which are therefore not directly
relevant. Likewise, we can often only measure properties in the linear regime, which
is not always representative of foam formation. Finally, in the majority of measuring
techniques on isolated substructures, we often suppress collective effects, associated
with confinement or with dynamics for example. Thus, the information obtained on
the intermediate scales, although always useful, often remains incomplete and the
conclusions that one draws are often inadequate.
Consequently, there still remains much to learn in order to be able to respond with
certainty in every case to the question “why does that foam?”.
3 Coarsening
A bubble with few neighbours has a higher pressure than its neighbours. If the gas that
it contains can diffuse through its films, the bubble shrinks and then loses neighbours,
and consequently empties more and more quickly (§3.1). In this way, bubbles disappear,
one after another, and so the average size of those which remain increases. Finally
there is only one large bubble remaining, which contains all the gas. This coarsening
mechanism and the resulting foam structure are generic (§3.2). However, the rate at
which coarsening occurs depends on the particular characteristics of the system being
considered; for a foam, it depends on the liquid fraction, the average bubble size, and
the physical chemistry of the gas and the liquid (§3.3). We begin with 2D (§3.1.1),
which is much simpler than 3D (§3.1.2).
3.1 Growth rate of a bubble in a dry foam

The following is specific to the growth of a bubble in a dry foam. A large part of the
literature on this subject in fact boils down to the study of film curvature, described
in §3.3.2, chap. 2.
3.1.1 Dry 2D foam: von Neumann coarsening We consider an ideal 2D foam

(see p. 26), that is, a dry foam with thin films. If the gas diffuses from a bubble i
to its neighbour j solely through the film ij, the flux from one bubble to the other
is proportional to ij (Pi − Pj ): it is the product of the driving force, here Pi − Pj ,
and the size of the region of gas exchange, here the length ij of the film. If the flux
is positive, the amount of gas contained in bubble i decreases, as does the bubble
area Ai (the gas density remains roughly constant). More precisely, the change in area
of bubble i, dAi /dt, results from the sum of the fluxes to each neighbour j (fig. 3.21):
dAi
n
= −a1 (Pi − Pj ) ij , (3.24)
dt j=1
where the coefficient a1 , which is the same for all films in a given foam, will be
interpreted in §3.3.
Surprisingly, the right-hand side of eq. (3.24) appears in eq. (2.24), relating pressure
to topology and curvature. By combining them we obtain:
2a1 λ
i n
dAi 2a1 λπ
=− κij ij = − (6 − ni ). (3.25)
dt e j=1 3e
This remarkable relationship between topology, geometry, forces, and the evolution
of bubble i, which is at least correct within the limit of the ideal foam model (see
§2.2.1, chap. 2), is readily confirmed experimentally (fig. 3.22a). It can be written in
the following abbreviated form, due to von Neumann [79], and even in terms of the
geometric charge defined in eq. (2.23):
dAi π
= −Deff qi = Deff (ni − 6). (3.26)
dt 3
The effective diffusion constant Deff is positive, and is expressed in m2 .s−1 . It is the
same as in 3D and is proportional to the rate of gas diffusion Df in the foaming liquid,
but also includes several other factors (see §3.3).
n<6 n=6 n>6

Convex Straight Concave
on average on average on average
P P P
P > P0 P = P0 P < P0
Fig. 3.21: A bubble with 5 sides or less has convex sides, so its pressure is greater
than the average pressure of its neighbours, P0 , and it shrinks. The curvature of the
sides of a bubble with 6 neighbours is zero on average, and it is therefore at the same
pressure as the average of its neighbours. A bubble with 7 sides or more has concave
sides, its pressure is lower than its neighbours, and it grows.
Coarsening 101
(a) (b)
10 1.0
8 48 min; 20 %
0.8 375 min; 20 %
48 min; 14 %
6
Cr(×102 μm2.s–1)
0.6 48 min; 17 %
dA . *
A
4 0.4
dt
2 0.2
0 0
–2 –0.2
4 6 8 10 12 14 16 0 5 10 15 20 25 30
n Number of faces
Fig. 3.22: Growth rate of a bubble as a function of its number of neighbours.

(a) Bubble area variation vs. number of sides in a dry 2D foam (from [18]). The
individual bubble growth rates are averaged over the duration of the experiment, and
fitted by a straight line that passes through 0 at n = 6, as predicted by eq. (3.26)). The
slope is Ȧ∗ = 140 mm2 .h−1 , giving Deff = 133 mm2 .h−1 . (b) Relative bubble volume
variation vs. number of faces in three wet 3D foams, with liquid fractions between 14
and 20% [44]. The graph is directly comparable with fig. 2.17, with the opposite sign,
in agreement with eq. (3.30). The anomalous negative slope at low f is probably linked
to the liquid fraction (see §3.3.3). Unlike in fig. 2.17, the data come from dynamic
experiments (X-ray tomography, see §1.2.4, chap. 5) which enables the prefactor Deff
to be determined.
Consequently, bubbles with positive geometric charge (3, 4, or 5 sides, see §3.3.2,
chap. 2), which have a higher pressure than the average of their neighbours (balanced
by the length of the shared edges), lose their gas to bubbles which have more than
6 sides (fig. 3.21). The geometric details (size and shape of the bubbles) do not
play a role, only the topology is important. This result, which had been predicted
theoretically before being confirmed experimentally [72], has simplified considerably
the study of coarsening in 2D.
If the foam has a free surface through which it can exchange gas with the envir-
onment, the peripheral bubbles are rather bulbous (fig. 2.12), and, since the foam
pressure is greater than ambient pressure, overall the foam gradually
N loses gas.
On the other hand, for a foam of fixed total area Afoam = 1=1 Ai (a foam in a
box, or even a numerical simulation of a foam using periodic boundary conditions),
any gas lost from one bubble is redistributed among it neighbours, and the average of
dAi /dt over the foam is zero:

1
dAi
N
dAi
= = 0. (3.27)
N i=1 dt dt
Eqs. (3.26) and (3.27) are compatible with the fact that the average value of n in the
foam is 6 (see p. 32). However, d Ai /dt is not zero: the average size increases, as
discussed on p. 105.
3.1.2 Dry 3D foam: Mullins coarsening The von Neumann relationship in

2D, relating the curvature of a bubble’s sides to its number of neighbours (eq. (2.24)),
cannot be generalized to 3D in an exact way (cf. eq. (2.27)). The same is therefore
true for the growth rate: in 2D it is an exact function of the number of neighbours
(eq. (3.25)), but this is not the case in 3D.
In 3D, the flux of gas across a film is proportional to the pressure difference across
it, this time multiplied by its surface area S. More precisely, dVi /dt is the sum of the
fluxes to each neighbour j. With a1 the same constant as in eq. (3.24), we obtain
dVi
n
n
= −a1 (Pi − Pj ) Sij = −2γa1 Hij Sij , (3.28)
dt j=1 j=1
where we have introduced the mean curvature Hij of the film ij using the Young–
Laplace law (eq. (2.4)). The term Hij Sij consists of a curvature (m−1 ) multiplied by
a surface area (m2 ), so it is a length. In order to compare experiments on different
1/3
foams it is natural to divide eq. (3.28) by a length, usually Vi . We consequently
r
define the relative growth rate C of bubble i as
1 dVi 3 d 2/3
Cir = 1/3 = V . (3.29)
Vi dt 2 dt i
As in 2D (§3.1.1), the proportionality constant of eq. (3.28) is an effective diffusion

coefficient, denoted Deff , expressed in m2 .s−1 . We thus write

n
Hij Sij
Cir = −Deff 1/3
= −Deff qi , (3.30)
j=1 Vi
in terms of the geometric charge qi (eq. (2.26)).

Bubble growth and number of faces As in 2D (eq. (3.26)), the change in volume
of a bubble depends only on its shape, and in determining the growth rate we need
not concern ourselves with its neighbours. Nevertheless the geometric characteristics
which influence the growth rate are not as simple as in 2D.
We note first that Cir (eq. (3.30)) is only the relative growth rate; the actual growth
1/3
rate, dVi /dt = Cir Vi , depends explicitly on the bubble’s volume Vi . Comparing
bubbles of the same volume, fig. 3.22 and eq. (3.30) together show that almost
all bubbles with a small number of faces, typically f ≤ 11, shrink over time, whereas
those bubbles with a large number of faces, for example f ≥ 16, get larger more quickly
as f increases. Bubbles which have between 12 and 15 faces are almost stationary,
neither losing nor gaining much gas. The same result can be seen in the (theoretical)
curve of geometric charge (fig. 2.17) and in numerical simulations [75]. However, in
contrast to the 2D case (eq. (3.26)), the equation linking the growth rate of a bubble
Coarsening 103
with its number of faces is not exact. If a bubble has faces which all have roughly
the same number of edges and the same shape, we observe that it grows faster (or
shrinks more slowly) than a more asymmetric bubble with the same number of faces
(fig. 2.17).
In particular, the sign of the growth rate of a bubble with f ≈ 13 depends on its
exact shape (fig. 3.22). It is even possible that it might show no change in volume
(C r = 0), as for the idealized bubble without curvature (with 13.4 faces, see p. 44).
Recall that, in a dry foam, the growth rate of a bubble satisfies
dVi 1/3
= −Deff Vi qi . (3.31)
dt
1/3
It is thus linked to the bubble diameter via Vi , to the different physical paramet-
ers at the local scale via Deff (see §3.3), and to the bubble shape via the geometric
charge q. The latter can be expressed exactly as a function of the total length of
the edges of the bubble and its mean width (see eq. (2.28)), and is well-correlated to
its number of faces. In the following, q(f ) will denote the average geometric charge
of all bubbles with the same number of faces in a disordered foam. Note also that
in a disordered foam, the number of faces of a bubble is closely correlated with its
volume (see p. 38), that is, large bubbles have more faces than small ones statistically,
and it is the largest bubbles that grow and the smallest bubbles which disappear (cf.
fig. 5.7b).
Evolution of the total foam volume For a foam of fixed volume (with no contact

with the outside, for example confined in a box), Vfoam = Ni=1 Vi is constant. Any gas
that leaves one bubble goes into another, and the gas flux must be zero on average at
any instant in time:

1
dVi
N
dVi
= = 0. (3.32)
N i=1 dt dt
However, in contrast to the 2D case (eq. (3.27)), this is not an automatic consequence
of eq. (3.30). In fact, eq. (3.30) determines dV 2/3 /dt, not dV /dt (see eq. (3.29)), and
besides, the average geometric charge on the foam in 3D is not zero, qi = 0.
In practice, the individual growth rate of each bubble (eq. (3.30)) and global gas
conservation (eq. (3.32)) are compatible if, and only if, the proportion of bubbles
which are growing, multiplied by their individual growth rates, equals exactly the
proportion of those that are shrinking, multiplied by their individual shrinkage rates.
More precisely, this imposes a constraint on the distribution of the geometric charge,
and thus on the numbers of faces of the bubbles and the correlations between neigh-
bours. Mullins [56] thus confirmed that his expression (eq. (2.27)), for an experimental
distribution of f , leads to gas conservation (eq. (3.32)).
On the other hand, if the foam has a free surface, as in a glass of beer, the aver-
age f is smaller, the peripheral bubbles bulge, the foam has a pressure greater than
atmospheric pressure, and overall it gradually loses gas.
3.2 Evolution of bubble distributions in a dry foam

Mullins coarsening affects the bubble size and the bubble size distribution in a dry
foam because of the occurrence of T2 topological processes (p. 81). The 2D case (von
Neumann), which is similar, is not described here. We assume here that Deff is constant
in time and uniform throughout the foam (its variations will be discussed in §3.3).
3.2.1 Inevitable disappearance of bubbles We follow a bubble i which has few

neighbours, for example fi < 12. It tends to lose gas, so its volume decreases. We first
assume that it can do so without changing shape, and therefore without changing its
geometric charge. Eqs. (3.29) and (3.30) would then imply that
2 Deff qi
Vi (t)2/3 = Vi (0)2/3 − t. (3.33)
3
An individual characteristic shrinkage time appears in this equation:
2/3
3Vi
ti = . (3.34)
2Deff qi
In practice, when the volume of the bubble decreases, it gradually loses neighbours
(fig. 2.15) through T1 topological changes (fig. 3.3). Eqs. (3.33) and (3.34) then
imply that Vi (t)2/3 = Vi (0)2/3 (1 − t/ti ). Each time a neighbour is lost, qi increases
2/3
and the shrinkage time ti decreases. Consequently, between two successive T1s, Vi
decreases like t, and Vi like t3/2 , but at each T1, the prefactor of t3/2 increases. So the
bubble shrinks even more quickly in a vicious circle. Eventually, the bubble disappears
in a T2 (see §1.2.3), usually in the form of a small tetrahedron (see fig. 5.7b).
We now consider a bubble j which was a neighbour of this small bubble i. It
had fj neighbours, but since its neighbour i disappeared, it finds itself with one less
neighbour (fj decreases by 1). If fj was large, larger than 15 say, and bubble j was
previously growing, it continues to grow, but now more slowly. If fj was 13 or 14, it
starts to shrink. If it was already shrinking, for example because fj was less than 12,
its shrinking accelerates.
To summarize, every bubble eventually loses neighbours, decreases in volume more
and more quickly, then disappears. For a bubble of volume V with a small number f
of faces which has begun to shrink, we estimate the time it takes to disappear from ti
(eq. (3.34)), up to an unknown factor of order one, taking into account the changes in
f which occur before the final T2:
V 2/3
tdisapp ≈ . (3.35)
Deff q(f )
3.2.2 From the individual to the group We now think about the foam as a
whole. At each moment in time, some bubbles are shrinking, whilst others are grow-
ing. From time to time, a bubble disappears (T2 process), the number of bubbles N
decreases by one, and the average bubble volume V = Vfoam /N increases discontinu-
ously. Since new bubbles are not created, N is always decreasing (step-by-step, at each
Coarsening 105
T2), and at the end of the process there remains just one large bubble containing all
the gas.
If the
foam is contained in a box, the coarsening proceeds at constant total volume
N
Vfoam = i=1 Vi (eq. (3.32)). Between two successive T2s, dVi /dt = 0 (eq. (3.32))
and V is constant, i.e. d Vi /dt = 0. At the instant at which a T2 occurs, we still
have dVi /dt = 0 (eq. (3.32)), but at the same time, because N varies, we also have
d
Vi
N
d V
= > 0. (3.36)
dt i=1 N dt
There is no simple expression for the rate of disappearance of bubbles (T2 rate) but
we can estimate it based on the lifetime of all the bubbles with f < 13 (which, since
they disappear most rapidly, are the most important contribution). The disappearance-
rate of bubbles with f faces is the number of bubbles of this type, N (f ), divided by
their characteristic lifetime given by eq. (3.35). We denote by .f the average over all
bubbles with the same f , and, by smoothing over time the discontinuities due to T2s,
obtain
dN
13
≈ −Deff N (f )q(f ) V −2/3 . (3.37)
dt f
f =4
Observe that the temporal evolution of N is not universal: it depends on the

distribution of the number of faces.
3.2.3 Asymptotic regime From the moment a foam is created, it begins to

coarsen and the average bubble size increases in time. We describe here the growth
of dry foams in 3D. The results also apply to wet foams (§3.3). They are analogous
to the coarsening of bubbly liquids (Ostwald ripening, see §1.1.3), and the 2D case is
comparable, although easier to study as all the bubbles are visible (fig. 3.23).
Evolution of the average size One experimental observation [57] (interpreted be-
low using eq. (3.39)) is that foams which have had time to coarsen reach an asymptotic
regime in which the average bubble size increases in a simple way. That is, at long
times, N varies as t−3/2 , the average bubble volume V varies as t3/2 , the average
surface area S varies as t, and the average diameter d varies as t1/2 (fig. 3.27). The
quantities V −1/3 dV /dt, dS/dt and ddd/dt become constant. More precisely,
it is observed that
d(t)2 = d(to )2 + K(t − to ). (3.38)
This averaged expression recalls the one for the growth of an individual bubble
(eq. (3.33)), but the actual link is not obvious since the growth of d is the result
of some bubbles growing and others shrinking. The averaged coarsening constant K
depends not only on Deff , and thus on the physico-chemical characteristics of the foam
(§3.3), but also on the bubble size distribution (via the parameter b in eq. (3.43) be-
low). In 2D, eq. (3.38) again holds: the average bubble radius grows as t1/2 and the
average bubble area as t (see exercise 8.1).
Fig. 3.23: Self-similar growth. Images of coarsening were obtained by leaving a 2D

foam on a photocopier. At the top: three successive images (1.95 then 21.5 and
166.15 hours) of a small part of an initially disordered dry foam. At the bottom:
with an initially ordered foam, the transition to self-similar growth takes longer, but
the final state is similar to the one in the top row. Reproduced by kind permission of
J. Glazier [33],
c 1987 The American Physical Society.
Evolution of distributions It is more difficult to study the change in the detailed

distribution of bubble sizes than the growth of the average size, both experimentally
(§1, chap. 5) and by simulation (§2, chap. 5). This is because it is necessary to observe
thousands of individual bubbles over a long time. In fact, the foam needs to coarsen
for a long enough time that it passes through a transient and reaches the limiting
regime; this requires that enough bubbles remain (despite coarsening) to be able to
characterize this limiting regime with good statistics.
So-called self-similar growth is defined as temporal evolution in which the dimen-
sionless statistical distributions remain stationary. That is, after the transient [45, 75],
in this limiting regime the foam structure at two different times has (geometric and
topological) statistical characteristics which differ only in their length-scale, i.e. the
average bubble size (eq. (3.38)). Two successive photos of the foam, if the scale is
not specified, appear alike (fig. 3.23). This self-similarity is demonstrated by sev-
eral statistical quantities, including the proportion N (f )/N of bubbles with f faces,
the relative volumes at a given time Vi /V (t), the correlations between the relat-
ive volume of a bubble and its number of neighbours fi , or even the correlations
between the relative volumes of two neighbouring bubbles Vi /V (t) and Vj /V (t)
(fig. 3.24).
The self-similar growth regime appears universal [45, 75]. The size distributions
and the number of neighbours don’t depend on the initial state of the foam (figs. 3.23
and 3.24). The transient is the time that the foam takes to completely lose any memory
of its initial configuration, through the disordering effects of T1s and T2s. This is why
the duration of the transient varies, and can be long if the foam is initially far from the
Coarsening 107
(a)
3 Normal Lognormal
2
1
P(V/<V>)
0
Bidisperse
3 1 t = 8,000 steps
100
2 500 Normal
1,000 Lognormal
1 5,000 Bidisperse
8,000
0
0 1 2 3 0 1 2 3
V/<V>
(c) 1 Lognormal
(b) 100
Lognormal 20 500
0.15 Normal 1 1,000
100 5,000
0.10 500 10 8,000
1,000
<V>f /<V>
0.05 Normal
5,000
0
P(f)
8,000
0.00 Bisperse t = 8,000 steps
0.15 Bidisperse t = 8,000 steps
Normal 20
0.10 Lognormal
Bidisperse 10 Normal
0.05 Lognormal
Bidisperse
0.00 0
0 15 30 45 0 15 30 45 0 20 40 60 80 0 20 40 60 80
f f
Fig. 3.24: Self-similar growth in 3D, simulated with the Potts model (§2.1.1, chap.
5) [75]. (a) Volume distribution; (b) distribution of number of neighbours; (c) volume
as a function of number of neighbours. Simulations are shown for three different initial
distributions: normal (Gaussian), log-normal, and bidisperse (mixture of two bubble-
volumes). The duration of the transient is different in each case. After 8,000 time
steps, all three have left the transient and show the same distributions, which remain
constant in time.
limiting regime. In extreme cases, for example in a very ordered foam, the transient
can be longer than the duration of the experiment, and the self-similar growth regime
is not observed.
The link between the evolution of bubble size and of distributions Mullins,
on the basis of dimensional analysis, suggested that in the self-similar growth regime
the average bubble volume should increase as t3/2 [57]. However, the converse is not
true: observing this exponent does not prove that, and in fact it often occurs before
[29, 45, 75], the self-similar growth regime has been reached. The growth of the average
volume with t3/2 (eq. (3.38)) is observed much more often than the self-similar regime,
and historically was recorded much earlier. More generally, in a self-similar growth re-
gime, any growth law compatible with the Young–Laplace law yields a growth exponent
of t1/2 for the bubble size [32], both in 2D and 3D.
The link between the self-similar growth regime and the exponent can be estab-
lished exactly in 2D (see exercise 8.1). In 3D, thanks to the approximate equation
(3.37), we can describe the self-similar growth regime by introducing the average
volume V = Vfoam /N , with Vfoam constant. We obtain
−2/3
N 5/3
N (f )
13
dN V
≈ −Deff 2/3 q(f ) , (3.39)
dt Vfoam f =4 N V (t)
f
from which we deduce that

dN
3D: ∝ −N 5/3 (3.40)
dt
with a prefactor that is constant in the self-similar growth regime, since N (f )/N and
(V /V (t))−2/3 f are then independent of time. Using eq. (3.40), we deduce that
N ∝ t−3/2 , and consequently the average volume increases as t3/2 . Thus the change in
the average volume governs all the time-scales of the system. Likewise, the topological
changes become less and less frequent during coarsening, the frequency varying with
t2/3 (fig. 3.25) [9].
The value of the prefactor is another piece of information that can be obtained from
eq. (3.39). The given form is already very approximate, and if we also take into account
the fact that q and V /V are of order one (see fig. 2.17), we obtain a prefactor of
2/3
the order of Deff /Vfoam . Then an approximate expression for the number of bubbles is
−3/2
−2/3
N (t) ≈ N (0) 1 + b Deff V 0 t (3.41)
where b is a constant of order one that depends on the bubble size distribution and
V 0 = V (0) is the average initial bubble volume. The bubble size then varies as
1/2
−2/3
d(t) ≈ d0 1 + b Deff V 0 t . (3.42)
70
50
30
τ (s)
10
8
6
0.66
4
100 101 102 103

Age of foam t (min)
Fig. 3.25: Average time interval between T1 topological changes, τ , in a coarsening

foam measured by multispeckle DWS (see §1.2.5, chap. 5). The sample is a Gillette
shaving foam with φl = 7.5%. After 20 min, τ ∼ t0.66 [9].
Coarsening 109

Finally, we can define a characteristic coarsening time tc such that d(t) = d0 1 + t/tc
by
2/3
V 0
tc ≈ . (3.43)
b Deff
tc depends on the effective diffusion constant Deff , and is related to the constant K in
eq. (3.38) by Ktc ≈ d20 .
3.3 Effects of different parameters

3.3.1 Time-scale and effective diffusion coefficient We explore the signific-
ance of the effective diffusion coefficient Deff , defined by eqs. (3.26) in 2D and (3.30)
in 3D, which fixes the time-scales associated with coarsening (eq. (3.43)). Following
Princen [63], we reason in the following way.
A film has one interface in contact with a bubble of pressure P1 and the other
in contact with a bubble of pressure P2 (fig. 3.26). On the liquid side of the inter-
face, the gas concentration is proportional to the pressure in the neighbouring bubble:
c1 = He P1 , c2 = He P2 , where the Henry coefficient He reflects the solubility of the
gas in the liquid. Within the liquid, the gas diffuses between these two values c1 and
c2 . Once a steady profile is established, the gas concentration therefore varies linearly
with z, with a gradient determined by the film thickness h:
∂c(z) He (P1 − P2 )
=− . (3.44)
∂z h
The flux of gas (per unit surface area of film) is determined by the actual diffusion coef-
ficient Df of the gas molecules within the liquid. The molar flux is Df He (P1 − P2 )/h
and so, with Vm the volume of a mole of gas at ambient temperature and pressure,
the volumetric flux is Vm Df He (P1 − P2 )/h. Eqs. (3.29) and (3.30) express the volu-
metric flux per unit surface area as Deff H12 = Deff (P1 − P2 )/2γ (using eq. (2.4)).
Comparing coefficients gives, for a dry foam,
2HeγVm
Deff = Df . (3.45)
h
Air Film Air
P1 flux P2 < P1
c1
c(z)
c2
z
0 h
Fig. 3.26: Profile of concentration of dissolved gas across a film (eq. (3.44)).
The effective diffusion coefficient for coarsening is thus proportional to the actual
diffusion coefficient of the gas molecules. The units of eq. (3.45) are correct: γ/h is a
pressure, while HeVm is the inverse of a pressure. A dimensionless geometric correction
factor for the liquid fraction must be taken into account (see §3.3.3), so that
2HeγVm
Deff = Df a(φl ). (3.46)
h
We will now comment on each term of this equation.
3.3.2 Effect of chemical composition The data in fig. 3.27 compare well to
an expression with an exponent 1/2 (eq. (3.42)), implying a characteristic time tc and
therefore a value of Deff (see eq. (3.43)) which depends linearly on the solubility He
and on the diffusivity Df , in agreement with expression (3.46).
It is generally assumed that the diffusion coefficient Df of the gas in the liquid
films is equal to that in the bulk liquid, and therefore independent of h, and inversely
proportional to the viscosity of the liquid. Since Df varies like the inverse square-root
of the molecular mass of the gas, in practice it does not depend much on the gas. In
pure water, Df lies between 10−6 and 10−5 cm2 .s−1 for common gases. The variation
of Df is responsible for only a factor of about 2 in the variation of the coarsening rates
shown in fig. 3.27.
3.5
3.0 SDS/N2
Casein/N2
2.5
d/do
SDS/C2F6
2.0
Casein/C2F6
1.5
1.0
0 4,000 8,000 12,000 16,000
Time (s)
Fig. 3.27: Effect of different surfactants and gases on coarsening. The liquid fraction
is kept roughly constant at φl = 0.15 by periodically inverting the sample to avoid
drainage. The average bubble diameter is measured as a function of time by diffuse
light transmission (see §1.2.5, chap. 5). An expression of the form d/d0 ∼ (1 + t/tc )1/2
(eq. (3.42)), where d0 is the average diameter at t = 0 (identical for the 4 foams),
is fitted to each set of data, allowing us to deduce a characteristic time tc = 190,
980, 4200, and 19100 s, respectively for the pairs SDS/N2 , casein/N2 , SDS/C2 F6 ,
and casein/C2 F6 . The coarsening dynamics therefore depend strongly on the physical
chemistry of the components (gas and surfactants) [66].
Coarsening 111
The Henry constant He depends strongly on the gas and the liquid. Thus, in pure
water, He = 3.4 × 10−4 mol.m−3 .Pa−1 for CO2 , whereas it is 6 × 10−6 mol.m−3 .Pa−1
for N2 and only 6 × 10−7 mol.m−3 .Pa−1 for a fluoro-carbon gas such as C2 F6 . The
variation of He is responsible for a factor of about 10 in the difference of the coarsening
rates between N2 and C2 F6 in fig. 3.27 (a CO2 foam would coarsen too quickly to
plot). An even more extreme effect of gas solubility is discussed in §3.3.4.
Surfactants affect eq. (3.46) directly via the surface tension γ. They also control
film thickness h via the disjoining pressure (see p. 48), and therefore affect the denom-
inator: the thicker the films, the slower the coarsening. For instance, in fig. 3.27, when
going from SDS to casein, the surface tension changes from 36 to 43 mN.m−1 , and
is responsible for only a small change in the coarsening rate. As shown in fig. 3.17,
casein results in inhomogeneous films a few hundreds of nanometres thick, which is
significantly different from the roughly 50 nm thick film for SDS and is thus responsible
for most of the change in the coarsening rate.
Experimentally, to test the effect of the physico-chemical conditions on the dynam-
ics of 3D foam coarsening requires that the liquid fraction φl is kept constant, which
is difficult (see §4). One possibility (provided the foam is not too wet and the bubble
diameter is less than a millimetre) is to periodically change the vertical orientation of
the foam sample, thereby changing the direction of drainage and keeping the liquid
fraction constant at the centre of the sample.
3.3.3 Effect of geometrical parameters We now fix the chemical components

of the foam and investigate how the dynamics depend on the initial bubble diameter d0
and on the liquid fraction φl .
As shown in eq. (3.43), the coarsening time tc depends strongly on the initial
bubble diameter: tc ∼ V 2/3 ∼ d2 . For bubbles of diameter 100 μm, tc is several tens of
seconds, but for bubbles of 1 cm in diameter, the coarsening time is multiplied by 104 !
Clearly there is practically no coarsening in such a foam.
Experimentally, for the reasons mentioned in §3.3.2, there are few measurements
of the coarsening rate at constant φl , either for different values of d0 (the initial
bubble diameter) or of φl . These two dependencies are coupled (see eq. (3.46)): Deff
depends on φl via the function a(φl ), but also because h depends on φl through the
value of d0 .
Film surface area The function a(φl ) is the proportion of the surface of the bubbles
covered by thin films rather than Plateau borders. It is through this effective surface
that gas diffuses from bubble to bubble. Diffusion across Plateau borders is much
slower because they are several orders of magnitude thicker. Two different expressions
for a(φl ) are used [36, 38], neither of which takes into account polydispersity. First,
a(φl ) = 1 − (φl /φ∗l )1/2 (model 1 in fig. 3.28). This comes from eq. (2.38), which
gives the size r of a Plateau border as a function of the liquid fraction: the area
of a film in a dry foam is of the order of S0 = d2 , which in a wet foam becomes
S(φl ) = d2 − d r = d2 (1 − (φl /φ∗l )1/2 ). Then a(φl ) = S(φl )/S0 and the expression fol-
lows [38]. This approach, which is based on orders of magnitude, does not enable
(ϕ1)
Model 1
0.1
Model 2
10–5 10–4 10–3 10–2 10–1

ϕ1
Fig. 3.28: Two models for the evolution of the effective surface area for gas transfer
as the liquid fraction varies. Model 1 has a(φl ) = 1 − (φl /φ∗l )1/2 , with φl = 0.36 [38].
2
Model 2 has a(φl ) = 1 − (φl /φ∗l )1/2 , with φ∗l = 0.43 [36].
a prediction of φ∗l , but explains the form of the dependence on φl . More recently,
2
numerical simulations have suggested a second expression, a(φl ) = 1 − (φl /φ∗l )1/2
(model 2 in fig. 3.28) [36].
Potts model simulations [29] (§2.1.1, chap. 5) indicate that the rate of gas dif-
fusion is the weighted sum of the contributions through films and Plateau borders.
Thus the rate at which an individual bubble grows is intermediate between the dry
and wet limits. Similarly, the exponent for the increase of the average bubble size
in the self-similar growth regime interpolates between t1/2 (eq. (3.38) and t1/3 , as
1/2 − φl0.2 /6 [29].
Film thickness The thickness h of a film depends on d and on φl , since they control
the capillary suction Pc ≈ γ/r of the Plateau borders (see §2.3.2). At equilibrium, the
disjoining pressure Πd , which governs the thickness of the√ films, is equal to the capillary
pressure, Pc . Using eq. (2.38) we find Pc = Πd ≈ γ/(d φl ).
However, a typical curve of Πd (h), as in fig. 3.15, shows that if Πd is high (of the
order of several hundreds of Pascals) a variation of Πd (induced by φl ) makes little
difference to h. This is the case for small bubbles (d ≤ 1 mm) or dry foams (φl ∼ a few
percent), where h(φl ) is almost constant, and in these cases the dependence of Deff on
liquid fraction is entirely due to the function a(φl ). In fact, a coarsening experiment
at constant liquid fraction φl , such as in fig. 3.27, allows us to infer (using eq. (3.46),
where the only unknown is h) that the thickness of the film is less than 100 nm (the
range of common black films) and independent of φl .
On the other hand, if the disjoining pressure is weak, of the order of several
tens of Pascals, the thickness h depends strongly on Πd , and consequently on φl .
In this case, which corresponds to wet foams or large bubbles, the dependence of Deff
on φl is then more complex than just the function a(φl ) and includes a non-trivial
contribution h(φl ).
Drainage 113
3.3.4 Can coarsening be controlled? In summary, which parameters have the

greatest effect on the dynamics of coarsening? Is it more effective to vary physical
parameters or the chemistry of the solution?
As we have seen, tc is strongly dependent on bubble size (see eq. (3.43)) and
on liquid fraction (see eqs. (3.43) and (3.46) and fig. 3.28). Changing the gas also
changes the dynamics, essentially by altering the solubility He (eq. (3.46)). Adding
to the usual gas (e.g. air or CO2 ) tiny amounts of a much more insoluble gas slows
down the coarsening rate, which is dominated by the slowest gas. It even changes the
bubble size distribution and its time evolution, which loses its self-similarity. Taken
to extremes, adding traces of the very insoluble gas C6 F14 dramatically changes even
the final state. In fact, the quantity of such an insoluble gas in each bubble remains
constant, and if there is enough of it then there is a coexistence volume at which
the Laplace pressure and the partial pressure of the trapped gas balance, and the
coarsening process stops [30, 83, 85].
Surfactants of low molecular mass appear to have a much weaker influence: in fact,
surface tension and film thickness only vary over a small range. Nevertheless, there
are particular systems in which surface concentration and surface tension depend on
bubble size. If the molecules adsorbed at the surface are irreversibly attached, then
when a bubble becomes smaller the surface concentration increases and we observe a
strong reduction in the coarsening rate. This is the case, in particular, for solid particles
or certain proteins [17]. In fact, the surface tension γ can continue to decrease and
coarsening can be stopped entirely [8]. This behaviour is linked to the elasticity of the
surface in compression (see §2.2.3).
Finally, while a foam is coarsening, gravity-driven drainage may also occur, redis-
tributing the liquid in the foam (see §4). It is therefore clear that the coarsening rate,
which is sensitive to φl , will be modified by drainage. Conversely, coarsening changes
the average bubble diameter, a parameter upon which drainage is strongly dependent.
The two processes are therefore strongly coupled. We will see (§4) that it is possible to
define a characteristic drainage time, so that comparing typical coarsening and drain-
age times for a given foam enables us to determine whether or not we need to take
this coupling into account.
4 Drainage
The aim of this section is to present a description and model of liquid drainage through
a foam. We will show that it is comparable to the problem of liquid flow in porous
media, but with two significant differences. First, in a foam, the liquid flows through
a network of pores whose diameter depends on flow itself: the bubbles can move apart
to allow liquid to pass and then move back. Second, the interfaces are fluid, and are
partially entrained by the flow. From this point of view, drainage is a good example
of coupling between different length-scales in a foam: to model this macroscopic phe-
nomenon, we will see that it is necessary to take into account the properties of the
interfacial monolayer that forms the pore surfaces and the characteristic size of the
pores themselves.
4.1 What is drainage?

Drainage is the passage of liquid through a foam. It is easily observed in daily life, in
beer foam for example. Gravity-driven drainage is a mechanism by which an aqueous
foam evolves: as noted in §1.1, under the effect of gravity the liquid contained in the
foam flows downwards and the gas bubbles are displaced upwards.
4.2 Free drainage

The evolution of the liquid fraction of a foam under gravity towards the equilibrium
profile (§4.3, chap. 2) is known as free drainage. Fig. 3.1 shows a foam sample of
height H0 and of uniform liquid fraction φl0 at t = 0 in a container with a closed base.
Over time, the downward flow of the liquid dries the foam and the liquid accumulates
at the base of the container. The interface between the foam and the drained liquid
rises over time (height L(t) on fig. 3.1). At a local scale, by watching the surface of
the foam through the wall of a transparent container, it is possible to see the liquid
network gradually emptying and the bubbles moving closer together.
The liquid fraction at a given vertical position z in the foam can be estimated
by the Plateau border size (see §1.2.3, chap. 5) or measured more precisely by foam
conductivity (see §1.2.6, chap. 5 and experiment 7.4), which gives the average liquid
fraction in a slice of foam of thickness dz. The profiles obtained are shown at successive
times in fig. 3.29. The behaviour is complex, but two important stages of the drainage
process are visible, defined by the time τ at which the drainage front reaches the bottom
of the foam:
(i) For t < τ , the liquid fraction is only modified in the upper part of the foam and
varies linearly with z between a small value and its original value, φl0 . The lower
limit of this region of dry foam is known as the drainage front. It moves downwards
at constant velocity as the foam drains, and the width of the front increases (the
slope dφl /dz is reduced). The lower part of the foam undergoes steady drainage,
with constant liquid fraction. That is, the same amount of liquid leaves as enters
this region and the velocity of the liquid there is constant.
(ii) For t > τ , the drainage is characterized by a slow relaxation towards the equilib-
rium profile described in §4.3, chap. 2. To explain this slowing down, we note that
gravity is not the only force acting on the liquid. In a drier foam, the radius of
curvature of the Plateau borders is smaller, and the Young–Laplace law indicates
that the capillary pressure is higher (see §4.1.2, chap. 2), so that the liquid frac-
tion gradient gives rise to a capillary pressure gradient. Dry regions imbibe liquid,
independent of the effects of gravity. The capillary pressure gradient (directed
from the bottom upwards) increases during drainage, and reduces the drainage
rate until the two are in balance, thus leaving the foam in static equilibrium. At
equilibrium, in contrast to the initial state, the part of a foam in direct contact
with the drained liquid is very wet. At the interface with the liquid, the liquid
fraction is close to 0.36, which corresponds to the void fraction between randomly
close-packed spheres (see fig. 4.17).
Drainage 115
t=0
e
Tim
z
t=
τ
0
0 0.14
Liquid fraction
Fig. 3.29: Free drainage experiment: vertical profile of liquid fraction as a function
of time measured by electrical conductivity (see p. 238) in a foam of height 40 cm.
At t = 0, the liquid fraction is uniform and constant, φl0 = 0.14. Drainage is evident
in the downward propagation of a front (shown by the arrow) that reaches the base
of the foam at t = τ . The bottom of the foam gets wetter and φl increases above
its initial value. The equilibrium profile is reached eventually. Graph courtesy of A.
Saint-Jalmes.
No liquid leaves the foam if the liquid fraction at the base is less than the value
φ∗l = 0.36. This wetting phase can take a relatively long time and if the total quantity
of liquid in the foam is too low (a dry foam, or a small volume of foam), it is even
possible that no liquid drains at all! Nonetheless, even if no liquid leaves, drainage
may still occur and give rise to liquid fraction gradients (see exercise 8.3).
4.3 Forced drainage

Free drainage, although conceptually straightforward, is not the best experiment
through which to understand drainage, since there are regions of very different liquid
fraction present simultaneously (from very dry to very wet). Moreover, the experiment
can take a long time, so that other ageing effects are often at work, complicating the
interpretation of the results. Finally, the initial profile of liquid fraction has a signi-
ficant influence on the subsequent dynamics, so that to generate reproducible results
requires that foams must be generated with an initial liquid fraction that is uniform
across the full height of the foam.
Instead, a forced drainage experiment can be performed, which consists of pouring
the foaming solution at constant flow rate into a foam that has already been allowed to
wet
foam
dry
foam
Fig. 3.30: Forced drainage experiment: liquid is introduced at the top of an initially
dry foam of height 10 cm in a vertical cylinder. Three photos taken at successive times
show, by diffuse light transmission (§1.2.5, chap. 5), the progress of a wetting front,
enabling the liquid velocity in the foam to be measured and its permeability to be
determined. Photograph courtesy of O. Pitois.
drain freely. A horizontal wetting front develops, separating wet foam (at the top) from
dry foam (fig. 3.30). The front moves downwards at a constant speed ufront which
is easy to measure. Above the front, φl is constant and depends on the imposed flow
rate, and the liquid therefore flows only under the force of gravity. The experiment
shows that ufront depends on the bubble size and the surfactant. We will make this
more precise later (§4.7.1), but for the moment it is worth noting that the influence
of the surfactant is explained by the partial mobility of the interfaces which confine
the flow, that is, the Plateau border walls. These interfaces may be entrained by the
liquid and then the surfactant plays an important role.
4.4 Modelling flows in solid porous media

Modelling of foam drainage is inspired by the field of flow in solid porous media, and
in the following we present some of the significant results in that field.
4.4.1 Flow at the scale of a pore—Poiseuille’s law We consider the flow of

liquid (with viscosity η and density ρl ) through a cylindrical pore of radius rp and
length , oriented at an angle ψ to the vertical (fig. 3.31). eψ denotes the unit vector
along the pore axis, oriented upwards. The velocity field in the liquid is approximately
of the form u = u(r)eψ , where r is the radial coordinate. The force acting on the liquid
contained in the pore due to the pressure p is πrp2 (p(0) − p()). Per unit volume, this
force is equal and opposite to the pressure gradient along the axis of the pore, −dp/d.
The force of gravity (per unit volume) along the pore is ρlg · eψ = −ρl g cos ψ. In a
foam we can safely assume that the Reynolds number Re, based on the characteristic
speed of the liquid u and from rp , is small compared to one, Re = ρl urp /η 1. In
Drainage 117
p(l)
ψ
u(r)eψ
g
p(0)
eψ
ez
Fig. 3.31: A cylindrical pore of length , oriented at an angle ψ to the vertical, thr-
ough which liquid flows under the action of pressure and gravity forces. The velocity
field is u = u(r)eψ .
this case, the flow is governed by the equilibrium between the driving forces (pressure
and gravity) and viscous forces, which oppose the flow, giving

1 d du dp
η r = + ρl g cos ψ. (3.47)
r dr dr d
Assuming that the velocity is zero at the wall (no-slip boundary conditions), the
solution is

rp2 r2 dp
u(r) = − 1− 2 + ρl g cos ψ . (3.48)
4η rp d
The direction of flow is determined by the sign and magnitude of the pressure gradient:
if it is positive, it enhances the effect of gravity; if it is negative, it opposes it and can
even negate it. The flux of liquid through the pore is obtained by integration, giving
the average velocity of liquid in the pore:

s dp
ū = u ds = − Kp + ρl g cos ψ , (3.49)
η d
where the cross-section s = πrp2 of the pore and a geometric coefficient Kp = 1/(8π)
have been introduced. This relationship indicates that the average liquid velocity is
directly proportional to the driving force, which is known as Poiseuille’s law.
4.4.2 Flow at the scale of the porous medium—Darcy’s law Henry Darcy
showed in 1856 that the relationship between velocity and driving force at the scale of
the pore (eq. (3.49)) remains true if we average the velocity and the pressure over a
collection of pores within a representative volume of the porous medium. This result
is known as Darcy’s law :
α

ume = ρlg − ∇p , (3.50)
η
where α is the permeability of the medium. The minus sign arises because the liquid
moves from regions of high pressure to low pressure. Here um is the flux, defined as the
liquid flow rate divided by the surface area of the sample transverse to the direction
of flow, which is denoted e. um depends on the average liquid velocity ū and the
volume fraction of the medium occupied by liquid, φl , according to
um = ū φl , (3.51)
where the average represented by the angle brackets is over all the pores in the
representative volume element.
Note that we assume that the pores of the medium are completely saturated with
liquid, so the porosity plays the same role as the liquid fraction, justifying our choice
of notation.
4.4.3 Permeability of a porous medium Whilst simple to use, eq. (3.50) still
requires knowledge of the permeability coefficient α. The latter expresses the ease
with which a liquid flows through a porous medium. Intuitively, we expect that α
increases with φl and hence with the radius of the pores. To estimate it, we consider a
model consisting of a random network of identical cylindrical pores. We will calculate
the flow of liquid in the vertical direction (ez ) under the action of a macroscopic
pressure gradient ∇p = dp/dz ez and the force of gravity ρlg = −ρl gez . In a pore
with orientation ψ, the microscopic pressure gradient is dp/d = (dp/dz) cos ψ and the
average vertical velocity is

s dp
ūψ
z = ū
e ψ ·
e z = ū cos ψ = − Kp + ρl g cos2 ψ. (3.52)
η dz
The average vertical velocity of liquid in the medium is the average of ūψ
z over all
orientations:
π/2
ūz = ūψ
z p(ψ)dψ. (3.53)
0
Here p(ψ)dψ is the probability that ψ falls between ψ and ψ + dψ. To determine
this probability, we calculate the relationship between the surface area at the ends
of those pores inclined at an angle between ψ and ψ + dψ, equal to 2πl2 sin ψdψ,
and the surface area of the hemisphere including the ends of all the pores. Thus
p(ψ)dψ = 2πl2 sin ψdψ/2πl2 = sin ψdψ, and the average vertical velocity is
π/2
s dp s dp Kp
ūz = − + ρl g Kp cos ψ sin ψdψ = −
2
+ ρl g . (3.54)
η dz 0 η dz 3
Drainage 119
To relate this to the flux um , we use eqs. (3.50) and (3.51) to give

um α dp
ūz = =− + ρl g . (3.55)
φl ηφl dz
Comparing with eq. (3.54) shows that the permeability is
sKp φl
α= , (3.56)
3
linking the macroscopic permeability of the medium to the geometric and hydraulic
characteristics of its pores. For the cylindrical pores considered here, the permeability is
rp2
α = φl . (3.57)
24
More elaborate models combine several pore sizes in order to form a more realistic
network. It is also possible to expand the cylindrical pore model by using the concept of
hydraulic radius, m, the ratio of the volume of a pore to its surface area. For example,
for cylindrical pores of radius rp , m = rp /2. We generally express m as a function of
the specific surface area of the porous medium, As , defined as the ratio of the surface
area of the pores to the total volume:
φl
m= . (3.58)
As
The expression for α retains the form given in eq. (3.57) even though the pore geometry
may change, so we can write
m2 φ3l
α = φl = . (3.59)
CK CK A2s
This relationship, known as the Carman–Kozeny model, introduces a constant CK
which represents the geometry of the network. For cylindrical pores we find CK = 6,
from eq. (3.57). In a more complex network this constant is determined by fitting
eq. (3.59) to experimental data. For a packing of spheres with diameter d, the specific
surface area is As = (1 − φl )(6/d) and CK is close to 5; then the permeability of a
medium consisting of spherical grains (a model often used for soil) is
φ3l d2
α= . (3.60)
180(1 − φl )2
This relationship provides a reference value for studying the permeability of many
different porous media.
4.5 Modelling the permeability of a liquid foam

4.5.1 Foam: a special porous medium Foam can be viewed as a porous me-
dium: the liquid forms the interstitial phase, and the bubbles play a similar role to
grains in a solid medium. There are nevertheless important differences between a foam
and a solid porous medium:
1. The liquid network in a foam is special, in that its geometry is fixed by Plateau’s
laws (cf. chap. 2). In particular, eq. (2.38) links the dimensions of a Plateau border,
i.e. its radius of curvature r and length , to the liquid fraction (or porosity) φl
defined at the macroscopic scale.
2. The fluid nature of the interfaces which make up the liquid network allows the foam
to expand or contract. Within this network, the Young–Laplace law directly links
the pressure p of the liquid, which causes the expansion, to the radius of curvature
r (eq. (2.30)):
γ γ
p=P − ≈ Patm − , (3.61)
r r
where P is the bubble pressure. In the context of deformable porous media, this
relationship between liquid pressure and the space available for flow is the simplest
of its kind.
3. The mobility of the liquid/gas interfaces poses a more significant challenge for
models of foam drainage. The flow of a liquid in contact with a solid wall is governed
by a no-slip condition (zero velocity at the wall), but at the surface of a bubble
there is instead a condition on the stress, since the flow in a Plateau border exerts
a tangential viscous stress on the interface, pulling it in the direction of flow. The
surfactant molecules adsorbed on an interface change the rheological behaviour of
the interface (see §2.2.3), changing the resistance to flow; the surface shear viscosity,
analogous to the bulk shear viscosity, appears to play the most important role in
this resistance [47].
We now take account of these different effects to estimate the permeability of a
liquid foam. We first consider the flow in a single Plateau border, and then modify the
models presented in §4.4.3 for solid porous media.
4.5.2 Flow in a Plateau border The permeability coefficient for a Plateau bor-
der must take into account not only its shape but also the way in which the liquid/gas
interface is entrained by the flow. For the problem of drainage, the effect of interfacial
mobility was considered in the 1960s by Leonard and Lemlich [47], who considered
an infinitely long Plateau border. Then the flow can be considered to be in the axial
direction and the surface concentration of surfactant is constant. The bulk viscous
stresses must be in equilibrium with the viscous shear stresses at the interface, a coup-
ling between bulk and surface flow that is described by the dimensionless Boussinesq
number
ηs
Bo = , (3.62)
ηr
where ηs is the surface shear viscosity (§2.2.3) and η is the bulk viscosity (see shaded
section below). Note that the value of Bo depends on the radius of curvature r, and
therefore on the geometry of the foam, not just the chemical composition. Leonard
and Lemlich make the additional hypothesis that the liquid velocity is zero along the
three lines where a Plateau border meets the films.
Drainage 121
Fig. 3.32: Element of an interface lying in the (x, y) plane subjected to a surface
stress.
Equilibrium of viscous stresses at the Plateau border walls

To better understand the distortion of the gas/liquid interface that constitutes the
wall of a Plateau border, we consider the example of steady parallel flow of two
fluids 1 and 2 which flow in the y direction with velocity vectors vy1 and vy2 . Any
given element of the interface between the fluids, of area dxdy, is subjected to the
action of forces on its edges, and must be in equilibrium. For this two-dimensional
s s
analysis, we introduce forces per unit length, σxy (x, y) and σyy (x, y), shown in
fig. 3.32. The equilibrium of forces is:
s s
σxy (x + dx, y) − σxy
s
(x, y) dy + σyy (x, y + dy) − σyys
(x, y) dx

− σ 1 dxdy + σ 2 dxdy = 0, (3.63)
yz 0 yz 0
1
where σyz |0 and σyz
2
|0 are the viscous stresses that the bulk exerts on the interface
at z = 0 for fluids 1 and 2 respectively. Rewriting the finite differences in eq. (3.63)
as gradients of the surface stress and dividing by the area of the element, we obtain
s
dσxy s
dσyy 1
2

+ − σyz 0
+ σyz 0
= 0. (3.64)
dx dy
The bulk shear stresses evaluated at the interface are

1
dvy1
2
dvy2
σyz 0 = η1 and σyz 0 = η2 . (3.65)
dz dz
0 0
s
If the interface is ideal then the surface stress σxy (x) is proportional to the interfacial
0
velocity gradient dvy /dx (eq. (3.14)), so
s
dvy0
σxy (x) = ηs . (3.66)
dx
s
Note that the surface stress σyy is none other than the surface tension γ (see
eq. (3.15)). If an elongation is coupled to the shear, a correction appears (eqs. (3.10)
and (3.12)) and induces a Marangoni effect (see §2.2.4), which we do not take into
account here. We then have

d2 vy0 dvy2 dvy1
η s 2 + η2 − η1 = 0. (3.67)
dx dz dz
0 0
If fluid 1 is water and fluid 2 is air, we set η1 = η and neglect η2 . We introduce a

characteristic length corresponding to the radius of curvature of the Plateau borders,
r, and a characteristic velocity r2 | dp
d |/η to write this equation in dimensionless form
using the Boussinesq number (eq. 3.62):

d2 ṽy0 dṽy1
Bo 2 = . (3.68)
dx̃ dz̃
0
With these hypotheses, solving the equations of fluid motion enables us to obtain
liquid velocity profiles [16, 42],shown in fig. 3.33, and to determine the average velocity
and the Plateau border permeability coefficient, Kp (eq. (3.49)), shown in fig. 3.34, for
different values of Bo. These figures highlight the influence of Bo on the flow through
the Plateau border:
• For Bo 1, the profile is close to a plug flow (i.e. it is flat over most of the cross-
section), and velocity gradients are only apparent in the corners of the Plateau
border. This corresponds to high mobility of the wall in response to the bulk flow.
It is accompanied by an increase in the average velocity of the liquid and in the
value of Kp .
• For Bo 1, on the other hand, the interface is almost immobile. It is barely
entrained and resists flow in such a way as to generate high velocity gradients
across the entire cross-section. This type of flow is close to a Poiseuille flow (with
zero velocity at the wall).
. . .
. . . . .
. . . .
. . .
. . . .
. .
Fig. 3.33: Velocity profiles in a Plateau border for three values of the Boussinesq num-
ber, Bo = 10, 1, and 0.1. The corners of the Plateau border are fixed. The greater Bo,
the more the interface resists the flow. Simulations by W. Drenckhan. See Plate 8.
Drainage 123
bulk Plateau border

surface Plateau border
1
Kp
0.1
0.01
0.01 0.1 1 10 100
Bo
Fig. 3.34: Evolution of the permeability coefficient Kp with Boussinesq number for a
bulk Plateau border (solid line) and a surface Plateau border (dashed line). After [42].
Some authors prefer to use the interfacial mobility M = 1/Bo [66, 73], which is small
for immobile surfaces and large for mobile ones.
More quantitatively, the minimum value of Kp , found when Bo 1, is 0.02
(fig. 3.34). This corresponds to pure Poiseuille flow in the Plateau border and is
less than for a pore with a circular cross-section, 1/8π
0.04. This difference in Kp
indicates that a greater proportion of liquid is in contact with the surface and con-
sequently that the flow velocity decreases. For smaller Bo, Kp can be approximated
by the analytic expression [42]

1 3 2 1 1
Kp = + arctan − arctan . (3.69)
6 25 Bo 8 Bo 2π Bo
In fig. 3.34, we also show Kp for a Plateau border in contact with a solid wall (a surface
Plateau border, see §5.1, chap. 2). Here again, Kp increases as Bo decreases, but less
than in the bulk case because the zero velocity condition is imposed along the entire
area of contact with the solid wall, irrespective of Bo. From these curves it is possible
to estimate the influence that the surface Plateau borders have on the permeability
of a foam (see exercise 8.4): to ensure that any measurement accurately represents
the bulk permeability, the liquid flow in the bulk Plateau borders must represent a
significant fraction of the total flow. We find that for a foam with about 30 bubbles
across the diameter of the container, this proportion reaches 80% for Bo = 10 and 95%
for Bo = 0.1.
4.5.3 Plateau border-dominated case Firstly, we will focus on the case of dry
foams (φl ≤ 0.02), in which the liquid network is composed of elongated Plateau bor-
ders with radius of curvature r and length . Eq. (2.38) allows us to describe the link
between r and the bubble radius RV :
r δb 1/2
= g1 (φl ) = φl , (3.70)
2RV 2
where δb = 1.73 is a geometric constant. We consider a precise structure, the Kelvin

foam (see p. 50), to estimate the relative length of the Plateau borders :

= 0.36. (3.71)
2RV
It is then possible to confirm that for φl = 0.01, for example, the Plateau borders are
indeed long and thin, since ≈ 4r.
The geometric analogy with the network of cylindrical pores described in §4.4
suggests the use of the expression for the permeability α of a solid porous medium
(eq. (3.50)) as a function of φl , the permeability coefficient Kp , and the pore cross-
section s (eq. (3.56)). For a Plateau border, we have derived an expression for Kp
above and we note that its cross-section is
√ π 2
s= 3− r ≈ δa δb2 RV2 φl , (3.72)
2
√
with δa = 3 − π/2 ≈ 0.161. Then
sKp φl δa δb2 Kp 2 2
α= = RV φl = Cp RV2 φ2l , (3.73)
3 3
which defines the constant Cp = δa δb2 Kp /3 ≈ 0.16 Kp . In the special case of completely
immobile interfaces, for which Kp = 0.02, the foam permeability is
α = 3.2 × 10−3 RV2 φ2l . (3.74)
As we will see in §4.7, this dependence of permeability on φ2l is compatible with

experimental results. Given the simplicity of expressions (3.73) and (3.74), we often
use this limiting case, based on just the Plateau borders (or channels), to estimate the
permeability, even though the hypotheses of the model limit its validity to relatively
dry foams.
Finally, note that when Bo decreases, Kp increases, and therefore so does the
permeability of a foam. There is a limit to this increase since Bo cannot decrease
indefinitely: for a foam with bubbles of one millimetre in diameter, liquid fraction
φl = 0.01, and in which the interfaces are characterized by a very low surface viscosity,
ηs = 10−8 kg.s−1 , we obtain Bo = 10−1 if the liquid which drains has a viscosity close
to that of water. Eq. (3.69) indicates that the liquid velocity will be around 15 times
greater than if the interfaces were completely immobile. Of course, this estimate is
based on the hypothesis that only the Plateau borders slow the liquid down, and that
the vertices which connect them do not participate. This hypothesis is legitimate when
the Plateau borders have low permeability and occupy a large proportion of the total
volume of liquid in the foam. If these conditions are not satisfied, it is necessary to
take into account the effect of the vertices [43], as we do below.
Drainage 125
4.5.4 Network model of Plateau borders and vertices We now derive expres-
sions for the permeability of an assembly of Plateau borders and vertices, considering
as the basic element of the liquid network a Plateau border with one quarter of a vertex
at each end (since a vertex connects four half-Plateau borders) [7, 43, 48, 59, 69].
The pressure gradient necessary to impose a flow rate Q = δa r 2 ū through a Plateau
border is
δp Rp + Rv /2 Rp + Rv /2
=− Q=− δa r 2 ū, (3.75)

where Rp and Rv are the hydraulic resistances of a Plateau border and a vertex,
respectively, which link the pressure difference between the ends of these elements to
the liquid flow through them (in the absence of gravity). We determine an expression
for the permeability by returning to the calculation in §4.4.3: eq. (3.49) is replaced by
eq. (3.75), and eq. (3.56) becomes
φl η
α= . (3.76)
3 (Rp + Rv /2)δa r 2
We introduce the dimensionless parameters
r 4 Rp 1
R̃p = = 2 (3.77)
ηpb δa Kp
and
r 3 Rv
R̃v = (3.78)
η
where pb is the Plateau border length, which differs from because it is necessary
(when the liquid fraction is greater than φl
0.02) to account for the decrease in
length of the Plateau borders as the vertices swell. In order to describe this change, we
generally resort to a function ξ(φl ), the form of which will be discussed after eq. (3.81),
which improves upon eq. (3.71):
pb r
= 0.36 − ξ(φl ) . (3.79)
2RV 2RV
In the same way, at high liquid fraction we obtain a more appropriate expression for
r by altering the exponent in eq. (2.37):
r
= g2 (φl ) = 0.62 φl0.45 . (3.80)
2RV
Then the foam permeability is
−1
0.48 RV2 φl g2 (φl ) R̃v 1 0.36
α= + 2 − ξ(φl ) . (3.81)
δa 2 δa Kp g2 (φl )
This expression is more general than eq. (3.73) and allows us to describe foam per-
meability over a wide range of liquid fractions, typically between 0.001 and 0.1, for
different types of surfactant.
We have introduced two additional dimensionless parameters, R̃v and ξ(φl ). The
former captures the viscous dissipation in the vertices of the network, for which there
is currently no precise model; several experimental and numerical results seem to
indicate that R̃v is of the order of several hundreds [10, 48, 69] and can reach several
thousand [43], without specifying its dependence on B0 (cf. eq. (3.69)). The latter,
ξ, accounts for the decrease in length of a Plateau border when the liquid fraction
increases. In dry foams (φl ≤ 0.02) ξ is treated as a constant between 1.5 and 2.3; for
larger values of φl , pb slowly decreases to zero at φ∗l ≈ 0.36. One choice for φl ≤ 0.1 is
a quadratic polynomial, ξ(φl ) = aφ2l + bφl + c with for example a = 52.5, b = −13.2,
and c = 2.24 [49].
In a dry foam, we can set g2 (φl )
g1 (φl ) (cf. eqs. (3.70) and (3.80)), ξ(φl )
0,
and R̃v ≈ 0 to recover the Plateau border-dominated case introduced in the previous
section (eq. (3.73)). Conversely, imagine a situation in which it is the vertices which
determine the flow rate (the resistance of the vertices is larger than that of the Plateau
borders), even in a dry foam where the contribution of the vertices to the liquid fraction
is negligible. Then we obtain the vertex-dominated case, with
5.2 2 3/2 3/2
α
RV φl = Cv RV2 φl , (3.82)
R̃v
3/2
which defines the constant Cv = 5.2/R̃v . Now the permeability varies with φl , in
contrast to the φ2l dependence found in the Plateau border-dominated case (eq. (3.73)).
3/2
Several experimental results show such a variation in permeability with φl over a
significant range of liquid fractions, particularly for large bubbles and mobile interfaces.
4.5.5 Carman–Kozeny model The Carman–Kozeny model, as we saw in §4.4.3,

expresses the permeability of a solid porous medium in terms of its specific surface area
As (eq. (3.59)). This macroscopic parameter can be measured directly (by determining
for example the isotherm of nitrogen sorption in the porous medium) and therefore
offers an attractive description of the geometry of the medium at a microscopic scale.
Adapting the model to liquid foams is relatively easy, as the surface area can be
obtained from simulations (§2.1.2, chap. 5). The interfacial area in contact with the
flowing liquid is the total surface area of the bubbles minus the area of the thin films.
This is because the films do not actively contribute to drainage: even if there is a liquid
flow there, the flow rate is negligible in comparison to that in the network of Plateau
borders and vertices. Thus [61]

3 S(φl ) Sf (φl )
As = − , (3.83)
RV So So
where S(φl ), Sf (φl ), and So are the total surface area of a bubble, the surface area
of this bubble that is covered by thin films (see fig. 3.28), and the surface area
of a spherical bubble of the same volume respectively. Fig. 3.35 shows the specific
Drainage 127
2.5
Non-dimensional specific surface area

2.0
1.5
1.0
0.5
0
0 0.1 0.2 0.3
Liquid fraction
Fig. 3.35: The dimensionless specific surface area of a Kelvin foam as a function of
liquid fraction [61].
surface area of a Kelvin foam as a function of the liquid fraction. Although wet foams
do not necessarily have a Kelvin structure, we expect the trends to be similar: (i)
when the liquid fraction increases, the interfacial area decreases (the bubbles are more
and more spherical); (ii) at the same time, the surface of each bubble covered by films
decreases strongly, giving rise to an increase in the specific surface area; (iii) at high
liquid fraction these areas only change very slightly, unlike the total volume of the
system, and so As reaches a maximum close to φl
0.2, as shown in fig. 3.35. The
foam permeability is then deduced from eq. (3.59) [61]. Since this approach does not
account for the mobility of the interfaces, the permeability obtained corresponds to
the case of immobile interfaces (Poiseuille flow).
4.6 Drainage equations

Knowing the permeability of a foam allows us to use mass conservation to write down
a corresponding drainage equation to describe the spatio-temporal evolution of the
liquid fraction φl (x, t).
Consider a fixed volume of foam with liquid fraction φl . We write ρl for the density
and u for the local velocity of the liquid, and assume that the liquid is incompressible.
The equation of continuity states that changes in liquid fraction are balanced by a flux
in or out of the foam:
∂φl
+ ∇ · (φl u) = 0. (3.84)
∂t
The liquid flows under the action of gravity ρlg and pressure −∇p. Darcy’s law (§4.4.2)
relates the liquid flux um = φl u to these forces in terms of the permeability α and the
viscosity η. In addition, the Young–Laplace law (eq. (2.4)) relates the liquid pressure
= −∇(γ/r),
to the curvature of the Plateau borders, ∇p so that
α
um = γ .
ρlg + ∇ (3.85)
η r
The radius of curvature r is linked to φl by eq. (3.70), giving

α γ −1/2
um = ρlg + ∇ φl . (3.86)
η δb RV
Inserting this in the conservation equation (3.84), we obtain the general drainage
equation

∂φl ρl γ · α l = 0,
+ ∇. (αg ) − ∇ 3/2
∇φ (3.87)
∂t η 2δb RV η φl
describing the evolution of the liquid fraction under gravitational and capillary forces.
The important parameter is the permeability α (eq. (3.81)). Eq. (3.87) is usually
studied in the limiting cases of Plateau border-dominated and vertex-dominated flow
(eqs. (3.73) and (3.82) respectively).
This drainage equation is three-dimensional; for simplicity we consider just the
vertical drainage by projecting eq. (3.87) in the z direction. In the Plateau border-
dominated case, we use eq. (3.73) to give

∂φl Cp ∂ γ 1/2 ∂φl
− ρl gφ2l RV2 + RV φl = 0. (3.88)
∂t η ∂z 2δb ∂z

This is rendered dimensionless by introducing a characteristic length ẑ = λc = γ/ρl g,
√
time t̂ = ηδb2 / γρl gCp , and liquid fraction φ̂l = λ2c /RV2 δb2 , giving [34, 78, 82]
1/2
∂ φ̃l ∂ 2 φ̃l ∂ φ̃l
− φ̃l + = 0. (3.89)
∂ t̃ ∂ z̃ 2 ∂ z̃
In the vertex-dominated case, we change the non-dimensionalization to use Cv instead

of Cp in the characteristic time, and use eq.(3.82); then eq. (3.87) becomes [43]

∂ φ̃l ∂ 3/2 1 ∂ φ̃l
− φ̃l + = 0. (3.90)
∂ t̃ ∂ z̃ 2 ∂ z̃
4.7 Comparison of theoretical predictions with experiments

Most drainage experiments are free drainage experiments, but forced drainage (§4.3)
results in a considerable simplification of the drainage equations. We will therefore
begin by comparing the models with forced drainage experiments of the type presented
in §4.1 before turning to free drainage.
Drainage 129
4.7.1 Forced drainage With a boundary condition of a constant flow rate at the
top of the foam and an initial condition of uniform liquid fraction φl0 at t = 0, the
solution of eqs. (3.89) or (3.90), expressed in terms of the function tanh2 (z) or exp(−z)
respectively [43, 82], shows a uniform value for the liquid fraction at the top of the
foam which decreases rapidly to φl0 in a transition zone or wetting front (fig. 3.36).
The front velocity, ufront , is the average vertical speed of the liquid in the foam,
that is ufront ≈ ūz , and is directly linked to the foam permeability α. The liquid
fraction is uniform above the front, so in this region the liquid flows solely due to the
force of gravity (there is no capillary pressure gradient). Then from eq. (3.55) we have
α ρl g
ufront ≈ . (3.91)
φl η
In the Plateau border-dominated case (eq. (3.73)), where the contribution of the ver-
tices is neglected, and in the vertex-dominated case (eq. (3.82)), where the contribution
of the Plateau borders is neglected, the front velocities are therefore
ρl gRV2 ρl gRV2 1/2

ufront,p = Cp φl ufront,v = Cv φl . (3.92)
η η
(a) z (b)
100
Vertical position (arbitrary units)
Front
80
60
40
g
20
0
0 0.2 0.4 0.6 0.8 1.0 1.2
Normalised liquid fraction
Time
Fig. 3.36: Forced drainage. (a) The advancing wetting front in an experiment. The
image is obtained by extracting a vertical line from each of a series of images (such
as those in fig. 3.30) and then placing them side by side. This spatio-temporal image
(time is the abscissa) enables the speed of the front to be determined directly from
the gradient of the line separating the light region (initially dry foam, with low liquid
fraction) from the dark one (wet foam). (b) The Plateau-border dominated drainage
equation (3.89) predicts that the wetting front moves at constant speed towards the
bottom of the foam, as represented by the arrow. The liquid fraction is normalized by
its value above the front. Numerical solution courtesy of S.J. Cox.
The exponents predicted by these two limiting cases are found experimentally,
depending on the Boussinesq number [69].
There is no strict equality between ufront and ūz because the front velocity is
generally measured within the laboratory frame of reference, which does not take into
account the upward displacement of bubbles due to the expansion of the foam (the
sign of this correction depends on the experimental conditions). However, for the liquid
fractions considered here (φl ≤ 0.1), this effect can be disregarded. Beyond a certain
value of imposed flow rate, and therefore above a certain liquid fraction, we observe
a phenomenon known as a convective instability [37], in which the bubbles themselves
move around. It occurs because the yield stress of a foam (see §3.2, chap. 4) depends on
both the liquid fraction and bubble size. The liquid fraction above which the instability
is triggered is inversely proportional to the bubble size; it is about 0.05 for bubbles of
several millimetres in diameter and 0.15 for submillimetric bubbles.
Foam permeability The liquid fraction above the front can be determined from
φl = Q/(S ufront ), where Q is the imposed liquid flow rate and S the cross-sectional
area of the container. Then eq. (3.91) gives information about the foam permeability α.
Fig. 3.37 shows the foam permeability for two surfactant solutions with differ-
ent interfacial rheological behaviour. The figure compares theoretical predictions with
experimental data from forced drainage experiments. The vertex-dominated case cor-
rectly describes the permeability of dry foams with mobile interfaces and the Plateau
border-dominated case is a good approximation, and the network model even better,
for foams with less mobile interfaces (note that a TTAB-dodecanol solution doesn’t
give perfectly immobile interfaces).2 Nevertheless, these two limiting cases are only
valid in the case of dry foams (φl 0.05). The network model (eq. (3.81)) is more
appropriate for predicting the permeability α over a larger range of liquid fraction.
Fig. 3.37 shows that this model reproduces the variation of α with φl and takes ac-
count of the difference in permeability between the two solutions. The dependence of
α on bubble size is less well understood: fig. 3.37 suggests that α ∼ d2 , but this is
not always the case. Eq. (3.81) predicts the dependence on d2 (via RV ), but this is
modified by the dependence of Kp and R̃v on d via Bo (eq. (3.69)). At the present
time, there isn’t a more satisfactory model, and the network model remains the one
which most accurately describes the drainage of liquid foams.
4.7.2 Free drainage We now return to free drainage, which, in practice, is the
more usual case. Consider a foam sample of height Z and cross-sectional area S, with
uniform initial liquid fraction φl0 at t = 0, and constant and uniform bubble size RV .
We seek the vertical profile of liquid fraction in the foam and its temporal evolution
during free drainage. The boundary conditions are that the flow rate is zero at the top
of the foam and that the liquid fraction at the bottom of the foam, in contact with the
liquid reservoir, is φl = φ∗l ≈ 0.36. The drainage curves can be calculated numerically,
as in fig. 3.38 which shows the result in the Plateau border-dominated limit.
2 Other surfactants, such as proteins, further reduce the differences between the data and the
Plateau border-dominated model for immobile interfaces.
Drainage 131
(a) (b) d = 1 mm
d = 1.1 mm d = 2 mm
d = 4.3 mm
Carman-Kozeny model
10–4 Vertex-dominated 10–5
Network model d = 2 mm
Network model
d = 1.1 mm Network model d = 1 mm
Network model
d = 4.3 mm Plateau-border dominated
10–5 10–6
α/d 2
α/d 2
10–6 10–7
10–7 10–8
0.001 0.01 0.1 0.001 0.01 0.1
Liquid fraction ϕ1 Liquid fraction ϕ1
Fig. 3.37: Permeability of foams with two different mobilities, determined from forced
drainage experiments. (a) TTAB foams (at 3 g.L−1 ), which have fairly mobile inter-
faces, for two bubble sizes. The line for the vertex-dominated case is eq. (3.82) with
R̃v = 2,000. The network model is eq. (3.81) with R̃v = 400, ηs = 5 × 10−8 kg.s−1 ,
and the function ξ(φl ) = aφ2l + bφl + c, given in §4.5.4. (b) Foams made from TTAB
(3 g.L−1 ) and dodecanol (0.2 g.L−1 ), giving less mobile interfaces, for two bubble
sizes. The line for the Plateau border-dominated case is eq. (3.74) and the line for the
Carman–Kozeny model is eq. (3.83). The network model is eq. (3.81) with R̃v = 800,
ηs = 5 × 10−7 kg.s−1 , and the function ξ(φl ) = aφ2l + bφl + c, given in §4.5.4. Note the
difference in the scale on the vertical axes. Data from [48].
A qualitative comparison with fig. 3.29 suggests that the foam drainage equation
predicts the experimental results well. From the numerical results we can again define
a characteristic time τ separating the first rapid phase, in which the liquid fraction is
only modified in the upper part of the foam, from a second slower phase of relaxation
towards the equilibrium profile.
The numerical results also give the shape of the liquid fraction profiles more pre-
cisely. For t < τ , the liquid fraction varies almost linearly with z, from the top of the
foam to the highest point that is still at the initial liquid fraction:
φl
= Ap (t)(Z − z), (3.93)
φl0
where Ap (t) is the observed gradient. The volume of drained liquid, V (t), is also linear
in time at first, and then tends asymptotically towards the value corresponding to the
total volume of liquid initially present in the foam: Vl0 = Sφl0 Z. The end of the linear
t=0
(a) (b)
Z
me
V(t)/Vl0
0.5
Ti
z
t=
τ
0 τ Time
0 1
Normalised liquid
fraction ϕ1/0.36
Fig. 3.38: Prediction of the Plateau border-dominated drainage equation (eq. (3.89)),
showing similar characteristics to experiments (fig. 3.29). (a) Liquid fraction profiles
at successive times. At time t = τ the drainage front has reached the base of the
column. The equilibrium profile, eq. (2.48), is visible in the bottom left-hand corner.
(b) The volume of liquid V that has drained out of the foam, normalized by the total
volume of liquid in the foam at t = 0, Vl0 . The variation is linear until t = τ , the
instant at which V = Vl0 /2. Numerical solution courtesy of S.J. Cox.
regime, at time t = τ , offers a natural time-scale for characterizing foam drainage. At

this time, Ap (τ ) = 1/Z and the volume of drained liquid is

V (τ ) Sφl0 Z φl 1
= 1− dz = . (3.94)
Vl0 Vl0 0 φl0 2
We determine the characteristic time τ by considering the lowest part of the foam.
For t < τ , the liquid fraction here is equal to the initial liquid fraction, φl0 , so that
the flow of liquid into this region is equal to the flow leaving it; thus the drainage rate
here is constant and characterized by the front velocity: ufront = V (τ )/Sφl0 τ = Z/2τ .
Using eq. (3.91) or eq. (3.92), the value of τ is therefore
φl0 Z η ηZ
τ= = . (3.95)
2 αρl g 2Cp ρl gRV2 φl0
The parameter Cp = 0.16 Kp (eq. (3.73)) depends on Bo (fig. 3.34). In the vertex-
dominated limit, a similar calculation shows that the liquid fraction profiles are
parabolic; eqs. (3.93) and (3.95) become
1/2
φl /φl0 = Av (t)(Z − z)2 τ = 2ηZ/(3Cv ρl gRV2 φl0 ), (3.96)
where Cv is given by eq. (3.82)). Here τ corresponds to the instant at which 2/3 of
the liquid has drained out of the foam (see exercise 8.5).
Drainage 133
Numerical solution of the drainage equations also allows a more detailed compar-
ison with experiments. For example, it confirms the experimental observation that
for t τ , the volume of drained liquid at time t differs from the volume of drained
liquid in the final, equilibrium, profile (eq. (2.48)) by an amount that varies in time
as t−1 (Plateau border-dominated) or t−2 (vertex-dominated), provided the system is
sufficiently far from equilibrium. In that limit, when t τ , the approach to the equi-
librium profile is very slow and is characterized by exponential dynamics, exp(−t/τ ).
Finally, note that the equilibrium profile, eq. (2.48), is a solution of the drainage equa-
tion (eq. (3.89)), shown in figs. 2.23 and 3.38a, and that there is a similar solution in
the vertex-dominated case in the long-time limit.
4.8 Summary and remarks

The Plateau border-dominated and vertex-dominated models introduced in §4.5 can
be viewed as limiting cases of a network model (§4.5.4). The transition from one
limiting case to the other is accomplished by changing the interfacial mobility (for
example via the interfacial viscosity) via the Boussinesq number: Bo 1 corres-
ponds to a foam with “immobile interfaces”, in which the drainage can be described
by the Plateau border-dominated case, while Bo 1 corresponds to a foam with
“mobile interfaces”, in which drainage can be described by the vertex-dominated
case. Real foams are rarely at one extreme and we must consider both contributions
(fig. 3.37).
We finish this section with some remarks on these results and on the limit of validity
of these drainage models. The theory is based on numerous hypotheses, involving
relatively few parameters. In particular, it should only apply to dry foams, in which
the bubble size is constant, and which are made from a simple chemical solution.
The agreement between prediction and experiment is satisfactory bearing in mind the
difficulty of the problem.
Nevertheless, at the present time, there remains a great deal to do in order to
describe drainage from every angle. Firstly, as we saw in §4.7, existing models are
not entirely satisfactory: either they depend on hypotheses which are too restrictive
or they do not faithfully reproduce certain experimental aspects. Current efforts fo-
cus on a more finely tuned model of the coupling between Plateau borders, vertices,
and films [20, 60, 62, 73]. A precise measurement of surface viscosity would also be
advantageous.
Secondly, research into very wet foams, which are often crucial in industrial
processes, is currently lacking [68]. Moreover, in many industrial applications the in-
terstitial liquid in a foam is often a non-Newtonian fluid. There are no models that
can describe the drainage of such foams, in which the liquid phase has a complex
rheological behaviour.
Thirdly, bubble size, d, is one of the parameters which most influences drainage
kinetics (in addition to the liquid fraction). Coarsening usually occurs at the same
time as drainage, and therefore modifies the drainage kinetics. This coupling between
coarsening and drainage [67, 77] is significant (fig. 3.39), but is absent from existing
models, which assume constant d.
0.8 0.25
0.04 0.25
Dimensionless volume of
0.6 N2
drained liquid
0.04
0.4
C2F6
0.2
0
0 1,000 2,000 3,000 4,000 5,000
Time
Fig. 3.39: Free drainage experiment for foams in which the bubbles contain nitrogen
and C2 F6 , for two values of initial liquid fraction (φl = 0.04 and 0.25). The foam
height is 30 cm, and the initial size of bubbles is d = 0.15 mm. The C2 F6 , which is not
very soluble in the foaming liquid, significantly slows down the ageing of the foam (see
fig. 3.27). The volume of drained liquid increases much more quickly in the case of ni-
trogen. This illustrates the strong coupling which exists between drainage and coarsen-
ing, the latter being responsible for the increase in average bubble size. After [66].
In extreme cases, the coarsening may completely control the drainage, and the
dependence on the initial liquid fraction disappears. An initial high liquid fraction
gives rise to rapid drainage, but slow coarsening (§3.3.3). Conversely, a dry foam
drains slowly, but coarsens more quickly: this is then accompanied by an acceleration
of drainage, as the bubble size increases, which can compensate for the reduction of
the initial liquid fraction. This has been termed “self-limiting” drainage: increasing
the initial liquid fraction doesn’t increase the drainage rate (observe the behaviour of
nitrogen-based foams in fig. 3.39).
Finally, we note that there are other experimental parameters which have an influ-
ence on the drainage dynamics, such as the shape of the container or the uniformity
of the foam [5, 66].
5 Rupture and coalescence

We have seen that a foam ages over time under the effects of drainage and coarsening.
The rupture of a film between two neighbouring bubbles is another process which
tends to make a foam disappear. In this section we will discuss the mechanisms by
which a single film can break, the collective effects at the scale of the foam, and the
stimulated rupture of a film by antifoaming agents.
5.1 Rupture at the scale of a single film

We saw in §2 that, in order to create a foam, significant repulsive forces must be
present in the films so as to induce a positive disjoining pressure to counterbalance
Rupture and coalescence 135
capillary suction. Without this positive disjoining pressure, the surfaces of a film do
not repel each other, its thickness decreases to zero, and, since there is nothing to
prevent coalescence, it breaks. This is clearly the simplest cause of film rupture.
It is also possible that repulsive interactions exist, but that they are weak (in other
words, Πd is positive for large thicknesses, but the maximum is small, see fig. 3.20).
As we saw in §2.3.2, capillary pressure Pc must be equal to Πd for a film to exist. This
allows the existence of foams in which the capillary suction from the Plateau borders
is weak, such as in the wet limit and for bubbles of large diameter. However, over
time, gravity-driven drainage dries the foam, reducing the cross-sectional area of the
Plateau borders and thereby increasing the suction applied to the films. The pressure
equilibrium condition (Πd = Pc ) becomes untenable, the repulsive forces in the films
are insufficient to counterbalance the increased Pc , and the films thin and then break.
The description of stability presented in §2.3.2 applies to a quasi-static situation.
Even in the case where there are strong repulsive interactions and a large disjoining
pressure, there are dynamic effects (for example during film formation) or fluctuations
which can cause rupture.
5.1.1 Film thinning and dynamic effects When two bubbles come into con-
tact, a thin film forms and its area increases while its thickness decreases towards
an equilibrium value h. During this dynamic process, there is a possibility of hydro-
dynamic instabilities causing film rupture at thicknesses greater than those at which
the disjoining pressure intervenes.
The simplest models assume that the interfaces are flat, parallel, horizontal, and
immobile (with no flow in the plane of the interface). This leads to the following
expression for the rate of thinning (vRe = −dh/dt), known as the Reynolds velocity, of
a circular, horizontal film of radius R:
2h3 (Pc − Πd )
vRe = , (3.97)
3ηR2
where Pc is the capillary pressure and Πd the disjoining pressure.

Experimental results show a significantly greater thinning velocity [40], which
means that these simple hypotheses are unrealistic. Increasing the interfacial mobility
and adding a coupling between the flow at the surface and within the film (as in the
Plateau borders, cf. §4) partially improves the agreement.
The film between two bubbles thins over time due to gravity, but more importantly
due to capillary suction into the Plateau borders. Experimentally, it is easy to observe
that there are complex flows in a film. A film exchanges liquid with the adjacent
Plateau borders, which accelerates film thinning, and the surface of the film becomes
distorted, with regions of different thickness. This leads to pinching at the periphery,
accompanied by an excess of liquid at the centre of the film known as a dimple,
illustrated in fig. 3.40. With immobile interfaces (high interfacial viscoelasticity), the
dimple thins symmetrically and remains at the centre of the film. On the other hand,
if the interfaces have high mobility (low viscoelasticity), the dimple may suddenly
move to one part of the periphery [71] (fig. 3.40) and then disappear by emptying
(a)
(b)
Fig. 3.40: Asymmetric drainage of a dimple in a horizontal film with high interfacial
mobility, studied with the film balance technique (§1.1.4, chap. 5). (a) A dimple is
stuck in the film (on the left), then empties into the peripheral Plateau border (on
the right). This occurs very rapidly (<1 s) and induces significant flows in the film.
(b) The dimple on the left is rapidly absorbed into the peripheral Plateau border,
resulting in the large-scale flow pattern seen on the right. Photographs courtesy of
A. Saint-Jalmes. See Plate 9.
into a Plateau border. This response provides a way to estimate the magnitude of the
interfacial viscoelasticity or mobility.
This exchange between a film and a Plateau border is also seen when a vertical film
drains, where it is known as marginal regeneration. It occurs when the interfaces are
very mobile, allowing rapid exchange between thick and thin regions (regeneration)
close to the boundary of the film (the marginal zone) [58] (fig. 3.41).
Even if all conditions are chosen so that an equilibrium thickness can be found,
the presence of dimples and the process of marginal regeneration increase the drainage
rate in the film and can even cause films to rupture.
5.1.2 The effect of fluctuations We reconsider the fluctuations which can occur
in a liquid film, that were introduced in §2.2.4 in relation to the interfacial viscoelasti-
city. Here again, even though all conditions necessary for stability are fulfilled, a film
may break because of natural (especially thermal) fluctuations in thickness or surfact-
ant concentration, which can either trigger a transition to a Newton black film (more
fragile), or cause rupture directly.
One of the main causes of film rupture is due to a non-uniform distribution of
surfactants at the surface and the creation of gaps large enough to destabilize the
film. A decrease in surfactant concentration is accompanied by an increase in surface
tension, and by a decrease in Πd , causing a local decrease in film thickness. Dilatational
elasticity stabilizes a film against these fluctuations by causing, through the Marangoni
effect, surfactant and fluid flow within the film (§2.2.4): the bulk flow induced by
(a) (b)
Turbulent
Black film regions
"Thinner"
region
Plateau border
Coloured
film
"Thicker"
region
Fig. 3.41: Marginal regeneration in a film held on a vertical frame. (a) General situ-
ation: the film drains from the top, creating a region of very thin black film; lower down
the flow is perturbed close to the frame. (b) Detailed situation, at the scale of the
Plateau border attached to the frame: regions of thick film rejoin the Plateau border,
whilst thinner areas detach and rapidly rise to the black film region. This exchange
between the thick and thin regions increases the film drainage rate. After [2].
the surface concentration gradient eventually reduces the fluctuations in interfacial

concentration (and the associated fluctuations in thickness).
Rupture of an initially stable film may also occur when the interface curves and
the thickness fluctuate locally [80], as sketched in fig. 3.42. These oscillations may be
the result of external (thermal or mechanical) perturbations, for example due to flow
or the bursting of neighbouring films. Curvature of an interface sometimes generates
hydrodynamic instabilities in the form of amplified capillary waves. In addition, al-
though the surface tension tends to smooth the interfaces, short-range attractive forces
between interfaces (§2.3.1) can dominate and increase the curvature and pinching of
the film. For example, if the local thickness of a film is less than that of a common
black film, it becomes locally unstable (§2.3.2). Film rupture is thus observed at the
point where the distance between the two interfaces is smallest: due to the pinching
of the film a hole is formed.
Analysing the stability of a film of average thickness h0 is possible if we assume
an initially sinusoidal fluctuation in the wavelength λ0 (see fig. 3.42) and follow its
λ0
h0 h(x)
Fig. 3.42: Oscillations of the interfaces of a film generated by mechanical or thermal

perturbations sometimes cause film rupture.
temporal evolution. Ivanov [40] and Vrij et al. [80] have shown that if the interaction
between the interfaces is dominated by van der Waals forces, then all wavelengths
greater than λc ∼ (4π 3 γh40 /Ah )1/2 are unstable, where Ah is the Hamaker constant
(see eq. (3.20)). Only films of radius R less than this critical length are stable, which
gives the film thickness hc at which rupture occurs spontaneously in terms of the size
1/4
of a film: hc = Ah R2 /(4π 3 γ) . This expression is based on the hypotheses that the
film is at equilibrium and that the average film thickness is constant in time.
5.1.3 Opening of a hole When a film breaks, a hole appears and grows. Can we
understand the dynamics of this growth? As soon as a hole is formed, surface tension
tends to enlarge it. For a Newtonian fluid, the displaced liquid accumulates at the rim,
which takes a toroidal form (fig. 3.43). In a cylindrical coordinate system (r, θ), with
origin at the centre of a hole of radius a, Newton’s second law applied to the part of
the rim lying between θ and θ + dθ, with mass dM and velocity wer , resolved parallel
to er , gives
d(dM w)
= 2γadθ. (3.98)
dt
In fact, the film is at rest before it is drawn into the rim and the only force acting at
the rim is the surface tension of the film. Moreover, it is reasonable to assume that
the liquid is entirely collected in the rim, which gives dM = dθ ρl a2 h/2. Finally, by
assuming that the velocity w = da/dt is constant, Culick [11] showed that

2γ
w= . (3.99)
ρl h
This predicts that in a film one micron thick the velocity at which a hole grows is
of the order of 10 m.s−1 . A circular film with a radius of 1 mm therefore disappears in
10−4 s! Assuming equilibrium between capillary and kinetic energies (inertial regime)
leads to an expression for the velocity that has the same dependence on the physical
parameters (ρ, γ, and h) but with a different numerical coefficient.
The experiments of McEntee and Mysels [53] validate Culick’s prediction
(eq. (3.99)) for films with thickness between 0.1 and 10 μm. Note that energy is not
conserved in this model: the lost capillary energy is twice the gain in kinetic energy.
The failure of the model to conserve energy is attributed to the presence of an inelastic
shock between the rim, which is displaced at velocity w, and the liquid contained in
the film at rest; this is a visco-inertial regime. This macroscopic argument is somewhat
a h
Fig. 3.43: The opening of a hole in a liquid film, represented here as a cross-section
through the film. For a Newtonian fluid the liquid gathers at the rim in a toroidal
shape.
limited: it doesn’t take into account either the surfactants, the precise shape of the
liquid at the rim, or the stability of the film.
If a film is very thin and/or the liquid is extremely viscous, then inertia is negligible.
Equilibrating viscous forces (dissipation) with capillary forces gives
2γ
w= . (3.100)
η
By comparing this expression for the velocity with the other limit (eq. (3.99)), we
find that the critical thickness hc of a film separating this viscous regime from the
visco-inertial one, corresponding to a Reynolds number of order one, is
η2
hc = . (3.101)
2γρl
For a typical foaming solution this thickness is of the order of tens of nanometres,
close to a common black film.
For films formed from non-Newtonian fluids (for example long-chain polymers
without surfactants), the rim does not thicken and the radius of the hole grows
exponentially in time [14].
Note also that it is not only holes that form in films, but also thin regions (§2.3.1).
They occur during drainage, following a jump in the local film thickness, and their
area increases in time, as shown in fig. 3.18.
Finally, the liquid rim may be unstable, and at certain velocities and thicknesses
may form a string of drops (fig. 3.44), illustrating another dynamic effect which may
compromise film stability.
Fig. 3.44: The thickness of a stratified film (fig. 3.18) decreases discontinuously.
When a thin region develops and grows, the liquid rim may become unstable, form-
ing several large droplets. The film shown consists of a mixture of surfactants and
polyelectrolytes (see §6.1), studied with the film balance technique (§1.1.4, chap. 5); its
diameter is 0.2 mm and its thickness is 20 nm. Photograph courtesy of A.Saint-Jalmes.
See Plate 10.
5.2 Rupture at the scale of a foam

The link between film rupture and the collapse of a foam is obvious: a foam disap-
pears via a succession of film ruptures. The bubble films at the surface of a foam
are most susceptible to rupture: they are the thinnest (due to drainage), the most
curved (bubbles with few neighbours are closer to a sphere), they evaporate the most
quickly (because they are in contact with ambient air), they are subject to external
perturbations (notably dust particles which, at the scale of a nanometric film, re-
semble enormous meteorites), and, lastly, they are the first to encounter environmental
variations (temperature for example).
Once the bubbles at the top have burst, the bubbles beneath come into contact with
the atmosphere and become in turn more susceptible to bursting. Thus a foam often
collapses from the top. However, if the walls of the container are dirty, for example
if particles or oil are present, a foam may start to rupture here. In addition, a film
within a foam is also sensitive to mechanical perturbations caused by its environment:
the liquid flow caused by drainage, ejection of droplets by the destabilization of the
rim of a hole in a nearby film, even sound waves emitted during film rupture (the
intermittent “crackling” of a dry foam that you hear in your bath). For all these
reasons, the probability of rupture of an isolated film is not the same as that of a film
in a foam, which seems to depend on its neighbours.
Observation of two-dimensional foams and recording the sound made by a rupture
event at the surface of a 3D foam allow the identification of individual rupture events.
They tend to occur in avalanches, that is, each rupture is not independent of the pre-
vious event; instead, rupture occurs in correlated bursts [55, 76]. In fact, if these events
were independent, they would follow Poisson’s law.3 Fig. 3.45 shows the distribution
of the number of events n in a dry foam, indicating that, at least in this case, the
experimental results are poorly described by Poisson’s law.
The mechanisms by which a foam collapses are still poorly understood, for example
the correlation between stability of an isolated film and the stability of the same film
within a foam [54], or even the origin of the avalanches of rupture events. Similarly,
it is not yet clearly understood how foam collapse depends on the bubble size or
on interfacial properties. The best approach has yet to be established: is the most
significant parameter a critical film thickness, a critical disjoining pressure, or a critical
liquid fraction?
5.3 Defoamers and antifoams

There are many situations where foam is considered a nuisance or a drawback. A
number of industries are directly affected by the problem of a too-stable foam. Foam
is sometimes a necessary part of a manufacturing process: its presence is therefore
temporarily desirable, but it is necessary to destroy it when it has served its purpose.
3 Recall that in this probability distribution, if n is the number of events occurring in a period of
time and n denotes the average number of events in this period, the probability p(n) of finding n
n
events is p(n) = e−n n
n!
.
103
Probability (arbitrary units)

102
101
100
0 5 10 15 20 25 30
Number of events per time interval
Fig. 3.45: Temporal distribution of rupture events in a dry 3D foam at different ages.
The distributions were recorded acoustically over time intervals of equal duration, and
have been normalized so that their maxima are equal. The white circles represent the
best possible fit of the data to a Poisson distribution [55].
We present here some of the strategies put in place either to prevent the formation
of a foam, or to destroy an existing, stable foam [31]. There is a subtle difference
between the two: in the first case the antifoaming agent must be present in the solution,
while in the second, the defoamer must be dispersed in the foam. Nonetheless, the
way in which they work, and the products used, turn out to be very similar. These
products are grouped into three categories: liquids (generally in the form of droplets),
solids (especially ones that are finely ground), and liquid/solid mixtures (the most
efficient).
5.3.1 Liquid antifoams A liquid antifoam consists of an immiscible liquid (e.g.

oil in aqueous solution) that is added to the foaming solution in the form of droplets.
In order to be effective, each droplet must reach the solution/gas interface, represented
schematically in fig. 3.46. As a droplet emerges from the solution, an antifoam/gas
interface (with surface energy γga ) appears, and a solution/antifoam interface and a
solution/gas interface (with surface energies γsa and γsg , respectively) disappear.
The difference in surface energy (per unit area) between the final state and the
initial state is ΔU = γga − γsa − γsg . For emergence of the droplet at the solution/gas
Gas
Solution
Antifoam
Fig. 3.46: The emergence of an antifoam droplet at the solution/gas interface. An

antifoam/gas interface appears, and a solution/antifoam interface and a solution/gas
interface both disappear.
interface to be energetically favourable requires that ΔU < 0. The entry coefficient is

generally defined as
Eγ = −γga + γsa + γsg , (3.102)
and the condition for emergence is Eγ > 0. To estimate the sign of Eγ requires the
values of surface energy for the three fluids, obtained when they are mutually saturated
and the surfactants have had time to adsorb. If the antifoam droplet does emerge at
the interface it may be energetically favourable for it to spread out. This is determined
by the spreading coefficient, defined as the difference in surface energy (per unit area)
between the non-spread and spread configurations, Sγ = γsg − γsa − γga , estimated in
the same way. The drop spreads out if Sγ > 0. We consider two cases, both of which
can cause a film separating two bubbles to rupture:
(a) Case Eγ > 0 and Sγ > 0 In this case an antifoam droplet emerges and spreads
across the solution/gas interface. There are at least two reasons for this to contribute
to the destabilization of the liquid film.
On the one hand, the solution/gas interface is replaced by the antifoam/gas inter-
face, which does not behave a priori like a surfactant. Consequently, this new interface
does not contribute to the stabilization of the film, so it is less stable than before.
On the other hand, the film is destabilized by the spreading out of the antifoam at
the solution/gas interface, in a process called Marangoni spreading (fig. 3.47). The
spreading of the antifoam droplet causes bulk flow in the film, by viscous drag of the
neighbouring liquid layers, leading to its thinning then rupture [27]. The driving force
for the spreading out of a lens-shaped droplet on a surface is the imbalance in the
surface tensions along the triple line. The net force per unit length (energy per unit
area) is none other than the spreading coefficient Sγ . For example, by applying con-
servation of volume to a silicone oil droplet of initial radius a that spreads up to some
maximum size, and assuming that a thickness δm of liquid is entrained, it has been
shown that [64]
1/3
ηa2
δm = a , (3.103)
Sγ ρa dm
where ηa and ρa are the viscosity and the density of the antifoam, and dm is the
thickness of the antifoam lens when it is fully spread (typically one molecule thick).
The equation predicts that the antifoam is more efficient the larger the droplet; this
is certainly true for small droplets, but in reality there is an upper limit (of the order
of a few microns [64]) to the optimal size of antifoam droplets.
(b) Case Eγ > 0 and Sγ < 0 This situation is very different from the previous one,
because the antifoam lens arrives at the solution/gas interface, but does not spread.
An antifoam effect is possible only if the lens can bridge the two interfaces of the film
(fig. 3.48).
If it can form, the stability of an antifoam bridge depends on the value of the
equilibrium contact angle θ between the antifoam and the solution. If it is greater
Fig. 3.47: Emergence and spreading of an antifoam droplet at the surface of a film
causing it to thin and the liquid to move. The film becomes so thin that it breaks,
releasing the antifoam droplets.
than π/2 then the curvature of the interface close to the droplet imposes a greater
pressure than is found in more distant parts of the film. This pressure gradient causes
liquid to move and the film thins until the two triple lines on each side of the lens
meet. Then rupture occurs and the antifoam is dispersed into several smaller droplets.
This reduction in the average diameter of antifoam droplets has been observed during
several cycles of use, and is correlated with a loss in efficiency of the antifoam, since if
the droplet becomes smaller than a critical value then bridging no longer occurs and
the antifoam no longer works.
5.3.2 Solid antifoams Small (micron-sized) solid particles with hydrophobic sur-
faces (fig. 3.50, right), for example carbon or pyrogenic silica, naturally play a role
in antifoaming, as follows. The way in which a film is ruptured by a solid particle is
similar to the situation Eγ > 0 and Sγ < 0 described above. The particle must emerge
Gas
θ Anti- Film
foam
Fig. 3.48: Bridging and rupture of a soap film by an antifoam droplet.
from the film, then bridge the liquid film in such a way as to cause it to thin. The li-
quid/solid contact angle is the parameter which determines whether or not the particle
can adopt a mechanically stable equilibrium position at the interface. For particles of
this size, capillary forces fc determine the equilibrium position, and a spherical particle
assumes its equilibrium position when fc = 0. This force is equal to the surface tension
of the liquid/gas interface integrated along the triple line forming the liquid/gas/solid
interfaces (fig. 3.49):
fc = −πγa cos ψ cos(ψ + θ)ez , (3.104)
where a is the particle diameter.

Equilibrium occurs when ψ = π/2 − θ. When the sphere bridges the two interfaces
of a film, the liquid must move so that the interfaces may remain flat. As in the case of
an antifoam droplet described above, it is the pressure generated by the curvature of
the interface which gives rise to film thinning.
If the contact angle θ is less than π/2, the interfaces reach equilibrium before the
film thickness decreases to zero (ψ > 0). On the other hand, if θ is greater than π/2,
the equilibrium position of the interface corresponds to an angle of ψ < 0. This is
incompatible with the presence of the second interface, and the film breaks when the
two triple lines meet.
a/2 ψ θ
Fig. 3.49: Solid sphere at a liquid/gas interface.

Appendices 145
θ θ
Fig. 3.50: Behaviour of a solid particle in a soap film as a function of the solid/liquid
contact angle. θ < π/2 (left): the interfaces remain attached to “their” half of the
sphere, and thinning of the film is prevented by the appearance of negative curvature
around the sphere. θ > π/2 (right): the interfaces are unable to simultaneously find a
stable position on the sphere, causing film rupture.
Solid particles thus have an antifoam effect if the solid/liquid contact angle is
greater than π/2, i.e. if they are hydrophobic (fig. 3.50, right). The process of rupture
is even more effective if the particles are large, because they emerge more rapidly.
5.3.3 Super defoamer It is generally much more efficient to use mixed antifoam-
ing agents, containing both oil and hydrophobic solid particles, rather than one or the
other. One of the most plausible explanations for this is related to the stability of the
pinched film between the antifoam droplet and the gas phase. In fact, the first step in
all of the mechanisms for rupture presented here is the emergence of the antifoam at
the solution/gas interface. During this motion the liquid trapped between the two in-
terfaces is expelled, a flow which occurs more slowly as the amount of the liquid in the
film that is entrained becomes smaller. Rupture can only happen when the thickness
of this region is equal to some critical distance (of the order of a hundred nanometres).
If there are solid particles on the surface of the antifoam droplet, their size fixes the
thickness at which the film separating the antifoam/gas phases breaks. This scenario
has been tested experimentally. The results have shown that films trapped between
a droplet and the gas phase exist for less time in the presence of solid particles, and
that this reduction can be correlated to the “roughness” of the interfaces.
6 Appendices
6.1 Stabilizing agents
Foams are stabilized by amphiphilic molecules, or surfactants (see p. 18), which are
diverse in nature: detergent molecules, proteins, polymers, and even solid particles. In
this section, we describe several large families of molecules that are found in the most
common detergents, cosmetics, or other products of daily life, classified as a function
of their molecular mass.
6.1.1 Surfactant molecules of low molecular mass Small surfactants with

a molecular mass Mw < 1,000 g.mol−1 move rapidly in solution through Brownian
motion; they adsorb (and desorb) more quickly on an interface than surfactants of
high molecular mass. Most of the surfactants used in industry fall into this category.
Chemical nature Depending on the charge of the polar head, we refer to surfactants
as anionic (negative charge), cationic (positive charge), non-ionic (no net charge), or
amphoteric (positive or negative charge depending on the pH of the solution). We list
below some of the surfactant families of low molecular mass which are currently used.
This list illustrates in particular the large chemical differences which exist between sur-
factants, and the variety of compounds used. Anionic surfactants are frequently found
in detergent products, washing-up liquids, etc., while non-ionic surfactants are present
in shampoos and other personal cleaning products. Cationic surfactants are often
chosen for their corrosive and bactericidal effects. Nevertheless, commercial products
generally have complex formulations, combining several surfactants (sometimes with
opposite charges), charged polymers, viscosifiers, etc. (see p. 148) with the aim of
meeting several objectives simultaneously.
Anionic surfactants Alkyl sulphates (like sodium dodecyl sulphate, or SDS, with the
formula C12 H25 OSO− +
3 , Na ), alkyl ether sulphates and alkyl benzene sulphates are
the standard ones. There are also alpha-olefin sulfonates and sulfoccinates (like sodium
bis(2-ethylhexyl) sulfosuccinate, or AOT: (C8 H17 −O−COCH2 )2 SO− +
3 , Na ).
Cationic surfactants These include alkyl trimethyl ammonium bromides (Cn TAB:
−
Cn H2n+1 N(CH3 )+3 , Br ) and alkyl trimethyl ammonium chlorides (like DTAC:
− −
C12 H25 N(CH3 ) , Cl ) and cetylpyridinium chloride (CPCl: C16 H33 N(CH2 )+
+
5 , Cl ).
Non-ionic surfactants These include polyoxyethylene alkyl ethers (Cn EOm ), alkyl
glucosides Ci Gj , spans and tweens (the latter two are also used for emulsions), Triton
X-100 (C14 H22 O(C2 H4 O)n ), and polyglycerol ethers.
Amphoteric surfactants These are sodium carboxylates, RCO2 Na (like natural soaps,
in which R represents a carbon chain), betaines (like cocoylamidopropyl betaine,
CAPB), and tetradecyl dimethylamine oxide (C14 H29 NO(CH3 )2 ).
Structure These surfactant molecules are distributed between the interfaces, where
they lower the surface energy, and the solution. At low concentrations they are isolated
when in the solution, but above a critical micelle concentration (cmc, see fig. 2.4b),
the surface concentration and the surface tension no longer change and the surfact-
ants then aggregate in the solution. In these aggregates the hydrophobic tails group
together and the polar heads form a shell around them, thus protecting them from
contact with the solution (fig. 3.9). As the surfactant concentration increases, dif-
ferent three-dimensional structures appear: (i) micelles are spheres, with a radius of
the order of the length dm of a molecule; (ii) cylindrical micelles are flexible filaments
of circular cross-section of radius dm ; (iii) bilayers are membranes of thickness 2dm ,
which can reach very large lateral dimensions compared to dm , and which sometimes
form multilayers. These structures, in particular cylindrical micelles which intermix
like polymers, modify the viscosity of the solution. The presence of salt also affects
Appendices 147
the structure of aggregates, and the addition of salt is a way to increase the number of
cylindrical micelles in a solution (of shampoo, for example) and therefore the viscosity.
At very high concentrations it is possible to obtain phases in which the surfactants
are arranged in crystalline structures.
The presence of micelles in a solution enables the dissolution of molecules that are
insoluble in the aqueous phase: the grouped lipid tails in a micelle behave like small
oil droplets in which some molecules can dissolve. Thus, certain molecules, called
co-surfactants, for example fatty alcohols such as dodecanol (C12 H25 OH), are not
surface-active by themselves (or only very weakly) and are also poorly soluble; but
when added to an ionic surfactant, for example, they move into the micelles and then
to the interfaces between surfactants. These co-surfactants improve foamability and
foam stability by significantly increasing the interfacial viscoelastic moduli, partly due
to their poor solubility (see §2.2.4).
6.1.2 Surfactant molecules of large molecular mass
Synthetic polymers Synthetic surfactant polymers are polymers that have been
designed or modified to become amphiphilic. Among the homopolymers (that show
the same repeated pattern), polyvinyl alcohol (PVA) and modified polysaccharides are
common. Copolymer blocks consist of hydrophilic polymers grafted onto hydrophobic
ones. Various assemblies of this type are known commercially as Pluronic. Diverse
configurations are possible at the interfaces, depending on the number of blocks, their
relative sizes, and the interfacial concentration. Diblocks, composed of one hydrophilic
block and one hydrophobic block, show a mushroom configuration of hydrophobic
chains at the surface at low densities, but above a certain density the chains extend
perpendicular to the surface, in a brush configuration. By introducing a hydrophobic
block between two hydrophilic blocks, it is possible to make a triblock which adsorbs
on the interface only by its hydrophobic part.
Natural polymers These are often edible proteins. Examples of globular proteins
include bovine serum albumin (BSA), β-lactoglobulin (BLG) from milk, or ovalbumin
from egg. Caseins (β and κ), which are also found in milk, are compact when in solution
but unfold at the interface.
These large surfactant molecules adsorb on an interface only very slowly (see
§2.2.2). This means that a highly concentrated solution is needed to make a foam.
Once in place, these molecules desorb with difficulty and can even remain at the
interfaces permanently. They also rearrange themselves at the interface, giving rise
to modifications of surface tension or viscoelastic moduli over time-scales of several
hours. In addition, proteins often form interfacial gels over long times, with large elastic
modulus.
Solid particles Aqueous dispersions of nano- or micrometric solid particles can

also be made to foam. If their wettability allows it, they adsorb at the interfaces
and stabilize them efficiently. It is mostly emulsions (Pickering emulsions) which are
produced in this way at the moment, but solid particles are being increasingly used to
stabilize foams [8].
6.1.3 Surfactant/polyelectrolyte complexes Many surfactant mixtures are

possible, and often necessary to optimize the properties of an industrial foam. Witness,
for example, the long list of ingredients in shampoo, shaving foam, washing powder,
and other household cleaning products. Moreover, in spite of their varied uses, these
products often have numerous ingredients in common!
These mixtures often act both as surfactants (of small size) and polymers (which
are usually charged and therefore known as polyelectrolytes). Fig. 3.51 shows the four
typical configurations that are found in the bulk and at the surface.
(a) In the case in which the components have the same charge, the surfactant generally
invades the surface first and prevents the adsorption of the polyelectrolyte by
repulsion. The interfacial properties are those of a pure surfactant, although the
liquid films can be stratified (see §2.3.1).
(b) In the case of oppositely charged components, both adsorb at the interface, often
through a synergistic co-adsorption: a surfactant/polyelectrolyte complex, more
surface-active than the components taken separately, first forms in the bulk and
then adsorbs to the surface. So even for very low concentrations of each component,
the surface tension is lowered significantly. The critical aggregation concentration
(cac) at which aggregates form in the bulk is much lower than the cmc of the
(a) (b)
Components with Components with
the same charge opposite charge
(c) (d)
Components which are
Components remaining
strongly hydrophobic
adsorbed at the surface
in the bulk
Fig. 3.51: Surfactant/polyelectrolyte mixtures in the bulk and at the surface. The
behaviour that arises from the various possible affinities that the molecules exhibit is
explained in the text.
Appendices 149
surfactant alone. Due to the mixing at the interfaces, the films obtained are thicker
than if they were stabilized by surfactants alone.
(c) In certain cases there are strong attractive interactions between the hydrophobic
parts of the polymers and the surfactants, known as hydrophobic interactions.
The surfactants adsorb onto the polymers, and therefore cover the surface less
thoroughly, leading to a reduction in foamability.
(d) Finally, if there is no interaction between the components, the interface is
partitioned between them.
6.1.4 Ferrofluids
Overview and physical principles In order to stabilize a foam, as we saw in §2,

it is sufficient to add surfactant molecules to the liquid phase. However, another way
of stabilizing a foam without adding surfactant is to create repulsive interactions in
the liquid phase.
One example is a Langmuir foam (fig. 2.24a), but it is also possible to produce
such foams with a ferrofluid, consisting of a colloidal suspension of magnetic particles
in a liquid carrier. The particles are 10 nanometres in diameter, and are made of
a ferrous or ferromagnetic material (generally maghemite or cobalt ferrite) with a
permanent magnetic moment. They consequently act as small nanoscopic magnets in
suspension. Under the effect of an external magnetic field of increasing intensity, the
magnetic moments gradually orientate themselves in the direction of the applied field,
whereas thermal agitation tends to disorientate them. As for a permanent magnet,
the interaction between particles depends on their relative position: they attract if
the particles are arranged head-to-tail, and repel if they are side-by-side. In order to
create a fluid which, when mixed with a second fluid, allows interfaces to form, the
interactions must on average be repulsive, so there must be more particles side-by-side
than head-to-tail.
Experimental system To create a 2D system, the ferrofluid is confined in a cell

consisting of two horizontal plates about 1 mm apart, and the magnetic field is oriented
vertically [25], a configuration in which more magnetic particles are side-by-side than
head-to-tail. In practice, two immiscible liquids are used: an aqueous ferrofluid and
a non-magnetic oil.4 The oil wets the plates better than the ferrofluid, so that a
microscopic film of oil separates the ferrofluid from the plates.
To create a foam, we start with the ferrofluid in a large drop in the middle of the cell
and zero applied field. The system is then subjected to an alternating magnetic field,
which creates holes in the drop. These holes fill with the oil to make a 2D magnetic
cellular structure, as shown in fig. 3.52. A foam created in this way exhibits many of
the characteristics associated with a 2D foam: it respects Plateau’s laws (§2.2, chap.
2), the Aboav–Weaire correlation (eq. (5.13)), and Euler’s formula (eq. (2.9), so on
average its bubbles have six sides) [25]. Moreover, since the oil can move between the
4 Such a mixture of two liquids is in fact an emulsion; here it is called a “ferrofluid foam” because
the pattern looks so foam-like.
Fig. 3.52: In this 2D ferrofluid foam the bubble walls are made of a ferrofluid (in
black) and the bubbles are filled with an immiscible, transparent, and non-magnetic
oil which wets the horizontal (upper and lower) plates. The foam is subjected to an
0 . Photograph courtesy of F. Elias.
external vertical magnetic field H
bubbles, by flowing through the wetting film, the bubble area can evolve in time, for
example due to pressure differences.
Equilibrium situation Unlike soap foams, which are metastable, a ferrofluid foam
can exist in equilibrium and therefore be truly stable. In fact, the van der Waals
interactions, which are attractive, tend to minimize the surface area of the oil/ferrofluid
interface, while the magnetic dipolar interactions, which are repulsive, tend to increase
it. So at equilibrium the total surface area of the interface has a value fixed by the
competition between these opposing forces. In contrast to a drop of a simple fluid, the
surface area is not minimal; instead, the creation of a pattern allows the interfacial
area to increase and reach equilibrium [3]. Depending on the way in which the system
is prepared, we may obtain labyrinthine patterns, a network of ferrofluid drops, or a
cellular structure.
In the latter case, which concerns us here, if the energy of the magnetic di-
polar interaction increases, so does the interfacial area and the number of bubbles
at equilibrium. Since the ferrofluid is paramagnetic, the magnetic interaction energy is
proportional to the square of the amplitude H0 of the magnetic field, so the number of
bubbles at equilibrium increases with H0 . Conversely, the number of bubbles decreases
if H0 is decreased, and the foam is completely destroyed when the field is removed.
The increase in the size of the bubbles when H0 is decreased recalls the coarsening of
a soap foam. The amplitude of the applied magnetic field is consequently the control
parameter for a ferrofluid foam, since it fixes all of its properties at equilibrium.
Out-of-equilibrium situation When H0 is increased in a given foam configuration,

a certain volume of ferrofluid is required at a vertex to allow a new bubble to appear
and a new equilibrium state to be found. However this volume of fluid is not always
available, especially when the foam is dry. In this case, equilibrium will not be reached
and the foam will evolve in time, with some bubbles growing and others shrinking
under the effect of the pressure differences between them.
Appendices 151
Fig. 3.53: 2D ferrofluid foam in an out-of-equilibrium situation. Observe that bubbles

with less than 6 sides grow in time while bubbles with more than 6 sides shrink, in
contrast to a coarsening 2D soap foam governed by von Neumann’s law. Photograph
courtesy of F. Elias [24].
However, unlike in a soap foam, a ferrofluid foam needs more interface to reach its
equilibrium state, and so the temporal evolution (“coarsening”) of cell areas follows
an inverted von Neumann’s law (§3.1.1): bubbles with less than 6 sides grow at the
expense of bubbles with more than 6 sides (and on average the area of a 6 sided
bubble doesn’t change) [24]. This process, shown in fig. 3.53, occurs because magnetic
dipolar repulsion leads to the existence of a macroscopic magnetic curvature energy,
which manifests itself as an additional pressure difference between the bubble and its
environment. This has the opposite sign to the radius of curvature, making a negative
contribution to the Laplace pressure difference.
Note that the interfaces of a soap foam also possess a curvature energy, due to
the electrical dipolar moments carried by the surfactant molecules, but the induced
pressure difference is negligible compared to the Laplace pressure. For ferrofluid foams
this curvature term dominates, and consequently reverses, the evolution of cell area
as a function of the number of sides. If the magnetic field increases further, the walls
between bubbles sometimes buckle, analogous to the response of a solid foam subjected
to compression [23].
One advantage of ferrofluid foams, besides the fact that they possess a control
parameter, is the possibility to locally manipulate them. It is possible to use a local
magnetic field gradient (for example, a magnetic needle pushed into the field H0 ) to
attract individual vertices, and so to artificially induce T1s [26] (see shaded section on
p. 80), or to induce anti-T2s, i.e. to create a new bubble (as in fig. 3.7 viewed from
right to left) by gathering sufficient ferrofluid at a vertex.
Finally, note that it is possible to use a magnetic field or force to manipulate a mag-
netic soap foam created from an aqueous ferrofluid containing surfactant molecules. It
is necessary to carefully determine the amount of surfactant to be introduced into the
ferrofluid to avoid destabilizing the colloidal suspension (i.e. to prevent the magnetic
particles aggregating and precipitating at the bottom of the solution) [15, 22].
6.2 Dissipation due to surfactant motion during the steady expansion

of a film
When a soap film of area S is stretched at a constant expansion rate ε̇ = d(ln S)/dt,
the surface concentration of surfactant decreases and concentration gradients appear.
Surfactants then move and repopulate the interfaces. In a thick film the bulk solution
can act as a reservoir of surfactant and this replenishment can occur in a direction
perpendicular to the interface (see fig. 3.12). However, if a film is very thin, surfactant
must migrate from the edges of the film (Plateau borders), which also induces dissip-
ation, the nature of which depends on the properties of the surfactant, and several
different modes of transport are possible, as illustrated in fig. 3.54.
Such surfactant-specific dissipation is in addition to the hydrodynamic dissipation
due to the velocity gradients in the bulk of the fluid, associated with the bulk viscosity
η (§3.3, chap. 4). Buzza, Lu, and Cates [6] predict the contribution of each mode
to the dissipated power D per unit volume of foam (in W.m−3 ). In every case, D
is proportional to ε̇2 (at least in a linear approximation), and the prefactor has the
dimensions of a viscosity. Calculating this effective viscosity, or equivalently D, thus
provides a convenient way to compare the contributions of different modes, even if we
estimate only orders of magnitude.
In the following we estimate for each mode the dissipation in the film per unit
surface area. Dividing it by the characteristic size of the bubbles, d (as in chap. 4,
§ 3.3.1 and 3.3.2), gives the dissipation per unit volume of the foam. Here d is also
the characteristic size (not thickness) of the film, and thus sets the scale of gradients
in surfactant concentration parallel to the plane of the film. In what follows, ηd , Γ,
and js are surface quantities, with dimensions of viscosity times length, concentration
times length, and flux times length, respectively.
6.2.1 Surface expansion The first contribution (fig. 3.54a) is due to the inter-
facial dilatational viscosity ηd associated with the expansion and compression of the
monolayer. This mode dominates when the surface moves together with the bulk solu-
tion, and the surface surfactant concentration Γ is not conserved: it is observed over
very short time-scales or for surfactants which are insoluble. In this case the dissipated
power is
ηd
Ddilat. visc. ∼ ε̇2 . (3.105)
d
6.2.2 Surface diffusion In the other limit, the surface concentration is kept con-
stant (at a value lower than its equilibrium value) by the flux of surfactants parallel to
the plane of the film. Film stretching then requires the transport of Ṡ/(Sd2m ) ∼ ε̇/d2m
surfactants per unit area and time, with d2m the area occupied by one molecule at the
surface.
We assume that there is no exchange with the bulk, and that surfactants mi-
grate along the interface with a speed limited by friction with the subsurface layer
(fig. 3.54b). The flux of surfactants in the surface is js = −Ds ∇Γ, where Ds is the sur-
face diffusion coefficient of the surfactants. Its 2D divergence is of order js /d ∼ ε̇/d2m .
The surfactant chemical potential is μ = μ0 + kB T ln Γ, where kB is Boltzmann’s con-
stant and T the temperature. The associated dissipation per unit area of film is of
order js .∇μ ∼ kB T (ΓDs )−1 j2s and so the dissipation per unit volume of foam is
kB T d 2
Dsurf. diff. ∼ ε̇ . (3.106)
ΓDs d4m
Appendices 153
d
(a)
Solution
h
Monolayer
Gas
(b)
(c)
Fig. 3.54: Stretching a film induces both a flow in the liquid (dashed arrows) and mo-
tion of surfactants to replenish the newly created interface. Dissipation can arise from
three contributions, shown schematically here on a cross-section through a film. (a)
Surface expansion: flow of surfactants within the interface (solid arrows). (b) Surface
diffusion: diffusion of surfactants within the interface (the most dissipative regions are
indicated by the thick solid lines). (c) Bulk diffusion: diffusion of surfactants within the
solution (solid arrows), followed by the less-dissipative migration of surfactant towards
and along the interface (open arrows).
6.2.3 Bulk diffusion in the plane of the film When surfactant movement to-
wards and along the interface is rapid enough, the dominant dissipation is due to bulk
diffusion in the plane of the film (fig. 3.54c). The average flux across a section of the
film is j = Dv ∇c, where Dv is the bulk diffusion coefficient and c is the bulk concentra-
tion. Typically j ∼ ε̇ d/(h d2m ) and the dissipation per unit volume of film is of order
j.∇μ, where μ is as above. Multiplying this dissipation by h/d yields the dissipation
per unit volume of foam:
kB T d 2
Dbulk diff. ∼ ε̇ . (3.107)
cDv hd4m
7 Experiments
7.1 Flow in a soap film

preparation time: 1 hr 30 experimental time: 10 min
Phenomenon demonstrated: film drainage.
References in the text: §2.3, §5.1.1 and §1.1.3, chap. 5.
Materials:
• a circular or rectangular frame, 5 to 20 cm in size, attached to a piece of string.
Ideally, the frame should be made from a rod with a screw thread, which creates
a liquid reservoir and means that the soap film lasts longer than if the frame is
smooth
• a bowl deep enough to contain the frame
• a large cardboard box at least twice the size of the bowl
• black paint or black paper
• a retort stand that will fit in the box
• a desk lamp (a source of white light),
• soap solution: 1% washing-up liquid in deionized water
• optional: glycerine.
1. Cut out one of the sides of the cardboard box. Paint the inside black a day in
advance, or cover the inside of the box with black paper. Place the box on a table
with its opening towards you. The inside of the box will be the stage for the
experiment, protected from air currents and external light.
2. Place the bowl containing the soap solution inside the box. Hang the frame from
the stand, just above the bowl, as close as possible to the back of the box. The plane
of the frame should be facing you, parallel to the opening of the box (fig. 3.55,
left).
3. Place the lamp outside the box, but directed towards the box, with an angle of
incidence of around 45 degrees (fig. 3.55, left).
4. Create a soap film by dipping the frame into the soapy solution, preferably by
moving the bowl upwards so as not to move the frame.
Experiments 155
45°
Fig. 3.55: Sketch of the liquid-flow experiment in a vertical soap film (left), and a
photograph of a film (right) in a frame of diameter 15 cm. Photograph courtesy of
F. Elias. See Plate 11.
5. Adjust the position of the lamp until you can see the reflection of the bulb on the
soap film (the angle of observation should be equal to the angle of incidence). It is
possible to place a diffusing screen (for example a sheet of white paper) between
the lamp and the frame to increase the illuminated area.
The soap film is at first white because its thickness, which is about ten microns, is
greater than the coherence length of white light. After several seconds horizontal bands
of colour appear at the top of the film and move downwards (fig. 3.55, right). These
iridescent bands of interference are oriented perpendicular to the thickness gradient.
Their colour is determined by the film thickness, between about one hundred nano-
metres and several microns, and their downwards movement is due to the thinning of
the film over time. If the film is sufficiently stable, you will see the appearance of the
black film above the coloured bands. It is generally at this point that the soap film
bursts.
By carefully observing the soap film close to the edges of the frame during drainage,
you should be able to see upward motion. This is marginal regeneration (see §5.1.1
and fig. 3.41).
Comments:
1. Glycerine can be added to the soap solution (around 10%) to increase the life of
the soap films and slow down drainage. Instead of using glycerine, the film can be
“fed” with soap solution (see experiment 6.3, chap. 2).
2. The larger the frame, the more spectacular the experiment (see fig. 2.33), but the
less stable the film.
3. By illuminating the soap film with a monochromatic light source (sodium vapour
or laser) it is possible to precisely measure the thickness and its variations.
4. If the soap film is created rapidly and protected from air currents, the drainage
dynamics will be more easily reproducible.
5. Depending on the washing-up liquid used, and its concentration, it is possible to
observe inhomogeneities in film thickness. This is due to micellar phases, or the
organization of physico-chemical complexes added to the surfactant molecules by
the manufacturer (see figs. 3.18 and 3.51). In the first case, you can get rid of
them by further diluting the solution. In the second case, you need to change the
washing-up liquid!
7.2 Free drainage in a foam and the vertical motion of bubbles

preparation time: 1 hr experimental time: 1 hr
Phenomenon demonstrated: Correspondence between downwards motion of the liquid

and upward motion of the bubbles during free drainage.
References in the text: §4.2, §4.7.2
Materials:
• a transparent, plastic cylinder, preferably with a square cross-section, with a
height of about 40 cm and a lid
• small pieces of laminated cardboard, about 1 cm in size
• soap solution: 1% washing-up liquid in deionized water
• a graduated ruler of the same height as the column
• adhesive tape
• a stopwatch.
1. Preparation of the cylinder: pierce a slit every 5 cm up one of the sides, in order
to be able to insert the cardboard markers in the foam. Fix the ruler to the same
side of the column.
2. Preparation of the foam: cover the slits with adhesive tape, pour all the solution
into the cylinder, and shake hard to produce a foam.
3. Uncover the slits and insert the pieces of cardboard as positional markers
(fig. 3.56).5
4. Measure the height L(t) of drained liquid at the bottom of the column (fig. 3.56)
as a function of time.
5. Measure the vertical movement of bubbles by tracking the position of the markers
relative to the slit as a function of time.
Under the effect of gravity, the liquid drains towards the bottom of the foam and the
much lighter gas bubbles rise. Using only a simple argument based on volume conser-
vation (the equilibrium between liquid movement downwards and bubble movement
5 As a precaution, the adhesive tape can be put back over the slits, although the foam is unlikely
to flow out.
Experiments 157
Marker
Slit
+ Slit
marker
L(t)
Time
Fig. 3.56: The free drainage experiment.
upwards), the movement of the markers allows the evolution of the liquid fraction at
different heights to be determined [52].
7.3 Forced drainage in a foam: observation of the wetting front
difficulty level: cost of material:

preparation time: 1 hr experimental time: 10 min
Phenomenon demonstrated: forced drainage

References in the text: §4.3, §4.7.1
Materials:
• a transparent cylinder of 2 to 4 cm in diameter, at least 15 cm high
• a retort stand to keep the tube vertical
• a large bowl
• 1 m of flexible rubber tubing
• a porous frit (an aquarium oxygenator for example) or a hypodermic needle
• an air pump
• a liquid pump
• two clamps to hold the rubber tubing
• soap solution: 1% washing-up liquid in deionized water.
1. Fill the bowl with soapy water. Fix the tube vertically and place its lower end in
the liquid (fig. 3.57).
2. Cut 30 cm of flexible tubing. Connect the air pump to one end of the tubing and
the porous medium or needle to the other end. Place the latter in the soapy water,
beneath the tube.
3. Fill the cylinder with foam by injecting air through the frit or needle, at constant
flow rate (to have bubbles of the same size). If the flow rate is too great it can be
reduced by pinching the rubber tubing with a clamp.
4. Once the cylinder is full, wait a few minutes for the foam to drain.
5. Connect some tubing to the water pump to take soapy water from the bowl.
Water
pump
Foam
Air
pump
Frit
Fig. 3.57: The forced drainage experiment.
6. Set the liquid flow to about 10 mL.min−1 , pinching the tube with the second clamp
if necessary.
7. Quickly place the free end of the tube at the top of the foam and keep it in this
position while the flow of soapy water continues.
A wet foam front should form at the top of the foam, and moves downwards at constant
speed. You have successfully carried out a forced drainage experiment.
Comments:
1. Flexible tubing, forceps, air, and liquid pumps and aerators can all be found in
shops selling aquarium products.
2. By illuminating the column of foam homogeneously and filming the light trans-
mitted through the foam, you can measure optically the speed of progress of the
wetting front ufront (see figs. 3.30 and 3.36). Next, redo the experiment with a
different liquid flow rate Q. Provided that the foam is very dry at the start of the
experiment (which might require a long waiting time, depending on the properties
of the foam), you can then obtain a measure of the liquid fraction in the foam
(φl = Q/(Sufront ), where S is the cross-sectional area of the cylinder), and you can
test the theory of foam drainage (§4.7.1).
7.4 Life and death of a foam measured by electrical conductivity

preparation time: 4 hrs experimental time: 4 hrs
Phenomena demonstrated: free drainage, coarsening, coalescence, electrical conductiv-

ity of a foam.
Experiments 159
References in the text: §3, §4 and §5, and §1.2.6, chap. 5.

Materials:
• materials necessary to fill a cylinder with foam, as in experiment 7.3

• a lid for the foam column
• 2 brass screws of the same diameter (about 5 mm)
• glue (preferably slow-setting Araldite)
• 2 crocodile clips
• a resistor with resistance close to 1 kΩ
• 5 cables with banana plugs
• 1 coaxial cable (BNC)
• 2 banana/BNC adapters (female/female)
• 1 male BNC T junction
• a low-frequency generator
• an oscilloscope
• optional: a computer to plot and analyse the results.
1. The day before the experiment, pierce two holes opposite each other in the cylinder
about two thirds of the way up. Insert a brass screw in each hole, so that the end
of the screw is flush with the inside wall of the column. Glue the screws in this
position: these will be the two electrodes.
2. Place the column vertically so that it is ready to be filled with foam (see
experiment 7.3). The electrodes must be in the upper part of the tube.
3. Set up the electrical wiring (see fig. 3.58) by connecting the generator in series with
the foam and the resistor. Connect the oscilloscope to the terminals of the resistor
(channel 2). In channel 1 of the oscilloscope, directly record the signal emitted by
the generator.
4. To avoid electrolysis of the foam at the terminals, use an alternating current. Set
the generator to pulse at a frequency of 1 kHz and amplitude of around 5 V.
5. Fill the cylinder with foam, adjust the oscilloscope over a long time base, and watch
the amplitude of the signal change on channel 2 of the oscilloscope.
6. Stop the bubbling and place the lid on top of the foam column.
Plot the signal over a long time (at least one hour) after the bubbling has stopped.
Over time the liquid drains out of the foam, so its conductivity decreases (elec-
trical resistance increases) and so does the amplitude of the signal being measured.
In fact, as shown on the electrical diagram in fig. 3.58, the foam not only has
an electrical resistance but also a capacitance, which depends on the amount of li-
foam. The complex impedance Z of the foam is thus Z = |Z|e , where
iϕ
quid in the
|Z| = R/ 1 + (RCω ) and ϕ = arctanh (RCω). The capacitive effects can therefore
2
be ignored at low frequency and the resistance R of the foam is
(U0 − Umeas )Rmeas

R= (3.108)
Umeas
U Umeas
R
Rmeas
C
U0
Foam
Fig. 3.58: Electrical set-up for measuring the electrical conductivity of a foam. The
generator provides a voltage U0 , and the foam has a resistance R and a capacitance C in
parallel. In channel 2 of the oscilloscope the voltage Umeas is measured at the terminals
of the measurement resistor (shunt) with resistance Rmeas . At low frequency, Umeas
doesn’t depend on C, enabling the measurement of R. The conductivity of the foam is
equal to the distance between the electrodes in contact with the foam, normalized by
their cross-sectional area, divided by the resistance R. In practice a calibration run is
performed with the foaming solution so that it is not required to know the geometrical
configuration of the electrodes precisely.
If you are using a computer, plot the amplitude of the signal as a function of
time on log-log scales. You should observe several different regimes (i.e. straight line
segments with different gradients). At short times (typically one minute) the signal
remains constant, because the liquid at the top of the column feeds the foam situated
at the level of the electrodes. The next regime corresponds to the thinning of the
Plateau borders under the effect of drainage. Finally, if the foam is unstable, the last
regime will reflect coalescence.
Comments:
1. If the signal-to-noise ratio is too low, you can add a little salt to the foam to
increase its conductivity (but not so much that you screen the interactions between
surfactant molecules, which would cause the films to rupture), or electrically isolate
the electrodes (screws) with thermo-retractable sheaths.
2. Experimentally, you can confirm that capacitive effects are negligible by varying
the frequency and confirming that the amplitude of the signal doesn’t change.
3. It is possible to link the conductivity to the liquid fraction (see §1.2.6, chap. 5). It
is then necessary to measure the conductivity of the soap solution, for example by
filling the cylinder with the solution and measuring the signal at the electrodes.
4. In the laboratory, for a more detailed analysis and a less noisy signal, we use a
lock-in amplifier instead of an oscilloscope.
5. This experiment enables you to determine the complete drainage dynamics by in-
stalling pairs of electrodes along the cylinder, and scanning the signal at each pair
of electrodes. It is then necessary to completely automate the experiment, as the
scanning time must be short compared to the evolution time of the foam.
Exercises 161
8 Exercises
8.1 Exponent in the scale-invariant regime

Consider a 2D foam.
(a) Integrate von Neumann’s law, eq. (3.26), to give an expression for the characteristic
time over which an individual bubble i, with ni sides, grows or shrinks (i.e. the
2D analogue of eq. (3.34)).
(b) Give an expression for the rate dN/dt at which bubbles with a positive topological
charge disappear.
(c) Assume that the foam has coarsened enough to reach the self-similar growth re-
gime, in which the various statistical measures of the foam no longer change in
time (see p. 105). Demonstrate that:
dN
2D: ∝ −N 2 (3.109)
dt
where the proportionality factor is a constant.
(d) Show that eq. (3.109) is equivalent to N ∝ t−1 . Show that the average bubble area
is proportional to t and so the average radius is proportional to t1/2 .
NB: the 3D case (eq. (3.40)), not described here, is similar but the formulae are more
involved. We refer instead to the approximate equations (3.37) and (3.39).
8.2 Frumkin equation of state

The Langmuir model describes the dynamics of adsorption of surfactant molecules at
the surface of a solution in the ideal case where molecules don’t interact with each other
(see eq. (3.3)). Frumkin corrected this model to account for non-negligible interactions
between molecules and when adsorption (or desorption) of a molecule at the surface
is associated with an adsorption (or desorption) energy Ea (or Eb , respectively). The
coefficients a and b in eq. (3.7) are

E 0 + Ea Γ E 0 + Eb Γ
a = a0 exp − a , b = b0 exp − b . (3.110)
kB T kB T
(a) Establish the relationship between bulk and surface concentrations by intro-
ducing the concentration ca = (b0 /a0 ) exp[(Ea0 − Eb0 )/kB T ] and the measure of
non-ideality HF = (Eb − Ea )Γ∞ /2.
(b) Use the Gibbs equation (3.2) to establish the relationship between the surface
pressure and the surface concentration.
8.3 Foam drainage and equilibrium height

Consider a foam with initially uniform liquid fraction φl0 , height Z, constant bubble
size d, and constant cross-section S.
(a) Calculate the initial volume of liquid.

(b) Calculate the volume at equilibrium.
(c) For what initial foam height are the initial and equilibrium liquid volumes equal?
(d) Calculate this foam height with φl0 = 0.01 or 0.1, d = 0.1 mm or 10 mm, and a
surface tension of 35 mN.m−1 .
8.4 Drainage in the bulk and at the wall

Using fig. 3.34, estimate the proportion of liquid flowing through bulk Plateau borders
and through surface Plateau borders in a cylinder of foam with radius a. Use the
assumption that in any given cross-section of the cylinder there is, on average, one
Plateau border per bubble of diameter d and consider the cases Bo = 0.1 and Bo = 10.
8.5 Free drainage: characteristic times and liquid fraction profiles

Eq. (3.95) gives the characteristic time τ for free drainage in the Plateau border-
dominated case.
(a) Based on the calculation presented in §4.7.2 for a foam of height Z, in particular
eq. (3.96), show that in the vertex-dominated case the characteristic time (assum-
ing that the initial liquid fraction is uniform, φl = φl0 ) corresponds to the time at
which 2/3 of the total volume of liquid has drained out of the foam.
1/2
(b) Show that τ = 2ηZ/(3Cv ρl gRV2 φl0 ). It may help to recall that the liquid fraction
profile is φl /φl0 = Av (t)(Z − z) , where Av (t) is a coefficient that depends on time
2
and z is the vertical coordinate, oriented upwards.
8.6 The true 3D pressure and 2D surface pressure

For a monolayer with (2D) surface pressure Πl , estimate the (3D) pressure within the
monolayer (pay attention to units!), and compare it with atmospheric pressure.
References
[1] A.W. Adamson and A.P. Gast, Physical Chemistry of Surfaces, Wiley, New York,
6th ed. 1997.
[2] A.A. Aradian, Quelques problèmes de dynamique d’interfaces molles, Thèse de
l’Université Paris 6, 2001. http://tel.archives-ouvertes.fr/tel-00001386.
[3] J.-C. Bacri and F. Elias, in Morphogenesis, origins of patterns and shapes,
P. Bourgine and A. Lesne (eds.), Springer-Verlag, Berlin, 2011.
[4] V. Bergeron, J. Phys.: Condens. Matter, 11, R215, 1999.
[5] A. Bhakta and E. Ruckenstein, Adv. Colloid Interface Sci., 70, 1, 1997.
[6] D. Buzza, C.-Y. Lu, and M.E. Cates, J. Phys. II, 5, 37, 1995.
[7] V. Carrier, S. Destouesse, and A. Colin, Phys. Rev. E, 65, 061404, 2002.
[8] A. Cervantes-Martinez, E. Rio, G. Delon, A. Saint-Jalmes, D. Langevin,
and B.P. Binks, Soft Matter, 4, 1531, 2008.
References 163
[9] S. Cohen-Addad and R. Höhler, Phys. Rev. Lett., 86, 4700, 2001.
[10] S.J. Cox, G. Bradley, S. Hutzler, and D. Weaire, J. Phys.: Condens. Matter,
13, 4863, 2001.
[11] F.E.C. Culick, J. Appl. Phys., 31, 1128, 1960.
[12] K.D. Danov, P.A Kralchevski, N.D. Denkov, K.P. Ananthapadmanabhan,
and A. Lips, Adv. Colloid Interface Sci., 119, 17, 2006.
[14] G. Debrégeas, P. Martin, and F. Brochard-Wyart, Phys. Rev. Lett., 75,
3886, 1995.
[15] W. Drenckhan, F. Elias, S. Hutzler, D. Weaire, E. Janiaud, and J.-C.
Bacri, J. Appl. Phys., 93, 10078, 2003.
[16] W. Drenckhan, H. Ritacco, A. Saint-Jalmes, D. Langevin, P.
McGuinness, A. Saugey, and D. Weaire, Phys. Fluids, 19, 102101, 2007.
[17] Z. Du, M. Bilbao-Montoya, B.P. Binks, E. Dickinson, R. Ettelaie, and
B.S. Murray, Langmuir, 19, 3106, 2003.
[18] J. Duplat, B. Bossa, and E. Villermaux, J. Fluid Mech., 673, 147, 2011.
[19] G. Durand, F. Graner, and J. Weiss, Europhys. Lett., 67, 1038, 2004.
[20] M. Durand and D. Langevin, Eur. Phys. J. E, 7, 35, 2002.
[21] D.A. Edwards, H. Brenner, and D.T. Wasan, Interfacial Transport Processes
and Rheology, Butterworth Heinemann, Stoneham, 1991.
[22] F. Elias, J.-C. Bacri, C. Flament, E. Janiaud, D. Talbot, W. Drenckhan,
S. Hutzler, and D. Weaire, Colloids Surf. A, 263, 65, 2005.
[23] F. Elias, I. Drikis, A. Cebers, C. Flament, and J.-C. Bacri, Eur. Phys. J.
B, 3, 203, 1998.
[24] F. Elias, C. Flament, and J.-C. Bacri, Magnetohydrodynamics, 35, 303, 1999.
[25] F. Elias, C. Flament, J.-C. Bacri, F. Graner, and O. Cardoso, Phys. Rev.
E, 56, 3310, 1997.
[26] F. Elias, C. Flament, J.A. Glazier, F. Graner, and Y. Jiang, Phil. Mag.
B, 79, 729, 1999.
[27] W.E. Ewers and K.L. Sutherland, Aust. J. Sci. Res., 5, 697, 1952.
[28] D. Exerowa and P.M. Kruglyakov, Foam and Foam Films—Theory,
Experiment, Application, Elsevier, Amsterdam, 1998.
[29] I. Fortuna, G. L. Thomas, R.M.C. de Almeida, and F. Graner, Phys. Rev.
Lett., 108, 248301, 2012.
[30] F. G. Gandolfo and H. L. Rosano, J. Colloid Interface Sci., 194, 31, 1997.
[31] P.R. Garrett, in Defoaming: Theory and Industrial Applications, P.R. Garrett
(ed.), Surfactant Science Series, Marcel Dekker, New York, 45, 1, 1993.
[32] J. Glazier, Phys. Rev. Lett., 70, 2170, 1993.
[33] J.A. Glazier, S.P. Gross, and J. Stavans, Phys. Rev. A, 36, 306, 1987.
[34] I.I. Gol’dfarb, K.B. Khan, and I.R. Shreiber, Fluid Dyn., 23, 244, 1988.
[35] F. Graner, B. Dollet, C. Raufaste, and P. Marmottant, Eur. Phys. J. E,
25, 349, 2008.
[36] S. Hilgenfeldt, S.A. Koehler, and H.A. Stone, Phys. Rev. Lett., 86, 4704,
2001.
[37] S. Hutzler, S.J. Cox, E. Janiaud, and D. Weaire, Colloids Surf. A., 309, 33,
2007.
[38] S. Hutzler and D. Weaire, Phil. Mag. Lett., 80, 419, 2000.
[39] S. Hutzler, D. Weaire, and R. Crawford, Europhys. Lett., 41, 461, 1998.
[40] I.B. Ivanov, Thin Liquid Films: Fundamentals and Applications, Surfactant
Sciences Series Vol. 29, CRC Press, Boca Raton, 1988.
[41] Y. Jayalakshmi, L. Ozanne, and D. Langevin, J. Colloid Interface Sci., 170,
358, 1995.
[42] S.A. Koehler, S. Hilgenfeldt, and H.A. Stone, J. Colloid Interface Sci.,
276, 420, 2004.
[43] S.A. Koehler, S. Hilgenfeldt, and H.A. Stone, Langmuir, 16, 6327, 2000.
[44] J. Lambert, I. Cantat, R. Delannay, R. Mokso, P. Cloetens, J.A.
Glazier, and F. Graner, Phys. Rev. Lett., 99, 058304, 2007.
[45] J. Lambert, R. Mokso, I. Cantat, P. Cloetens, J.A. Glazier, F. Graner,
and R. Delannay, Phys. Rev. Lett., 104, 248304, 2010.
[46] D. Langevin, Curr. Opin. Colloid Interface Sci., 3, 600, 1998.
[47] R.A. Leonard and R. Lemlich, AIChE J., 11, 18, 1965.
[48] E. Lorenceau, N. Louvet, F. Rouyer, and O. Pitois, Eur. Phys. J. E, 28,
293, 2009.
[49] N. Louvet, Etude multiéchelle du transport de particules dans les mousses li-
quides, Thèse de l’Université Paris-Est Marne-la-Vallée, 2009. http://tel.archives-
ouvertes.fr/tel-00541198
[50] J. Lucassen and M. van der Tempel, Chem. Eng. Sci., 27, 1283, 1972.
[51] A.H. Martin, K. Grolle, M.A. Bos, M.A. Cohen-stuart, and T. van Vliet,
J. Colloid Interface Sci., 254, 175, 2002.
[52] G. Maurdev, A. Saint-Jalmes, and D. Langevin, J. Colloid Interface Sci.,
300, 735, 2006.
[53] W.R. McEntee and K.J. Mysels, J. Phys. Chem., 73, 3018, 1969.
[54] D. Monin, A. Espert, and A. Colin, Langmuir, 16, 3873, 2000.
[55] W. Müller and J.M. Di Meglio, J. Phys.: Condens. Matter, 11, L209, 1999.
[56] W.W. Mullins, Acta Metall., 37, 2979, 1989.
[57] W.W. Mullins, J. Appl. Phys., 59, 1341, 1986.
[58] K.J. Mysels, K. Shinoda, and S. Frenkel, Soap Films, Studies of their
Thinning, Pergamon Press, London, 1959.
[59] S.J. Neethling, H.T. Lee, and J.J. Cilliers, J. Phys.: Condens. Matter, 14,
331, 2002.
[60] O. Pitois, C. Fritz, and M. Vignes-Adler, J. Colloid Interface Sci., 282, 458,
2005.
[61] O. Pitois, E. Lorenceau, N. Louvet, and F. Rouyer, Langmuir, 25, 97, 2009.
[62] O. Pitois, N. Louvet, and F. Rouyer, Eur. Phys. J. E, 30, 27, 2009.
[63] H.M. Princen and S.G. Mason, J. Colloid Interface Sci., 20, 353, 1965.
[64] A. Prins in Food Emulsions and Foams, E. Dickinson (ed.), Royal Society of
Chemistry Special Publication, 58, 30, 1986.
[65] D. Reinelt and A. Kraynik, J. Rheol., 44, 453, 2000.
[66] A. Saint-Jalmes, Soft Matter, 2, 836, 2006.
References 165
[67] A. Saint-Jalmes and D. Langevin, J. Phys.: Condens. Matter, 14, 9397, 2002.
[68] A. Saint-Jalmes, S. Marze, H. Ritacco, D. Langevin, S. Bail, J. Dubail,
L. Guingot, G. Roux, P. Sung, and L. Tosini, Phys. Rev. Lett., 98, 058303,
2007.
[69] A. Saint-Jalmes, Y. Zhang, and D. Langevin, Eur. Phys. J. E, 15, 53, 2004.
[70] A. Salonen, M. In, J. Emile, and A. Saint-Jalmes, Soft Matter, 6, 2271, 2010.
[71] G. Singh, G.J. Hirasaki, and C.A. Miller, J. Colloid Interface Sci., 184, 92,
1996.
[72] J. Stavans, Rep. Progr. Phys., 56, 733, 1993.
[73] H.A. Stone, S.A. Koehler, S. Hilgenfeldt, and M. Durand, J. Phys.:
Condens. Matter, 15, S283, 2003.
[74] C. Stubenrauch and R. von Klitzing, J. Phys.: Condens. Matter, 15, R1197,
2003.
[75] G.L. Thomas, R.M.C. de Almeida, and F. Graner, Phys. Rev. E, 74, 021407,
2006.
[76] N. Vandewalle, J.F. Lentz, S. Dorbolo, and F. Brisbois, Phys. Rev. Lett.,
86, 179, 2001; N. Vandewalle and J.F. Lentz, Phys. Rev. E, 64, 021507, 2001;
N. Vandewalle, H. Caps, and S. Dorbolo, Physica A, 314, 320, 2002.
[77] M.U. Vera and D.J. Durian, Phys. Rev. Lett., 88, 088304, 2002.
[78] G. Verbist, D. Weaire, and A. Kraynik, J. Phys.: Condens. Matter, 8, 3715,
1996.
[79] J. von Neumann, in Metal Interfaces, American Society for Metals, Cleveland,
1952, p. 108
[80] A. Vrij, Discuss. Faraday Soc., 42, 23, 1966; A. Vrij and J.Th. Overbeek, J.
Am. Chem. Soc., 90, 3074, 1968.
[81] A.F.H. Ward and L.J. Tordai, J. Chem. Phys., 485, 63, 1946.
[82] D. Weaire, S. Hutzler, G. Verbist, and E. Peters, Adv. Chem. Phys., 102,
315, 1997.
[83] D. Weaire and V. Pageron, Phil. Mag. Lett., 62, 417, 1990.
[84] D. Weaire and N. Rivier, Contemp. Phys., 25, 59, 1984.
[85] A.J. Webster and M.E. Cates, Langmuir, 17, 595, 2001.
4
Rheology
Rheology is the study of the deformation and flow of “complex” fluids which exhibit both
liquid and solid behaviour. Such materials are common in our everyday lives (cf. §1, chap. 1).
Foams, although mostly air and water, are certainly complex fluids. A small amount of shaving
foam or whipped cream doesn’t spread out under its own weight, but keeps the shape it is
given, yet it can flow if sufficient force is applied. This complex behaviour forms the basis of
this chapter.
1 Introduction
The mechanical properties of foams are important in numerous industrial applications
(cf. §3, chap. 1). In oil recovery, for example, aqueous foams are used as a lubricant
and to keep solid debris in suspension when drilling stops. Moreover, a foam’s vis-
cosity decreases as the strain rate increases, which decreases the dissipative losses. In
addition, the low density of foam reduces the hydrostatic pressure in the wellbore, a
significant advantage for wells that are several kilometres deep.
Aqueous foams typically exhibit a combination of the following three types of
behaviour:
• elastic: an aqueous foam can deform reversibly; it can store up mechanical energy
that it will release on returning to its initial shape.
• plastic: beyond a certain point the deformation is irreversible and a foam will
acquire and retain a new shape. The amount of energy dissipated depends on how
much of the deformation is irreversible, but not on the rate of deformation.
• viscous: an aqueous foam can flow like a liquid, dissipating an amount of energy
that depends on the deformation rate.
The special rheological properties of a foam are a consequence of their high interfacial
area. Due to the many different length- and time-scales, foam rheology is particularly
difficult to model. At the macroscopic scale (i.e. large compared to the bubble size),
the relationships between stress, deformation, and flow are described with constitutive
laws, which take into account the characteristic time-scales of the external forcing (for
example the strain rate of a steady flow, or the frequency of an oscillatory shear).
However, it is also essential to understand how the behaviour at the (local) scale of
the films and bubbles is linked to the macroscopic response. This enables a better
description and control of the rheological behaviour of a foam, and the possibility to
give to it the characteristics required by a given industrial process [13].
This chapter is divided into three parts. The first (§2) consists of a brief over-
view of topics in the rheology of complex fluids that are relevant to foams. We will
168 Rheology
introduce constitutive laws and common rheological tests, and show how to define the
strain (especially at large deformations) in a complex fluid and the stress tensor in a
mixture of two immiscible fluids (especially water and air).
In the second part (§3) we consider the foam at the film-scale, in order to determine
the origin of the local processes that give rise to the macroscopic rheological response.
The elasticity of a foam is due to the fact that a small applied stress increases the film
area, and thus the surface energy, per unit volume. Beyond the elastic limit, an applied
stress induces irreversible bubble rearrangements (T1s) and hence plasticity. We will
show how to calculate the change in film area in a dry foam and then we will calculate
the elastic modulus and the yield stress, both of which depend on the foam structure.
Finally, we will describe the various processes by which a foam loses mechanical energy,
whether through hydrodynamic dissipation in the liquid (films, Plateau borders) or
dissipation linked to the movement of surfactants at the interfaces.
The third and final part of the chapter (§4) is concerned with the link, as it is
currently understood, between mechanisms at the local scale and the macroscopic
rheological properties. We will see why a foam is much softer and flows more easily
when it is wet. For small slow deformations, the response is dominated by elasticity
and the shear stress is linear in the strain. Yet there are also non-linear elastic effects,
for example the normal stress difference varies quadratically in the strain, as for an
elastic solid. We will discuss the effects of plasticity at large strains and, noting that a
foam is elastic over short times and liquid-like over long times, the dependence of the
rheological response of a foam on the characteristic time-scale of the perturbation.
We will identify different mechanical relaxation processes and show how they are
coupled to coarsening. Finally, we will explain why the steady flow of a foam may
be inhomogeneous.
2 Overview of the rheological behaviour of complex fluids

We will give a general overview of the rheology of complex fluids, restricting ourselves
to those parts which are useful for describing the mechanical behaviour of liquid foams.
The reader can find further details in specialist books [41, 43].
2.1 Constitutive laws

In order to model the macroscopic rheological behaviour of a complex fluid consisting of
a large number of bubbles, it is necessary to treat the fluid at a length-scale sufficiently
large that the details of the microstructure can be ignored. That is, we must consider
the foam sample as a continuous medium with a constitutive law that relates stress,
strain, and strain rate. Other parameters may appear in this relationship, for example
temperature, liquid fraction, etc. We start by considering shear stresses before giving
general expressions for constitutive laws in §2.3 and §2.4.
Simple shear can be described with respect to Cartesian axes x, y, z, as shown
in fig. 4.1. Each point of the cube is displaced in the x direction while its y and z
coordinates remain the same. The displacement is linear in y, and we define the shear
strain as ε = u/d. The shear stress σxy is the force per unit surface area which acts
Overview of the rheological behaviour of complex fluids 169
σyy
y df
n
σxy u
σxx d
x
σzz
z Before After
Fig. 4.1: Simple shear of a cube of material in the x direction. The strain is given
by ε = u/d. The normal stresses σxx , σyy , and σzz correspond to the forces per unit
surface area which act perpendicular to the faces of the cube in each of the three
directions. n is a unit vector normal to the upper face of the cube, and df indicates
the force which acts on this face.
tangentially on the top surface of the sample (i.e. the surface with normal in the
y direction) in the x direction. The relationship between σ and ε can be linear or
non-linear, as we shall see.
Linear response The linear static response of an elastic solid is described by
Hooke’s law,
σ = G ε, (4.1)
where the constant G is the static shear modulus. G is related to the elastic energy
density E by

∂ 2 E
G= . (4.2)
∂ε2 ε=0
Elastic behaviour is represented symbolically with an elastic spring (fig. 4.2).

On the other hand, a liquid is called Newtonian if the shear stress is proportional
to the shear rate ε̇, that is
σ = η ε̇, (4.3)
where the constant η is the viscosity. A Newtonian liquid is represented as a dashpot

(fig. 4.2). In an elastic solid, an oscillatory stress and the oscillatory strain that it
stimulates are in phase, whereas for a liquid we observe a phase shift of π/2 between
them. We can describe the relationship between stress and strain in terms of phase
and amplitude using the complex shear modulus G∗ (§2.2) whose real and imaginary
parts, written G and G , describe, respectively, the elastic part and the viscous part
of the response of a material.
In general, whether a complex fluid behaves like a solid or a liquid depends on
the characteristic time-scale of the external forcing. For example, a Maxwell liquid
170 Rheology
Response to a step stress Oscillatory response

Model σ or ε
and σ
shear modulus t
t
ε
Elastic solid G'
Modulus
G''
G* = G
t ω
Newtonian liquid ε
Modulus
G''
G* = i ηω G'
t ω
Kelvin–Voigt solid
ε
G'
Modulus
G''
G* = G + iηω t ω
Maxwell liquid ε
G'
Modulus
iGηω G''
G* =
G + i ηω t ω
Fig. 4.2: Examples of linear rheological behaviour, showing the time and frequency
response of elementary models (see eqs. (4.8) and (4.9)). For the Maxwell liquid,
the maximum of G corresponds to the frequency ω = G/η, with the characteristic
relaxation time τ = η/G.
is elastic at short times (high frequency or large applied strain rate) and viscous at
long times (low frequency or small applied strain rate). On the other hand, a Kelvin–
Voigt solid is viscous at short times and elastic at long times. These two material
responses are modelled by placing a spring and a dashpot either in series or in parallel
(fig. 4.2).
Non-linear response If we gradually increase the strain applied to a material, the

elastic stress becomes a non-linear function of the strain and in the end the material
yields. If it is fragile, then it fractures, whereas if it is ductile, it undergoes a perman-
ent irreversible deformation εp , known as plastic deformation. If the applied strain
increases rapidly, there is an additional stress due to viscous friction. Elastic, viscous,
and plastic behaviour can be modelled with springs, dashpots, and sliding friction
elements.
a)
σy σ
G
σy
G
ε
εp
b)
ηp σ
σy ηp
σy
.
ε
Fig. 4.3: Examples of non-linear rheological behaviour. On the left, the spring rep-
resents an elastic response, the sliding element represents solid friction, and the
dashpot represents a viscous response. The mechanical response is shown on the
right. (a) Response of an elasto-plastic solid to an applied stress that increases to
the yield stress and is then reduced back to zero. (b) Representation and flow curve
of a Bingham fluid.
For quasi-static applied strains, i.e. strains that are applied sufficiently slowly that
the viscous stresses in the complex fluid are negligible, the response is “elasto-plastic”,
a combination of elastic and plastic responses. This is represented by a spring and
a sliding friction element in series (fig. 4.3a). The yield stress σy , above which the
friction element slides, is the maximum static stress that the material can withstand
before it flows. The corresponding value of strain is the yield strain εy . Above the yield
strain, the strain can be separated into two contributions,
ε = εe + εp , (4.4)
the sum of a plastic part εp and an elastic part εe ; the latter is reversible and is thus
εe = σy /G, at least in the linear elastic regime.
In order to model the steady flow of a “yield stress fluid” (i.e. a material that
behaves as a solid or a liquid depending on the applied stress) over a large range of
strain rates, we must account for viscous stresses. The Bingham model is often used.
It is represented by a sliding friction element in parallel with a viscous dashpot that
has a “plastic” viscosity ηp (fig. 4.3b). The constitutive law is
ε̇ = 0 if σ < σy
σ = σy + ηp ε̇ if σ > σy . (4.5)
When σ > σy , the effective viscosity σ/ε̇ of a Bingham fluid decreases as the shear
rate increases. This is known as shear-thinning behaviour.
172 Rheology
The empirical Herschel–Bulkley constitutive law adds to the Bingham model a fur-
ther free parameter β, which allows a more accurate fit to the shear-thinning response
found experimentally. Above the yield stress it is given by
σ = σy + ηp ε̇β if σ > σy . (4.6)
Since for β = 1 the units of ηp are no longer those of a viscosity, ηp is sometimes known
as the consistency of the fluid.
More complex configurations of springs, viscous dashpots, and frictional elements
can be constructed to describe the relationships between stress, strain, and strain rate,
not only of foams but of many other complex materials [3, 9, 55]. In §3 we will identify
the microstructural origin of these elements.
2.2 Shear tests

We next describe a few shear tests that are used to probe the rheological response
of complex fluids, particularly foams. The equipment used for these experiments is
described in §1.2.7, chap. 5.
Shear start-up This test consists of applying to the material an increasing strain,
ε(t) = ε̇ t, at constant shear rate ε̇, and measuring the resulting stress to give both the
yield stress and the effective viscosity, σ/ε̇.
Steady flow A steady flow experiment is determined by both the stress and the
strain rate. We set one, wait until a steady state is reached, then measure the other.
This results in a flow curve σ(ε̇). For a Newtonian liquid the flow curve is a straight
line through the origin with a slope that corresponds to the viscosity η (eq. (4.3)). The
flow curve for a Bingham fluid (fig. 4.3b) is also a straight line, but it does not pass
through the origin: instead it intersects the stress axis at the yield stress σy (eq. (4.5)).
For a Herschel–Bulkley fluid the flow curve also intersects the stress axis at the yield
stress, but it is not straight (eq. (4.6)).
Relaxation and creep In a relaxation test a strain ε is applied and the change in
the stress σ(t) is measured as a function of time. The response is characterized by the
relaxation modulus, G(t) = σ(t)/ε. For a perfectly elastic solid, G(t) is constant and
equal to the static shear modulus G (eq. (4.1)).
Conversely, during a creep experiment the stress σ is prescribed and the change
in the strain ε(t) is measured as a function of time (fig. 4.2). The mechanical re-
sponse is characterized by the compliance J(t) = ε(t)/σ. For a perfectly elastic solid,
J is constant and equal to 1/G. For a viscoelastic material, J(t) differs from 1/G(t)
(eq. (4.10)), and a creep experiment is an appropriate method for studying slow
mechanical relaxations, as we shall see in §4.1.5.
Oscillatory response The rheological properties of a viscoelastic material may be
probed over a given time-scale by applying oscillatory shear, that is, a sinusoidal strain
of amplitude ε0 and frequency ω, and measuring the resulting stress. If the amplitude
of the strain is sufficiently small, the stress will also be sinusoidal, with an amplitude
σ0 proportional to ε0 , and a phase shift of δ with respect to the strain.1 In complex

notation this is written
ε(t) = ε0 R[eiωt ] σ(t) = σ0 R[ei(ωt+δ) ], (4.7)
where R[ ] denotes the real part. The constitutive law,
σ(t) = ε0 R[G∗ (ω)eiωt ], (4.8)
is expressed in terms of the complex shear modulus
G∗ (ω) = G (ω) + iG (ω). (4.9)
The storage modulus G describes the capacity of a material to store elastic energy,
and the loss modulus G describes its dissipative character. For an elastic solid, G∗
is real and equal to the static shear modulus G = G, G = 0 (eq. (4.1)), while the
complex modulus of a Newtonian liquid is purely imaginary: G = 0 and G = η ω.
Fig. 4.2 illustrates these viscoelastic moduli for several common models.
In the linear regime the information obtained from a relaxation (or creep) ex-
periment is equivalent to that obtained by applying an oscillatory shear. In fact, in
this case the relaxation modulus and the compliance are linked to G∗ by an integral
transform [23], for example
∞
1
= iω J(t)e−iωt dt. (4.10)
G∗ (ω) 0
If we apply a sinusoidal stress of frequency ω and an amplitude σ0 or ε0 that in-

creases from zero, we go beyond the linear regime, and even exceed the yield strain.
Consequently the response includes not only a Fourier component at the funda-
mental frequency ω, but also a whole spectrum of harmonics at multiples of ω. Under
these conditions, the complex shear modulus is defined in terms of the fundamental
component and depends on the strain amplitude ε0 :
2π/ω
∗ ω
G (ω, ε0 ) = σ(t)e−iωt dt. (4.11)
π ε0 0
Then G∗ (ω, ε0 ) does not completely characterize the non-linear rheological response
because the harmonics are not taken into account.
Finally, note that if G (ω, ε0 ) > G (ω, ε0 ) the predominant behaviour is that of an
elastic solid, whereas if G (ω, ε0 ) < G (ω, ε0 ) it is that of a viscous liquid.
2.3 Small and large strains

We have seen how to describe simple shear by its amplitude ε (fig. 4.1). In order
to characterize any general strain, we must specify, for a point with initial position
1 An equivalent experiment (in the linear regime) consists of applying an oscillatory stress and
measuring the strain.
174 Rheology
vector x, the new position x (x) that results from applying the strain. We can then
characterize the displacement field X(x) = x − x, as a function of the initial positions.
This is simplified by considering the displacement field at a much larger scale than
that of the microstructure. The material can then be regarded as a homogeneous,
continuous medium.
An infinitesimal deformation is then described by the infinitesimal strain tensor

1 ∂Xi ∂Xj
εij = + with i, j = 1, 2, 3, (4.12)
2 ∂xj ∂xi
where we identify the 1 direction with x, the 2 direction with y, and the 3 direction
with z. In this notation, a simple shear strain ε (cf. fig. 4.1) is 2 εxy .
This is appropriate in the linear regime. At greater strains, in the non-linear re-
gime, linear functions of εij are no longer sufficient to describe the kinematics. To
characterize the strain here we introduce the deformation gradient tensor Fij and the
Finger tensor Bij :
∂xi
Fij = and Bij = Fik Fjk . (4.13)

∂xj
k
Non-linear elastic constitutive laws are expressed in terms of Bij [43]. Beyond the
yield strain, this total strain differs from the elastic strain (eq. (4.4)) due to plastic
deformation.
2.4 Stress tensor in a complex fluid

2.4.1 Definition To define the stress tensor in 3D we consider one face of a small
cube of material (fig. 4.1) of area dS and unit normal vector n pointing outwards.
The total force df acting on the cube across the surface dS has components

dfi = σij nj dS, (4.14)

j
where σij is the stress tensor (and a shear stress corresponds to the element σxy ,
cf. p. 169). The pressure p exerts a stress −p δij , but for an incompressible medium,
the pressure affects neither the shape nor the behaviour of the material. We therefore
subtract from the stress an isotropic part −p δij and the rest, σij + p δij , is known as the
deviatoric stress tensor. Its off-diagonal elements represent shear, and its three diagonal
elements are the differences of the normal stresses from their average value p = (σxx +
σyy + σzz )/3. The latter are the forces per unit surface area acting perpendicularly to
the faces of the cube, in each of the three coordinate directions (fig. 4.1). Hooke’s law
(of which eq. (4.1) is a simplified version) in tensor form is
σij = 2 G εij − p δij , (4.15)
and Newton’s law (eq. (4.3)) can be generalized to
σij = 2 η ε̇ij − p δij . (4.16)

To characterize the normal stresses in a way which does not depend on changes in
ambient pressure, we construct two independent normal stress differences:
N1 = σxx − σyy and N2 = σyy − σzz . (4.17)
For example, consider the compression of a cube in the y direction (fig. 4.1), leaving
its lateral faces free. Then N1 and N2 are independent of the pressure and vary linearly
with the (uniaxial) strain according to Hooke’s law. But, remarkably, normal stress
differences can also appear during shear. This non-linear effect will be discussed in
§4.1.2.
2.4.2 Relationship between stress and structure Understanding the micro-

structure of a material is key to understanding the origin of its stresses. For example,
in a crystal, macroscopic stresses are the consequence of particular interatomic
forces. Similarly, the rheological properties of a foam are essentially governed by the
interfaces.
Batchelor [2] established the general expression for the stress induced in a mixture
of two immiscible phases by a quasi-static strain. The stress tensor averaged over
a volume V , denoted σij , is linked to the fluid microstructure, the pressures in the
dispersed phase, and the interfacial tension γ according to

1 γ
σij = − Vk Pk δij − φl pδij + (δij − ni nj ) dS. (4.18)
V V Sint
bubbles k
The first two terms represent the average pressure in the dispersed and continuous
phases, with Vk the volume of bubble k, Pk its pressure, p the pressure in the liquid,
and V φl its volume. The last term represents the contribution of the interfaces of
tension γ (recall that a liquid film has two interfaces). The integration domain Sint
consists of the total interface in a volume V , and δij − ni nj represents projection on to
the surface with local normal n; the orientation of the normal can be chosen arbitrarily
since the product ni nj does not depend on it. When we apply a macroscopic strain,
the areas and the orientations of the interfaces change, and eq. (4.18) predicts the
resulting stress. The shaded section below derives the stress tensor for the case of a
dry 2D foam.
The Batchelor expression shows that if we evaluate the average stress in a volume
V smaller than a bubble, the result will fluctuate strongly depending on where this
volume is located. For this reason, the rheology of a foam is often modelled at an
intermediate scale between that of the bubble and that of the whole foam. This scale
must be sufficiently large that the stress is independent of the microstructure, and thus
continuous and differentiable; however, it must remain small enough to demonstrate
the spatial variations of stress within a sample. Finally, we note that the total stress
in a flowing foam can include, in addition to this static stress, dynamic contributions
due to viscous forces.
176 Rheology
The stress tensor for a dry 2D foam

For a dry foam, the second term in eq. (4.18) vanishes. The first term gives the
average gas pressure in the bubbles and the last term is the stress due to the sur-
face tension of the soap films. We demonstrate here its derivation in 2D, where
a stress is a force per unit length, and the line tension λ is a force (cf. §5.2,
chap. 2).
Due to the microstructure of a foam, the stress is not defined locally and
eq. (4.14) is only meaningful with a finite value of dS, much larger than the typical
bubble size. To calculate the average stress in a rectangular foam of size Δx by
Δy and area A, we consider the force acting across an arbitrary line L in the foam
(fig. 4.4). We denote by f(xL ) the total force exerted by the foam to the right of
L on the foam to the left. If the foam were continuous, fj (xL ) would be given by
eq. (4.14),
fj (xL ) = σ1j Δy. (4.19)
The average stress tensor depends on the value of f(xL )/Δy, averaged over position
xL . To obtain the components of the stress tensor we must therefore integrate
over xL :
1
Δx
1
σ1j = fj (xL )dxL . (4.20)
Δx Δy
0
Δy
dx
t
n
Lk
0 x
x Δx
Fig. 4.4: A 2D foam with area A = ΔxΔy. The zoom shows a film that crosses
the straight line L, with unit tangent and normal vectors t and n, respectively.
We first estimate the contribution due to interfacial forces by writing eq. (4.20)
as an integral along the films. The length d of a film element which crosses a strip
of width dxL (fig. 4.4) satisfies
dxL = tx d, (4.21)
where t = (tx , ty ) is a unit vector tangent to the film where it intersects L. The
interfacial force exerted by this film on the foam to the left of L is 2λ t. Using
eq. (4.20), we find the surface tension contributions to the stress:

1
σxx = 2 λ tx tx d, (4.22)
A A

1
σxy = 2 λ tx ty d. (4.23)
A A
The integrals are taken over the total length of films in the area A. A similar
reasoning gives σyx = σxy (the stress tensor is clearly symmetrical, as expected [43]),
and

1
σyy = 2 λ ty ty d. (4.24)
A A
Now consider the pressure Pk in a bubble k, which intersects L over a length

Lk (which may be zero). The gas to the right of L exerts on the line segment Lk
a force −Pk Lk in the x direction. Since the area of bubble k is Ak = Lk dx1 , we
obtain
1
σxx = σyy = − Pk A k . (4.25)

A
bubbles k
(The reasoning required to obtain this expression for σyy is the same, but with L
parallel to the y axis.)
We recover Batchelor’s expression in 2D (eq. (4.18)) by adding up the con-
tributions due to interfacial tension (eqs. (4.22), (4.23), (4.24)) and to pressures
(eq. (4.25)), and by noting that in 2D we have ti tj = δij − ni nj .
2.4.3 Trapped stresses Even if a complex material is at mechanical equilibrium

and the average deviatoric stress is zero, it can nevertheless experience variations in
stress at the scale of structural heterogeneities.2 These trapped stresses result from
the history of the strains applied to the sample, for example during its manufacture.
2 Likewise, in atomic solids, trapped stresses are often found in the vicinity of defects such as
dislocations.
178 Rheology
Fig. 4.5: Trapped elastic stresses in a network of springs, each with the same stiffness
and a different relaxation length.
To illustrate this phenomenon, we consider a group of springs whose ends are fixed
to two parallel bars (fig. 4.5). If each spring has a different rest length, they can
never all be relaxed simultaneously, no matter what the distance between the bars. So
there are always trapped elastic forces that can’t be relaxed without modifying the
structure of the system. This idea can be generalized to assemblies of springs with
different lengths in a regular network, or to identical springs in a disordered network.
At a macroscopic scale, the mechanical state of such a network is characterized by
trapped stresses. By analogy, it is the bubbles in a foam which play the role of springs
and are subject to trapped stresses.
3 Local origin of rheological properties

We seek to understand the origin of the mechanical properties of a foam at a local
scale (of bubbles or soap films). Why is it elastic? What is the value of strain or stress
above which it flows? How is mechanical energy dissipated? In §4 we will describe
the link between the mechanisms at the local scale and the macroscopic rheological
response.
3.1 Elastic shear modulus of a dry monodisperse foam

In the absence of an applied stress, the structure of a dry foam at equilibrium corres-
ponds to a minimum of surface energy, since the volume of gas in each bubble is fixed.
In order to deform a foam, it is necessary to apply a stress. As long as this stress is
not too great, the structure returns to its initial state as soon as the stress is removed:
this is a foam’s elastic response.3 An elastic response is easily observed experimentally,
as long as care is taken to avoid T1s (cf. §1.2.2, chap. 3). In this section, we consider
simple shear of small amplitude (linear regime) in the quasi-static limit, i.e. at each
instant, the structure, although deformed, is at equilibrium (a more precise definition
is given in the shaded section on p. 186). This includes elongational strains at constant
volume, since at small amplitude simple shear is in effect equivalent to a uniaxial strain
followed by a rotation.
3 This is the usual elasticity, which should not be confused with Gibbs’ surface elasticity (cf.
eq. (3.11)). Here, surface tension is assumed constant.
Local origin of rheological properties 179
3.1.1 Dimensional analysis and orders of magnitude The elastic response of

a foam is influenced by pressure, capillary forces, and the characteristic length-scale
of the structure, although pressure does not affect the shear stress (cf. eq. (4.18)). In
the dry limit, Plateau border radii and film thicknesses are negligible compared to
the bubble size and it can be shown that they do not influence the static response
(although in wet foams the liquid fraction does play an important role, see §4.1.1).
The shear modulus G (eq. (4.1)) thus depends primarily on surface tension γ and the
characteristic size of the bubbles d. As a modulus is an energy density or a pressure, we
thus expect G ∼ γ/d. We will show that this scaling is indeed appropriate for a range
of dry foams (2D, 3D, ordered, and disordered), with a different numerical coefficient
for each structure.
Note that for 2D foams, the line tension (with the dimensions of a force of the
order of γe, cf. §5.2, chap. 2) is denoted λ, Ĝ is the 2D shear modulus (also in N.m−1 )
and σ̂ is the 2D stress (in N.m−1 ). For simplicity, the .̂ will usually be omitted.
The shear modulus γ/d is of the order of 300 Pa for d ≈ 100 μm and
γ ≈ 30 mN.m−1 . Therefore foams are certainly “soft matter” compared to ordinary
solids (metal, glass, etc.) which have a modulus of the order of 1010 to 1011 Pa! When
a sample of aqueous foam is squashed, it flattens like a soft incompressible material.
In fact, the bulk (compression) modulus of a foam is dominated by the response of the
gas, being of the order of the external pressure (≈ 105 Pa) and certainly greater than G
(see exercise 7.6). In practice compressive deformations arise for example during the
propagation of sound4 or shock waves (which we do not discuss here).
3.1.2 Ordered 2D foam In his pioneering work on foam rheology, Henry Princen
considered a model 2D foam consisting of an ordered network of regular, identical
hexagons. This simple model gives a good idea of the effect of shear at the bubble
scale [49]. Fig. 4.6 shows how the shape of the bubbles changes as an increasing
(a) ε=0 (b) ε = 0.5
θshear
h
Fig. 4.6: Quasi-static shear of a 2D hexagonal foam with edge length h at rest. The
structure remains ordered as the strain ε increases from (a) 0 to (b) 0.5. All bubbles
therefore have the same pressure, and the edges are straight and meet at 120◦ . The
stress exerted across the line L, calculated in the text, is non-zero if θshear is different
from 90◦ , i.e. for ε = 0.
4 Measuring the speed of sound in a dry foam is one way of determining its liquid fraction, since the
two are strongly correlated. At a frequency of a few kHz, it is about 50 m.s−1 for φl = 0.1 and reaches
a minimum of the order of 20 m.s−1 for a bubbly liquid which is half water and half air [19, 25, 47].
It is much lower than the speed of sound in water (1500 m.s−1 ) or in air (340 m.s−1 ).
180 Rheology
shear strain ε is applied. The disordered case is dealt with in more detail in [37] and
exercise 7.2.
Stress equilibrium The two states in fig. 4.6 both satisfy Plateau’s laws (§2.2,
chap. 2). In fact, the foam is free to relax locally to a minimum energy structure.
However, the applied shear reduces the number of degrees of freedom of the foam by
fixing the points where the edges meet the top and bottom walls of the “box”. A shear
stress σ is exerted across lines parallel to the x direction. In this way equilibrium states
with non-zero stress (σ = 0), distinct from the stress-free state (σ = 0 and ε = 0), are
found.
Stress and shear modulus The foam located on one side of a line L parallel to
the x axis (fig. 4.6b) exerts a shear stress σ on the foam on the other side, as a result
of the edges which are “cut” by L and pull in the θshear direction. As the strain ε
increases, the angle θshear decreases, and so σ grows. Writing h for the length of an
edge when ε = 0 and λ for the line tension, the 2D stress σ̂ (the force per unit length)
is [40]
2λ ε
σ̂ = √ √ . (4.26)
2
3h ε + 4
Linearizing this expression gives the shear modulus G = σ/ε:
λ λ λ Lint
Ĝhex = √ = 0.52 = . (4.27)
3 h RA 4 Afoam
RA is the radius of a disc of the same area as a hexagon with edge length h
(RA ≈ 0.909 h ) √ and Lint /Afoam is the length of interface per unit area when ε = 0,
Lint /Afoam = 4/( 3h ). Ĝ is clearly of the order of the line tension divided by the
bubble size. Note that in the linear regime, the energy, and thus the total quantity of
interface, varies as ε2 (p. 169). In this example, the modulus Ĝ is independent of the
shear direction [60]. As we will see, this is not the case for a Kelvin foam.
3.1.3 Ordered 3D foam The Kelvin cell forms the unit cell of an ordered dry
foam (see shaded section on p. 50) in 3D, as the hexagon does in 2D. Under the effect
of shear, the area and orientation of the faces change (fig. 4.7). Due to the cubic
symmetry, the magnitude of these variations depends on the direction in which the
shear is oriented. Consequently, the elastic response of a Kelvin foam is anisotropic.
Using the Surface Evolver (§2.1.2, chap. 5), Kraynik and Reinelt showed that the shear
modulus averaged over all orientations is [33]
γ Sint γ
GKelvin = 0.15 = 0.50 , (4.28)
Vfoam RV
where Sint is the total interfacial area and Vfoam the total volume of foam. Here again,
the prefactor is of order one, as in eq. (4.27) for a hexagonal foam. Depending on the
shear direction, the prefactor varies between 0.35 and 0.60.
ε=0 ε = 0.5 ε y = 1.1
Fig. 4.7: A numerical simulation using the Surface Evolver gives the shape of a Kelvin
cell for different strains ε. The shear direction corresponds here to the (100) direction
of a Kelvin foam’s bcc lattice.
3.1.4 Disordered 3D foam: affine deformation When a 3D foam is disordered,

it is difficult to predict the precise deformation of each film under the effect of a
shear. The simplest hypothesis assumes that the deformation is affine: a point initially
situated at position A(x, y, z) is found at position A (x + εy, y, z) after a shear strain
ε. In an isotropic foam the films are oriented randomly, which leads to Derjaguin’s
expression [18]:
4 γ Sint γ Sint
Gaffine,3D = = 0.26 . (4.29)
15 Vfoam Vfoam
Numerical simulations with the Surface Evolver show that in a disordered mon-
odisperse dry 3D foam [35] Sint /Vfoam ≈ 3.3/RV (cf. eq. (2.19)).5 From this we
deduce
γ
Gaffine,3D = 0.88 . (4.30)
RV
An affine deformation does not satisfy Plateau’s laws—the angles between films are
no longer 120◦ and the angle at which edges meet at a vertex is no longer 109.5◦ —
so that the resulting configuration does not consist of minimum energy surfaces; we
will see that the modulus of a disordered 3D foam (eq. (4.35)) is actually lower
than Gaffine,3D .
Affine deformation of a 2D foam

The 2D version of Derjaguin’s expression for the shear modulus (eq. (4.29)) is easily
derived for an isotropic foam. We assume that the foam consists of a collection of
straight edges whose initial orientations are random and uncorrelated with their
length. The deformation modifies the orientation and the length of each edge affinely
5 Contrast this with the case of spherical bubbles, for which Sint /Vfoam = 3/RV .
182 Rheology
(fig. 4.8), and the shear stress σ̂ can be calculated using Batchelor’s expression
(cf. shaded section on p. 176):
2λ
σ̂ = i cos θi sin θi , (4.31)

Afoam
edges i
where the sum is over all the edges in the sample. Using the geometrical calculation
illustrated in fig. 4.8 we find

2λ
sin3 θi 3λ ε
σ̂ = i sin θi cos θi 1 + ε = i . (4.32)

Afoam cos θi 4Afoam
edges i edges i
The second equality in eq. (4.32) is obtained by averaging over θi , assuming that the
are uniformly distributed in [0, π] and uncorrelated with the edge lengths.6
angles
With edges i i = Lint /2, we deduce that the shear modulus G = σ/ε is
λ Lint
Ĝaffine, 2D = 0.375 . (4.33)
Afoam
As in 3D (eq. (4.30)), the Derjaguin approximation overestimates the shear

modulus.
θ
120°
0 θ
x
Fig. 4.8: Affine shear (with strain ε) of two edges in a 2D foam, of lengths 0 and ,
that initially intersect at 120◦ . The edge 0 , parallel to the direction of shear, remains
unchanged. The length and orientation of become, to leading order in ε: = (1 +
ε cos θ sin θ), sin θ = sin θ (1 − ε cos θ sin θ), and cos θ = cos θ (1 + ε sin3 θ/ cos θ).
3.1.5 Disordered 3D foam: non-affine deformation To estimate the shear

modulus more precisely it is necessary to take into account Plateau’s laws: during a
quasi-static shear, the angles between films and the angle at which edges meet must re-
main constant and equal to 120◦ and 109.5◦ , respectively (cf. p. 30). In the Stamenovic
6 For the same reason, the stress is initially zero (eq. (4.31) with θi = θi ).
α
β
Fig. 4.9: Non-affine deformation of a tetrahedron. The bottom triangle is reduced by

a factor 1 − ε/2 in both directions of the plane, and the total height of the tetrahedron
increases by 1 + ε. The central vertex is placed in the position that preserves the
tetrahedral angles (denoted α). The angle β, which is not the angle between edges, is
not preserved.
model the foam is described as a collection of tetrahedra which are deformed

according to Plateau’s laws (fig. 4.9). A geometric calculation predicts the variation
in film surface area, and therefore the surface energy density, giving [57]
1 γ Sint
Gtet = . (4.34)
6 Vfoam
Expressing Sint in terms of the bubble size (p. 181) gives

γ
Gtet = 0.55 . (4.35)
RV
This is currently the best prediction for disordered monodisperse dry foams, and is in
close agreement with numerical calculations using the Surface Evolver [34] which give
G = 0.51γ/RV .
3.2 The elastic limit of a dry foam

At larger deformations, the response of a foam is no longer elastic and it begins to
flow. We consider here the plastic regime under quasi-static deformation, in which the
bubbles slide past each other in a succession of small jumps. Each of these plastic events
is a T1 topological process (§1.2.2, chap. 3), the mechanical consequences of which will
be discussed. Experiment 6.1 illustrates the T1s which accompany 2D foam flow.
3.2.1 The elastic limit In general, bubbles are not greatly deformed, that is, their
aspect ratio remains close to 1. Consequently, the yield strain εy (defined in §2.1), also
known as the elastic limit, is of order one for a dry foam. In consequence the yield
stress σy , which is approximately G εy if the linear regime extends up to εy , scales in
the same way as the shear modulus:
γ
σy ∼ . (4.36)
d
184 Rheology
Ordered 3D foam A Kelvin foam (p. 50) leaves the elastic regime when the length
of an edge first goes to zero (fig. 4.7). A vertex is then formed at which six edges
meet; this is unstable (by Plateau’s second law, §2.2.2, chap. 2) and the structure
relaxes towards a new equilibrium configuration through a cascade of T1 topological
changes [51, 52]. The strain at this point is the yield strain εy , which depends strongly
on the shear direction because of the anisotropy of the structure [51]. Beyond εy , in
the plastic regime, the stress varies discontinuously with the strain due to the jumps
at each T1.
Ordered 2D foam A more detailed calculation is possible for a hexagonal 2D foam
(fig. 4.10). Under the effect of shear, some edges stretch and others shorten. When the
length of an edge goes to zero, a vertex with four edges forms (fig. 4.10C). This is un-
stable (by Plateau’s first law, §2.2.2, chap. 2): there is a T1, and then the foam relaxes
irreversibly (§3.2.3) to a new equilibrium state (fig. 4.10D). Point C corresponds to
the yield strain. In this case εy can be calculated geometrically (cf. exercise 7.2), and
then we may deduce σ̂y using eq. (4.26). For the shear direction in fig. 4.10, we find
2 λ
εy = √ and σ̂y = √ . (4.37)
3 3h
3.2.2 Energy landscapes, basins of attraction, and hysteresis The example

of a hexagonal 2D foam enables us to illustrate some important concepts in plasticity.
Fig. 4.11a shows how the shear stress σ̂ changes as a function of strain ε (eq. (4.26));
the letters correspond to the different states of the foam shown in fig. 4.10. Fig. 4.11b
shows the energy,7 or energy landscape, of the foam, again as a function of strain ε.
The landscape consists of two (quasi-parabolic) potential wells which partially overlap.
A and D are the stress-free equilibrium states, B is an equilibrium state (p. 180),
and C is an out-of-equilibrium state. If we release the stress at B, the foam relaxes
spontaneously towards the energy minimum at A. Generally, the points situated on
the parabola centred at A return towards A and constitute the basin of attraction
A ε= 0 B ε = 0.5 C ε = εy = 1.15 D ε = ε y = 1.15
θshear
h
Fig. 4.10: Quasi-static shear of a hexagonal 2D foam in the x direction. The edges
meet at 120◦ , except in C where unstable four-fold vertices appear. In D, T1s occur
at each vertex and neighbouring layers are shifted by one bubble in relation to each
other, as the two grey bubbles illustrate.
7 which

can be calculated as E = 6λh + A σ̂ dε, where A is the bubble area.
(a)
0.6 C
0.4
B
0.2
^ /( λ / )
h
A
0 D
σ
–0.2
–0.4
E
–0.6
–1 0 1 εy = 1.15 2
ε
(b)
E C
3.4
3.3
E/(2 λ h )
3.2
3.1 B
A D
3.0
–1 0 1 εy = 1.15 2
ε
Fig. 4.11: Stress and energy in a hexagonal 2D foam. (a) 2D shear stress (eq. (4.26)),
normalized by λ/h , as a function of the strain ε. Points A, B, C, and D (with strains
εA = 0, εB = 0.5, εC = εD = 1.15) correspond to fig. 4.10. Point E is a state with the
same magnitude of elastic strain as C. (b) Energy per bubble, normalized by the
energy 2λh of an undeformed edge.
186 Rheology
of A. State B is in this basin of attraction and would relax towards A, even if εB were
closer to εD than εA . Only a T1 transformation can lead to a change of the basin
of attraction (as from C to D). The abscissa of the graphs (fig. 4.11) is the total
strain of the foam. It coincides with the elastic strain for the points situated on the
parabola centred at A, but not for the parabola centred at D, where the elastic strain
is εe = ε − εy , the plastic part of the strain is εp = εy , and we again have σ = Gεe .
There are several possible structures for a given strain, which leads to the phe-
nomenon of hysteresis. If, having arrived at D, we reduce the applied strain, the foam
does not return via the previous structures. Instead, the structure in D is deformed
without changing the topology until E, at which point T1s occur and the foam returns
to A.
3.2.3 Energy dissipated by a T1 A T1 is always accompanied by dissipation of

mechanical energy. In order to understand this, we return to the transition from state
C to D in a hexagonal foam (fig. 4.10). This transition consists of an instantaneous
change in topology followed by a relaxation. The T1 dissociates the unstable four-
fold vertex into two stable three-fold vertices then, during the ensuing relaxation, the
structure deforms to reach the equilibrium state D. The relaxation happens without
the boundaries of the system moving, and thus without exchange of work with the
outside. The difference in energy between C (just at the yield strain) and D is thus
entirely dissipated as heat during the relaxation. In this hexagonal case we have ΔE =
EC − ED ≈ 0.46 × 2λ h ≈ Ebub /6, where Ebub = 6λh is the energy per bubble; for a
bubble of edge-length h and a foam of thickness e, both of the order of a millimetre,
and a surface tension 25 mN.m−1 we thus find ΔE ∼ 3 × 10−8 J.
The amount of energy dissipated per bubble is fixed by the energy accumulated
during the elastic loading, which itself depends only on the equilibrium properties of
the foam (see exercise 7.5). Thus the dissipation, per unit volume and for a strain
of fixed amplitude ε, doesn’t tend towards zero in the limit of small strain rates. It
is proportional to the number of T1s (and thus, on average, to ε) and independent
of ε̇. We can therefore calculate the dissipation per unit volume and per unit time,
i.e. the dissipation rate. In the quasi-static regime, summarized in the shaded section
below, this rate is proportional to ε̇ as in solid friction, and not to ε̇2 as in viscous
friction or, more generally, as in any of the dissipative modes at the scale of the films
and interfaces, which will be described later (§3.3.2). This is because we control the
macroscopic deformation rate but not the relaxation time (and therefore the local
strain rate) of the foam between C and D. This time, which is intrinsic to the foam,
depends on local modes of dissipation.
Quasi-static flow
The following is a summary of the conditions necessary for quasi-static flow:
• the foam must be at mechanical equilibrium at each instant, except during the
brief relaxation after a T1;
• the relaxation time must be negligible compared to the characteristic time of

the macroscopic deformation (1/ε̇);
• the new equilibrium structure reached after a T1 must not depend on
dissipation at the local scale (films and interfaces).
Under these conditions, the succession of structures adopted by the foam is in-
dependent of the applied strain rate and, in this sense, the notion of time is no
longer relevant. Note that in a macroscopic sample, consisting of a large number
of bubbles, there is almost always a relaxation taking place: the first condition is
thus rarely strictly respected. In order to resolve this difficulty, it is necessary to
introduce the notion (as yet poorly defined) of the range of a T1, cf. fig. 3.5 [14],
and to determine the extent to which T1s “overlap”.
During rapid flow, many T1s occur and a foam never reaches an equilibrium state.
The imposed strain, which tends to elongate the bubbles, dominates the relaxation
after a T1, which tends to reduce the bubble deformation. The yield strain thus
increases with the strain rate [53] (§4.2.2).
3.3 Dissipative processes

As we have just seen, liquid foams are only capable of accumulating a finite amount of
energy: the elastic energy density (eq. (4.2)) is at most of the order of σy2 /G ∼ γ/d, i.e.
several J.m−3 for bubbles of diameter d ∼ 1 mm. The kinetic energy per unit volume
of foam is of the order of ρl φl v 2 —i.e. around 1 J.m−3 for a liquid fraction of φl = 10%
and a speed v of about 10 cm.s−1 . The additional energy supplied to the system, in
order to impose a steady flow for example, is continually dissipated as heat. In the
quasi-static limit, it was possible to predict the amount of energy dissipated (§3.2.3);
here, we discuss the way in which it is dissipated.
The processes that dissipate energy in the rapid flow regime or during the fast
relaxation following a T1 in a quasi-static flow are the same. The dissipation modes
can be grouped into two categories: the hydrodynamic dissipation in the liquid phase
due to the liquid shear viscosity; and the dissipation associated with surfactant motion
due to the interfacial viscosity of the monolayer or the diffusivity of the surfactants [6].
The coupling between these different modes is complex and there is currently no
simple and general criterion to identify the dominant modes in any given situation.
A weakly dissipative mode contributes little to the production of heat, while a strongly
dissipative mode is probably not stimulated and does not contribute either, in which
case we speak of a frozen mode. In the following we extend the discussion of §6.2,
chap. 3 to give for each process the order of magnitude of the dissipative power per
unit volume of foam, denoted D, as if they were independent of each other.
3.3.1 Dissipation in the liquid The viscous dissipation D in a Newtonian li-

quid subject to a local shear rate ε˙l is η ε˙l 2 , where η is the viscosity, of the order of
10−3 Pa.s for foaming solutions. The dissipation in a foam is difficult to evaluate, as
188 Rheology
the macroscopic shear rate ε̇, determined by the movement of the boundaries at the
scale of several bubbles, is not directly related to the velocity gradients in the liquid
phase. The solution is sheared locally, over the thickness of the films and over the
cross-section of Plateau borders, and in general ε˙l is different from ε̇. Moreover, even
if the exact shape of the bubbles (itself difficult to measure in situ in the flow) were
known precisely at each moment, the problem still would not be resolved since for a
given evolution of the structure, there may be several possible velocity fields in the
liquid.
For a small strain of amplitude ε, in the absence of topological changes, the relative
change in surface area ΔS/S of an individual film (which may be positive or negative)
is proportional to ε. For the whole foam, since those films which shrink and those
which expand almost compensate each other, we have ΔSint /Sint ∼ ε2 (p. 180). We
disregard this global expansion.
We first distinguish two modes of deformation at the local scale, assuming affine
flow of the bubbles:
Mode A If an adsorbed surfactant monolayer is incompressible, for example if the
dilatational surface modulus Es∗ is high (cf. eq. (3.13)), strong surface stresses make
it impossible to create or destroy interface. This is known as a rigid or immobile
interface. These interfaces behave like flexible inextensible membranes, moving at the
speed of the bubble that they enclose and sliding past one another (fig. 4.12A). The
relative speed of two interfaces situated on either side of a film is thus of the order of
the difference in velocity between two bubbles in contact, ε̇ d. The velocity gradient
in a film of thickness h is therefore of the order of ε̇d/h. In a volume d3 (the bubble
size), the volume of sheared liquid is hd2 , which implies that the dissipation rate per
unit volume of foam is
2 2
ε̇d hd d
Dhydro,A ∼ η ∼ η ε̇2 . (4.38)
h d3 h
Mode B In contrast, for a perfectly compressible interface the surface stress is due
only to the equilibrium surface tension γ. This is known as a fluid or mobile interface,8
which can deform as represented in fig. 4.12B. The film length increases (or decreases)
at a rate dε̇ inducing velocity gradients of the order of ε̇ in a film of volume d2 h. The
dissipation rate per unit volume of foam is then
d2 h h
Dhydro,B ∼ η ε̇2 ∼ η ε̇2 . (4.39)
d3 d
Mode B induces less hydrodynamic resistance to the applied strain than mode A,
but it also requires that a significant amount of interface is created on one side of a
Plateau border and destroyed on the other. It is therefore associated with a transport
of surfactants from within the film, which also induces dissipation. Mode A, on the
8 In the case of drainage (p. 119), interfaces are classified as mobile or immobile depending on
the interfacial shear viscosity (or, more precisely, the Boussinesq number Bo). Here, the mobility or
rigidity of an interface refers to its response to compression and expansion; we thus prefer to refer to
perfectly compressible or incompressible interfaces.
− + −
− + −
Fig. 4.12: Movement of the gas/liquid interfaces in a sheared 2D foam. The interfaces
are notionally separated into different parts represented by the shading, to enable
a schematic visualization of local deformations and displacements during shearing.
State 0 is the reference state: it can be deformed according to either mode A or B.
In mode A, the length of each part of the interfaces is fixed (to first order in ε) and
they slide over one another, which induces velocity gradients in the films. In mode B,
the interfaces change length (according to the sign + or –), the interfaces on each
side of a film undergo the same displacements, and the velocity gradient in each film
is zero.
other hand, does not induce surfactant-related dissipation. In practice, it is therefore

the competition between hydrodynamic flows and surfactant transport which dictates
the flow.
The expressions for the dissipation (eqs. (4.38) and (4.39)) correspond to rather
unrealistic velocity fields, but illustrate the coupling between the liquid and the sur-
factant. In particular, they don’t take into account the fact that film thickness h itself
depends on the flow, and can vary over several orders of magnitude. Moreover, thick-
ness variations are coupled to strong heterogeneities in velocity gradients within the
190 Rheology
same film. Schwartz and Princen [56], taking this effect into account, have sugges-
ted that the dissipation comes mainly from the movement of liquid from the Plateau
border into the film. They predict that the liquid is entrained over a distance of the
order of x∗ ∼ rCa1/3 , with a film of thickness h∗ ∼ rCa2/3 , where r is the radius of
curvature of the Plateau border, and Ca = η ε̇d/γ is the capillary number, which com-
pares viscous to capillary effects. For a bubble of volume d3 , the sheared volume is
thus x∗ h∗ d, with a shear rate ε̇d/h∗ . The bulk dissipation rate is then
2 ∗ ∗ −1/3
ε̇d x h d −1/3 η ε̇d
DS.P. ∼ η ∼ η ε̇ Ca
2
∼ η ε̇2
. (4.40)
h∗ d3 γ
This process predicts that the shear stress is a non-linear function of the strain rate,
σ ∼ ε̇2/3 , which contradicts the linear response observed macroscopically in the limit
of small strains.
3.3.2 Dissipation associated with surfactant motion We next determine

the dissipation associated with the transport of surfactants and the viscosity of a
monolayer.
Shear The upper and lower films in fig. 4.12 (mode A or B) don’t change in area,
but are subject to the global shear rate ε̇. The surface area per unit volume is of the
order of 1/d, so the dissipation associated with shear can be estimated from a scaling
argument:
ηs
Dshear ∼ ε̇2 , (4.41)
d
where ηs is the surface shear viscosity defined in §2.2.3, chap. 3.
Expansion and compression Expansion and compression are important in mode
B of fig. 4.12, where the interfaces undergo relative changes in area at a rate of
the order of ε̇. The surface area per unit volume is again of the order of 1/d, so the
dissipation is
ηd,film 2
Dexp/comp ∼ ε̇ , (4.42)
d
where ηd,film is the dilatational viscosity of the film. As discussed in §2.2.3, chap. 3, a
film may expand in different ways, associated with the exchange of surfactants between
the interface and film. These different modes and the corresponding dissipation are
explained in fig. 3.54, §6.2, chap. 3. Comparing those expressions (eqs. (3.105), (3.106),
and (3.107)) with eq. (4.42) allows the effective viscosity ηd,film corresponding to each
mode to be estimated.
Other models, based on surfactant-linked dissipation, are also proposed in the
literature [4, 5, 20].
3.3.3 Dissipation at the wall Lastly, we investigate the dissipation due to the
wetting film which is present between a foam and a wall. Two limiting cases arise, de-
pending on whether the gas/liquid interface is perfectly compressible or incompressible,
as defined in the footnote in §3.3.1 [8, 16, 58]:
Perfectly compressible interface In this case, a surface Plateau border slides

along the wall without causing the wetting film to move; the bubble “rolls without
sliding” (fig. 4.13). The velocity gradient is localized in the region where the surface
Plateau border and the wetting film meet (i.e. the region where the liquid is expelled
or absorbed by the Plateau border). Although the characteristic size of this region
increases with the velocity, the following argument, analogous to that of Schwartz and
Princen (p. 190), shows that the dissipation is not a quadratic function of the imposed
shear rate.
Consider a surface Plateau border (with radius of curvature r), which slides at speed
U ex along a wall.9 The flow is quasi-parallel, and a lubrication analysis is appropriate
[1]. The velocity is zero at the wall and the tangential stress is zero at the interface
(because we are assuming that the surfactants are mobile). The pressure depends
only on x, and is directly related to the curvature by the Young–Laplace law: p(x) =
P − γ(∂ 2 h/∂x2 ). The pressure gradient induces a parabolic velocity field in which
the flux per unit length is Q ∼ ∇p h3 /η ∼ (γ/η) (∂ 3 h/∂x3 )h3 . Conservation of mass
requires that ∂Q/∂x = ∂h/∂t. Finally, at steady state the space and time derivatives
are linked by ∂h/∂t = −U ∂h/∂x and so
3

∂h γ ∂ 3 ∂ h
U ∼− h . (4.43)
∂x η ∂x ∂x3
We can determine the properties of h(x) without solving this differential equation,
but simply by examining the orders of magnitude of its terms. The thickness h is
negligible and constant in the wetting film, and has constant curvature 1/r in the
surface Plateau border. The distance over which we go from one to the other is x∗ ,
the length of the transition zone. In this zone, h(x) is of the order of h∗ , and its
nth derivatives are of order h∗ /x∗n . We must determine h∗ and x∗ . Eq. (4.43) gives
U r
u =U
Transition
zone
u =0 h*
x
x*
Fig. 4.13: A surface Plateau border sliding along a wall in the case of a perfectly
compressible interface. The arrows represent the local velocity field u in the liquid.
The surface Plateau border—of which only half is shown—moves with velocity U ; the
wetting film is at rest on the wall (y ≤ 0). There is a strongly dissipative region where
the wetting film meets the Plateau border, known as the transition zone, of thickness
h∗ and length x∗ . The flow is assumed invariant along the axis of the Plateau border
(in the z direction, perpendicular to the plane of the figure).
9 The problem is analogous to that of a plate withdrawn from a reservoir of liquid [15].
192 Rheology
the relationship h∗ /x∗ ∼ (ηU/γ)1/3 . Expressed as a function of the capillary number

Ca = ηU/γ, this is h∗ /x∗ ∼ Ca1/3 .
On the other hand, the continuity of curvature between the transition zone and
the surface Plateau border dictates that h∗ /x∗2 ∼ 1/r. Thus h∗ ∼ rCa2/3 and x∗ ∼
rCa1/3 . From this we may deduce the order of magnitude of the dissipation associated
with the motion of a surface Plateau border, which is localized in the transition zone.
The velocity gradients are ∂u/∂y ∼ U/h∗ and the volume of the zone is Ω ∼ h∗ x∗ d
for a surface Plateau border of length d. Then the dissipation per unit surface area of
foam in contact with the wall is
2
1 ∂u η U 2 Ca−1/3
Dwall, m ∼ 2 η dΩ ∼ . (4.44)
d Ω ∂y d
So in the case of a perfectly compressible interface the effective viscosity of a foam

sliding over a solid wall decreases with velocity as Ca−1/3 . This is because the faster
the foam goes, the more the films inflate, facilitating the slippage. Nevertheless, the
dissipation increases with the speed. A similar reasoning leads to a (dimensional)
expression for the shear stress on the wall:

1 ∂u γCa2/3
σwall, m ∼ 2 η dS ∼ . (4.45)
d x∗ d ∂y d
Incompressible interface If the monolayer is incompressible, then interface is

neither created nor destroyed. The bubbles undergo uniform translation (the bubble
“slides without rolling”) and the gas/liquid interface slides at the speed of the bubble.
This increases the velocity gradient in the wetting film (which has characteristic size
d2 ): the side of the wetting film in contact with the wall is stationary, while the other
side moves at the speed of the bubble, so that the dominant contribution to the hy-
drodynamic dissipation comes from the shear of the wetting film. The flow in the
film induces an overpressure which leads to an increase in film thickness10 and leads
once again to a non-quadratic dependence of the dissipation on velocity [17]. The
experimental data are well fitted by the expression
ηU 2 Ca−1/2
Dwall, i ∼ . (4.46)
d
In this case the stress at the wall is:
γCa1/2
σwall, i ∼ . (4.47)
d
Fig. 4.14 shows experimental results that illustrate both behaviours [16]. The
missing prefactors in eqs. (4.45) and (4.47) depend strongly on the liquid fraction and
on the nature of the surfactants, but not on the viscosity of the solution, which is
accounted for in the capillary number.
10 This is the principle of aquaplaning.
The multiscale character of foam rheology 193
10 –1
0.5 wt % K Myristate
+ 0.25 wt % K Palmitate
+ 0.25 wt % K Laurate
σwall / (γ/d)
10 –2
m = 1/2
1 wt % Na Laurate
+ 2 wt % K Cocoyl Glycinate
+ 3 wt % Betaine
m = 2/3
10 –3
10 –6 10 –5 10 –4 10 –3
Capillary number
Fig. 4.14: Wall stress as a function of capillary number for two different foaming solu-
tions. The different symbols for each surfactant mixture correspond to different glycerol
concentrations. Solutions with perfectly compressible interfaces exhibit a power law
with exponent m = 2/3 (lower curves, eq. (4.45)) and solutions with incompressible
interfaces exhibit a power law with exponent m = 1/2 (upper curves, eq. (4.47)).
The wall stress for a given surfactant mixture barely depends on the glycerol concen-
tration, indicating that the role of solution viscosity is correctly represented in the
predictions [16].
4 The multiscale character of foam rheology

A foam behaves as a solid or as a liquid depending on the liquid fraction, the applied
stress, or the typical time-scale of a perturbation (fig. 4.15). As we saw in §3, elasticity
is due to the surface tension of the gas/liquid interfaces, and plasticity is associated
Applied
stress
Liquid
Liquid
0 fraction
Solid
Time
Fig. 4.15: Schematic diagram of the mechanical behaviour of a foam as a function of

the applied stress, the liquid fraction, and the characteristic time-scale of a mechanical
perturbation.
194 Rheology
with T1s. During flow, the dissipation of mechanical energy is dominated by creep or
plasticity at low shear rates (quasi-static regime), and by (surface or liquid) viscosity
at high shear rates. We now look at how the mechanisms acting at the scale of films
and bubbles influence the elastic, plastic or viscous behaviour of “real” foams, which
have a finite liquid fraction and are often polydisperse.
4.1 Solid behaviour

4.1.1 Static linear elasticity We have seen (§3.1) that dry foams, whether
ordered or not, become softer (lower shear modulus) as the bubbles become larger,
in both 2D and 3D. We now turn to the shear modulus of wet polydisperse foams, and
the non-linear elastic response at large strains, for example due to normal stresses.
Effect of polydispersity The shear modulus G of a monodisperse foam depends
on the inverse of the bubble radius (§3.1.1). Which characteristic length determines G
in a polydisperse foam? The average bubble radius RV (with RV = (3V /4π)1/3 ) is
not always the best choice. This is clear in the case of a bidisperse foam constructed by
replacing each vertex of a monodisperse foam with a small three-sided bubble (in 2D)
or a tetrahedron (in 3D): this construction significantly changes the average bubble
size but does not modify the structure of the surrounding foam.11 Shearing this foam
barely deforms the small bubbles (see exercise 7.2), and so the shear modulus of this
decorated foam is close to that of a foam consisting only of large bubbles. For this
special case the right length-scale needed to estimate the shear modulus is hence the
size of the large bubbles.
For real disordered foams, there is no clear rule. When a foam is sheared, the
interfacial area increases in proportion to Sint , the area of the interfaces at rest. In
dimensional terms, the surface energy density E varies in proportion to the surface
to volume ratio RV2 /RV3 , and we know that E = Gε2 (eq. (4.2)). We therefore
introduce the Sauter mean radius, R32 = RV3 /RV2 (see §3.2.2, chap. 5), which is
Fig. 4.16: Small tetrahedral bubble decorating a vertex of a large bubble. Photograph
courtesy of V. Labiausse.
11 by analogy with the 2D decoration theorem (§5.3, chap. 2).

often about 10% to 20% greater than RV for real foams (fig. 5.19 and eq. (2.19)).
Then for a disordered 3D foam G ≈ γ/R32 . The 2D case is dealt with in more detail
in reference [37] and exercise 7.2.
Effect of liquid fraction A dry foam is stiffer than a wet foam for a given bubble
size. When the liquid fraction increases, G decreases, and finally goes to zero at the
volume fraction of close packing, denoted φ∗l . In fact, at this point, the contacts between
bubbles disappear since the bubbles are spherical (fig. 4.17 gives the value of φ∗l
for different structures). Moreover, at φ∗l , under the effect of shear, the bubbles re-
main spherical, the interfacial energy is constant, and no restoring force opposes the
deformation: the foam loses its elasticity.
By reasoning in 2D, we can understand how G decreases with φl . We consider a
hexagonal foam “decorated” with Plateau borders (fig. 4.18). Batchelor’s expression
Structure ϕl∗
π
Ordered 2D (hexagonal) 1– 0.09
2 3
Disordered 2D 0.16
Ordered 3D 1– π
(face-centred cubic) 0.26
3 2
Disordered 3D 0.36
Fig. 4.17: Liquid fractions at which monodisperse foams lose their elasticity. For dis-
ordered structures, φ∗l corresponds to the density of a random close packing of solid
discs (in 2D) or spheres (in 3D).
(a) (b)
Fig. 4.18: A hexagonal 2D foam “decorated” at a liquid fraction of a few percent.

(a) At rest. (b) After shear: unlike the films, the Plateau borders are not affected by
the shear.
196 Rheology
for the shear stress due to the interfaces (eq. (4.18)) consists of two contributions:
one associated with the films (the contact zones between bubbles) and one with the
Plateau borders. However, in this case the latter contribution is zero because of their
three-fold symmetry (which is preserved during shear, see exercise 7.2). Consequently,
only the films contribute to the shear modulus. As φl increases, the bubble shape
changes from a hexagon to a disc, and the film length decreases whilst the Plateau
border areas increase (§4.1, chap. 2). From this we deduce that G decreases with φl .
For a hexagonal wet foam (fig. 4.18), Princen [49] gave a geometric argument to
show that the variation in the modulus with liquid fraction is
λ
for φl < φ∗l ,
1
Ĝhex = 0.52 (1 − φl ) 2 (4.48)
RA
where RA is the equivalent circular radius (RA = (A/π)1/2 for a bubble of area A),
and λ is the line tension. As expected, the modulus decreases as φl increases. At
φl = φ∗l all the contacts between bubbles disappear and Ĝ drops discontinuously to
zero (fig. 4.19).
For a disordered 2D foam, the loss of contacts is a statistical phenomenon: the foam
continuously loses its elasticity with increasing φl and G eventually vanishes at φ∗l ,
when the network of contacts between bubbles is no longer capable of transmitting a
static stress [21, 61]. Likewise, for disordered 3D foams, G decreases continuously until
φ∗l . Experimental observations [46] of the variation of G with φl and with the Sauter
mean radius R32 are well described by the phenomenological expression (fig. 4.19)
γ
G = 1.4 (1 − φl )(φ∗l − φl ) . (4.49)
R32
200 200
Hexagonal 2D structure
150 Disordered 3D structure 150
G (N/m)
G (Pa)
100 100
^
50 50
0 0
0 0.1 0.2 0.3 0.4
l
Fig. 4.19: Shear modulus G as a function of liquid fraction φl for a disordered 3D

foam (R32 = 66 μm and γ = 20 mN.m−1 , data from [54]). The continuous line is
eq. (4.49), and the dotted line is for a 2D hexagonal foam (eq. (4.48)) calculated with
RA = 66 μm and λ = 20 mN (right-hand axis).
This expression is in good agreement with numerical simulations for dry 3D foams [34]
(φl = 0 and φ∗l = 0.36) which give G = 0.51γ/R32 .
Physical origin of the rigidity loss at φ∗l In the vicinity of φ∗l , the coupling
between strain and surface energy depends on the average number Zc of contacts
between bubbles, and on the force required to bring the bubbles together (the “contact
force”), opposing the repulsion between bubbles as they seek to reduce their surface
area and be as round as possible. To make this simpler, we describe this repulsion
schematically using a model of a compressed spring in the shaded section below. The
spring constant k and Zc decrease when the liquid fraction (and hence the distance
between bubble centres) increases. In an ordered 2D foam, the springs are harmonic
and of constant stiffness k ∼ γ/RA . As all the contacts disappear simultaneously at φ∗l
(Zc jumps from 6 to 0), the modulus is discontinuous at this liquid fraction (fig. 4.19).
In an ordered 3D foam, the contacts also disappear at φ∗l , and G goes to zero, but
we expect a continuous variation in G because of the anharmonic character of the
capillary forces; indeed, k depends on the distance between bubbles and goes to zero
as φl tends towards φ∗l .
Interaction between neighbouring bubbles in a foam

When φl < φ∗l , neighbouring bubbles repel each other because of the surface energy
required to deform them. This is really an interaction between many bubbles,
but to understand it at a given liquid fraction,12 we take an initially circular 2D
bubble of radius RA and squeeze it between two parallel planes (to imitate the
contact between bubbles) whilst conserving its area (fig. 4.20). When equilibrium
is reached, the perimeter of the bubble is again minimal, implying that the two
ends are circular arcs with radius L; if L is the length of contact with each plane
(exaggerated in the figure), we have to leading order (δ/RA 1) L L ∼ δRA .
Consequently, the force required is F ∼ (λ/L)L ∼ λδ/RA , with λ the line tension.
Close to φ∗l therefore, the repulsive force between neighbouring bubbles can be
represented as a harmonic spring with stiffness k ∼ λ/RA . Here again, we observe
that a foam is stiffer when the bubbles are smaller.
F δ = RA− L
2L
L
Fig. 4.20: A 2D circular bubble of initial radius RA squeezed between two planes,
keeping the area constant, with a force per unit length F .
12 The changes in surface energy associated with a change in liquid fraction are described by the
osmotic pressure of the foam in §4.2, chap. 2.
198 Rheology
In 3D, an initially spherical bubble squeezed between two planes adjusts its
two radii of curvature in order to minimize its surface area at constant volume. A
numerical calculation shows that for small deformations (δ/RV 1), the energy
(per unit surface area) required to trap the bubble is [39] ΔE ∼ γ(δ/RV )α where
α > 2 (α = 2.5 for a monodisperse wet foam with a face-centred cubic structure).
Thus in 3D, the contact between two bubbles is described by anharmonic springs
with stiffness k ∼ γ(δ/RV )α−2 that goes to zero at φ∗l , the liquid fraction at which
δ = 0.
4.1.2 Static non-linear elasticity A shear deformation close to the yield strain
but insufficient to trigger T1s induces stretches and rotations of films which creates nor-
mal stress differences (p. 175). In general, the first normal stress difference (eq. (4.17))
induced by a quasi-static shear in an isotropic elastic material obeys Poynting’s law
(exercise 7.3):
N1 = σ ε = G ε2 . (4.50)
In a sheared foam, the soap films are stretched and oriented in a preferential direction
(fig. 4.21a). When the stress σ = σxy is large, the film normals tend to be aligned
in the y direction. Consequently, the normal stresses σxx and σzz , due to the surface
tension of the films, are large compared to σyy . As the x and z directions are equivalent
in this respect, we have σxx ≈ σzz , and we expect that qualitatively [40] N1 ≈ σxx and
N2 ≈ −σzz ≈ −N1 . This effect is different from that observed for polymers, in which
the chains tend to be aligned only in the direction of the applied shear (fig. 4.21b),
so that σxx is large compared to σyy and σzz , and N1 ≈ σxx and N2 ≈ 0.
A model based on individual films, generalizing the approach of Derjaguin (p. 181),
leads to the following tensorial constitutive law, valid as long as the yield strain is not
exceeded [30]:
G −1
σij = −p δij + (Bij − 6Bij ). (4.51)
7
In this (Mooney–Rivlin) law, finite deformations are described by the Finger tensor
Bij (eq. (4.13)). It predicts a linear relationship between shear stress and strain, ac-
companied by a non-linear variation of normal stresses. For a finite applied shear, the
expression for Bij is given in exercise 7.3, and eq. (4.51) predicts
6
σ = Gε N1 = Gε2 N2 = − Gε2 . (4.52)
7
As expected, N2 is of the same order of magnitude as N1 and of opposite sign. The
quadratic dependence of N1 on ε is confirmed in numerical simulations (fig. 4.22a)
and experimentally up to the yield strain (fig. 4.23). Moreover, eq. (4.52) predicts
that the normal stress difference varies in proportion to the elastic shear modulus
G. Consequently, in foams which are soft and have a low yield strain, N1 is always
(a) y y
x x
z z
(b)
y y
x x
z z
Fig. 4.21: Shear applied to a cube containing elastic elements that are randomly
oriented in space and represented by: (a) facets which represent the films in a foam;
(b) strands which represent polymer chains. Under the effect of shear, the facets tend
to be aligned parallel to the (x, z) plane and the strands tend to align in the x direction.
small compared to the shear stress. Typically for G ≈ 100 Pa and ε ≈ 0.1, we have
N1 ≈ 1 Pa! Polymers, which are more elastic (with G of the order of MPa) and can
undergo strains ε of order one (think of rubber), show much greater normal stress
differences, since we can have N1 ≈ σ. Finally, we note that these normal stresses also
induce an effect in steady shear flow: the Weissenberg effect, observed in polymeric
liquids and, to a lesser extent, in foams, in which the fluid climbs the rotating rod
driving the flow.
4.1.3 Linear viscoelasticity

Frequency response In §3 we identified the different modes of dissipation asso-
ciated with bubble motion, either with or without topological changes (T1s): the
viscosity of the solution, interfacial viscosity, and surfactant transport. In the limit of
small strains, much lower than the yield strain, they cause a foam to behave as a linear
viscoelastic medium whose response depends on the time-scale of the perturbation, of
the order of 1/ω for an oscillatory experiment with frequency ω. The response is char-
acterized by the complex shear modulus G∗ (ω) (eq. (4.9)), and the variations of G∗ (ω)
are similar to those observed in other disordered soft materials such as concentrated
emulsions and pastes. They show a large spectrum of relaxation times, associated with
200 Rheology
(a) 0.4 (b)

σ
y
Normalised stress
0.2
N1
N2
–0.2
0 εy 0.5 1.0 1.5
Strain ε
Fig. 4.22: (a) Shear stress and first and second normal stress differences as a func-
tion of the (quasi-static) strain for a monodisperse, disordered, dry 3D foam (Surface
Evolver simulation [34]). The stresses are normalized by γ/V 1/3 , where V is the bubble
volume. Below the yield strain εy , we observe that σ is linear in ε, while N1 and N2
vary as ε2 (eq. (4.52)). Beyond εy , the stress fluctuates around the yield stress σy due
to T1s (see §5). (b) The structure of a foam of 216 bubbles at a strain of ε = 1.4.
Image courtesy of A. Kraynik.
10 100
1
N1(Pa)
G'(Pa)
0.1
ϕ1 = 0.03
Rv = 80 μm
0.01 2 γ = 27 mN/m
0.001 10
0.01 0.1 1
Amplitude ε0
Fig. 4.23: First normal stress difference N1 (•) and shear modulus G (◦) as a function
of the amplitude of the applied strain ε0 (the yield strain is εy ≈ 0.4). N1 and G are
measured by an oscillatory method at a frequency of 2 Hz (cf. fig. 5.11). G is constant
if the amplitude is much lower than the yield strain εy , then decreases beyond that.
N1 is quadratic in ε (eq. (4.52)), except at very low amplitudes where the deviation is
due to residual trapped stresses (cf. p. 177). Data from [38].
10 3
G'
Shear modulus (Pa)
10 2
1/2
G''
10 1
10 –4 10 –3 10 –2 10 –1 10 0 10 1 10 2
Frequency f (Hz)
Fig. 4.24: Complex shear modulus of a foam (eq. (4.9)) as a function of frequency,
in the linear regime. The sample is a shaving foam that has been left to coarsen for
100 min, with φl = 0.08. The dashed curves are the Laplace transform of the creep
compliance (eq. (4.54)) whilst the straight line of slope 1/2 is eq. (4.53). Data from
[11, 12, 27].
different dissipative mechanisms, which have been measured over approximately six
decades in frequency. Fig. 4.24 shows that G grows monotonically (with a plateau at
intermediate frequencies) while G goes through a maximum, then a minimum, then
a power law increase.13 We next discuss these different responses one-by-one.
Very rapid relaxations In order to estimate the shortest relaxation time, we consider
an individual soap bubble in air. A bubble of size d behaves roughly like a spring with
stiffness G ∼ γ/d (cf. the shaded section on p. 197). If the movement is restricted by
the liquid flow in the film, the effective viscosity is ηeff ∼ ηd/h (eq. (4.38)), where h
is the thickness of the soap film and η the viscosity of the solution (cf. §3.3.1). The
relaxation time is τ ∼ ηeff /G (cf. fig. 4.2), i.e. τ ∼ ηd2 /γh; it is of the order of 10−6 to
10−3 s for γ = 30 mN.m−1 , η = 1 mPa.s, d = 30 μm and h between 20 nm and 1 μm,
which corresponds to a characteristic frequency f = 1/(2πτ ) of 102 to 105 Hz. Such a
relaxation cannot be observed in the measurement window of fig. 4.24.
Rapid relaxations At high frequency (between 1 and 102 Hz on fig. 4.24), we can fit
the data by
√
G∗ (ω) = G(1 + α iω), (4.53)
13 Note that G does not give information about the viscosity of a foam in steady flow, which we
will discuss in §4.3.1.
202 Rheology
where G is the static elastic (shear) modulus (of the order of 250 Pa in fig. 4.24).
The dependence of G∗ on ω 1/2 has been suggested to be a consequence of collective
processes at the scale of a few bubbles; that is, the structural disorder creates some
weak, anisotropic regions, somewhat like cracks in a continuous medium. During shear,
the bubbles slide past the weak regions and energy is dissipated by viscous friction. The
random orientation of these regions causes a distribution of relaxation times, which
leads to square-root dependence [42].
A very different process, due to the interfacial dilatational elasticity Es∗ (§2.2.3,
chap. 3), could also contribute to the dissipation. In fact, dimensionally, G∗ varies as
Es∗ /d. If we assume that the interfacial elasticity is controlled by the bulk diffusion of
surfactants, Es∗ is given by eqs. (3.17) and (3.18), and, if the period of the deforma-
tion is sufficiently long √that the surfactants have time to diffuse and repopulate the
interface, we find Es∗ ∼ iω. We then presume that this contribution is added to the
static modulus G to give eq. (4.53) [4].
Elastic plateau At intermediate frequencies (10−3 to 1 Hz), we observe in fig. 4.24

a plateau in the storage modulus G and a minimum in G . In this regime, elasticity
dominates the response and G corresponds to the static shear modulus G (p. 178). In
describing static elasticity (§4.1.1), we have implicitly assumed that this regime fully
describes the low-frequency response.
Slow relaxations The quasi-static regime (seen in §4.1.1) does not apply at very low
frequencies (f < 10−3 Hz). Here G tends towards zero and G reaches a maximum
before tending towards zero. This maximum of G is the signature of a process in
which the characteristic time τ = 1/(2πf ) (which is around 300 seconds in fig. 4.24)
is linked to coarsening. At these long times, the foam coarsens during each oscillation,
and it is therefore difficult to decouple the ageing from the relaxation.14 A better way
to probe the long-time response is to perform a creep test (cf. p. 172) as we will see
in the next paragraph.
Time response In the linear regime, the frequency and time responses are equi-
valent (§2.2). Consequently, a creep test is an appropriate way to examine the slow
relaxations of a foam as it coarsens: it can be short enough to give information about
the viscoelasticity at a specific foam age. Recall that this test probes the linear re-
sponse by applying to the sample a step stress σ lower than the yield stress, so as not to
trigger T1s (§2.2). The strain ε(t) is then measured over a sufficiently short time that
the properties of the foam (bubble size, elastic modulus, rate of coarsening-induced
T1s) remain constant, but long enough that coarsening induces a significant number
of T1s.
Fig. 4.25a shows the compliance J(t) = ε(t)/σ measured for three foams with
different coarsening rates in response to a step stress applied for Δt = 100 s. In all
three cases, J(t) exhibits an instantaneous compliance J(0) = 1/G. Then we observe
14 We presume that the effects of drainage are negligible, which a comparison of the orders of
magnitude (eq. (2.48)) suggests is possible in practice for a sample of this height over these time-
scales, cf. fig. 2.23. Coarsening can be considerably slowed down by adding insoluble gases, see
fig. 3.27 and §3.3.4, chap. 3.
(a)
σ (t)
2
104
A C
103
τ0(s)
B
B A
Normalised compliance J0(t)/Je0
102 1
C 103 104 105

1 Time (s)
τ1
A
Δt B
C
0
0 100 200 300

Time (s)
η1
(b)
η0
G
G1
Fig. 4.25: (a) Linear creep response of a foam subjected to a constant stress σ σy
for Δt = 100 s. The three foams A, B, and C (with φl = 0.07) are made with gases of
different solubilities, in such a way that the coarsening rate, and therefore the rate of
topological changes induced by coarsening, decreases from A to C. The compliance is
normalized by Je0 = (1/G + 1/G1 ) (eq. (4.54)). The inset shows the time τ0 = η0 Je0
deduced from the experiment as a function of the average time T between T1s in
a volume (2RV )3 measured independently by multiple light scattering (see §1.2.5,
chap. 5). Data from [12]. (b) Diagrammatical representation of the behaviour observed
in (a).
204 Rheology
a transient relaxation lasting for a time τ1 , of the order of a few seconds. The flow
then reaches a steady state in which J(t) increases linearly with time, more quickly if
the foam coarsens quickly. Finally, at t = Δt, the stress is brought to zero. The foam
recovers the elastic part of the strain and retains a residual strain due to the flow. In
the time interval 0 < t < Δt, the compliance behaves as
1 t 1 − exp(−tG1 /η1 )
J(t) = + + . (4.54)
G η0 G1
A Laplace transform of J(t) gives the complex modulus G∗ (ω) at low frequency
(p. 173). This is in good agreement with oscillatory experiments, as in fig. 4.24,
which is a signature of the linearity of the response. More specifically, J(t) may be
represented by a Maxwell liquid (G, η0 ) and a Kelvin–Voigt solid (G1 , η1 ) in series
(fig. 4.25b). In the stationary regime (over a long time), the foam flows slowly with
a viscosity η0 such that
t
J(t) = Je0 + (4.55)
η0
with Je0 = (1/G + 1/G1 ). In this way the creep test generates two relaxation processes,
with characteristic times τ1 = η1 /G1 for the transient process (τ1 is of the order of a
few seconds in fig. 4.25a), and τ0 = η0 Je0 ≈ η0 /G for steady creep (τ0 is typically a
few minutes in fig. 4.25a). The parameters G, G1 , η0 , and η1 depend on the physico-
chemical characteristics of the foam. Experimentally, we find that: (1) G ∝ γ/RV , in
agreement with eq. (4.49); (2) G1 ∝ γ/RV , but with a larger prefactor (G1 G);
(3) steady flow slows down (η0 increases) when coarsening slows; (4) η1 ∝ 1/RV .
Although η0 and η1 have the dimensions of a viscosity, neither depends significantly
on the viscosity of the solution [12, 45]. Likewise, τ0 increases during coarsening but
τ1 is constant. These two relaxations therefore have very different origins at the scale
of bubbles and films, as we will now describe.
4.1.4 Coupling between viscoelasticity of a foam and interfacial rheology

Under stress, a foam is deformed rapidly and acquires a non-equilibrium structure
which is governed by viscous forces and in which surface area is not minimized. The
transient regime with characteristic time τ1 observed in fig. 4.25a corresponds to a
relaxation towards an equilibrium structure, during which the area and orientation of
the films evolves in order to minimize the interfacial energy. The forces exerted by
the films on the Plateau borders drive the dynamics: a film of length L pulls on a
border with a force of the order of the surface tension γ, and resists the deformation
principally by its interfacial dilatational viscosity ηd (eq. (3.10)). A scaling argument
for the dynamics of the film suggests that γ ∼ (ηd /L) dL/dt [22] with a characteristic
time τ1 ∼ ηd /γ independent of bubble size. Typically, γ ≈ 30 mN.m−1 and ηd ≈ 10−3
to 10−2 kg.s−1 for interfaces with high dilatational modulus, which gives a time τ1
of the order of a fraction of a second, close to the value measured for shaving foam
(around 1 s). Note that this is similar to the dynamics of the relaxation of a film after
a T1 [5, 20]. Depending on the foaming solution, other dissipative processes associated
with the surfactants can play a role in the coupling between the viscoelasticity of the
foam and that of its interfaces (§3.3.2).
4.1.5 Coupling between slow relaxation and coarsening-induced T1s Gas

diffusion between neighbouring bubbles intermittently leads to unstable mechanical
configurations (cf. fig. 3.3) and the bubbles undergo T1s (to minimize the surface
energy locally) with an average frequency characteristic of coarsening (fig. 3.25). A
T1 causes a jump in the macroscopic strain: in a random direction in the absence of
stress, but biased in the shear direction if a stress is applied. Thus, at steady state
a foam slowly flows under a low stress like a Maxwell liquid (cf. p. 204). A scaling
˙ = ε̇(t)/σ is proportional to the compliance
argument suggests that the creep rate J(t)
Je0 , the average volume Vr of the foam affected by a T1, and to the number ωr of T1s
induced by coarsening per unit time and volume [12, 59]:
˙ ≈ Je0 ωr Vr ,
J(t) (4.56)
where we have omitted a prefactor of order one.
The phenomenon of creep was described at the macroscopic scale on p. 202.
Comparing the creep rates given by eqs. (4.55) and (4.56) allows us to link the
macroscopic viscosity to the bubble scale process, giving
1
= Je0 ωr Vr . (4.57)
η0
According to this relationship, the characteristic creep time, τ0 = η0 Je0 , varies in pro-
portion to the average time interval 1/ωr between T1s in a unit volume of foam. By
simultaneous in situ measurement of J(t) by rheometry and ωr by diffusing-wave spec-
troscopy (DWS, see §1.2.5, chap. 5), this linear variation can be observed, illustrated
in the inset to fig. 4.25. We deduce that, for a foam of 7% liquid fraction, Vr is of the
order of (6RV )3 which is a volume of around 50 bubbles, but it is not yet understood
why [14].
Scaling law for the complex shear modulus During coarsening, the average
bubble size RV slowly increases (eq. (3.38)) and the frequency ωr of T1s decreases.
As the age of the foam ta increases, we observe in figs. 3.27 and 3.25 that RV ∼ ta0.5
and τ −1 ∼ ωr Vr ∼ ta−0.66 . The (Maxwell-type) relaxation induced by T1s is coupled
to coarsening and the foam softens, with G ∼ γ/RV ∼ ta−0.5 . From eq. (4.57), with
G1 G (p. 204), the characteristic relaxation time τ0 ≈ η0 /G = 1/(ωr Vr ) increases
as ta0.66 , which shifts the maximum of G (located at f = 1/(2πτ0 )) towards lower
frequencies (fig. 4.24). We therefore find a simple scaling law for the evolution of the
complex shear modulus as a function of frequency and age, illustrated in fig. 4.26:
G∗ (ω, ta ) = b(ta , t0 ) G∗ (ω a(ta , t0 ), t0 ). (4.58)
The factors a(ta , t0 ) (for the frequency) and b(ta , t0 ) (for the moduli), determined in
comparison to a reference age t0 , allow the moduli at a later time ta to be determined.
See the shaded section overleaf. The linear viscoelasticity of foams at low frequency
thus shows a “time–age” equivalence, reminiscent of the equivalence between time and
temperature.15
15 It is known as the Williams–Landel–Ferry law in polymer rheology [41].
206 Rheology
1,000
normalised G'
Normalised modulus (Pa)
1/2
100
normalised G''
AGE t a
15 min
30 min
1h
2h
t 0 = 15 min 4h
8h
10
0.01 0.1 1 10 100
Normalised frequency (Hz)
Fig. 4.26: Master curves for the complex shear modulus (eq. (4.58)), representing the
normalized moduli G (ω, ta )/b(ta , t0 ) and G (ω, ta )/b(ta , t0 ) as a function of normal-
ized frequency f a(ta , t0 ). The sample is a shaving foam with φl = 0.08, and the moduli
are measured at ages from 15 min (reference age) to 8 hours. Data from [30].
“Time–Age” equivalence
At low frequency, the complex shear modulus G∗ (ω, ta ) depends on the frequency
ω and the age ta (eq. (4.58)). By definition, the factors a(ta , t0 ) and b(ta , t0 ) are
equal to one at the reference age ta = t0 (fig. 4.27). In this regime, if we ignore the
relaxation due to the interfaces (§4.1.4), the complex modulus is that of a Maxwell
element, with elastic modulus G and viscosity η0 :

iωη0 /G
G∗ (ω, ta ) ≈ G . (4.59)
1 + iωη0 /G
Comparing eqs. (4.58) and (4.59) gives
η0 (ta )
a(ta , t0 ) ∝ and b(ta , t0 ) ∝ G(ta ), (4.60)
G(ta )
suggesting that, according to eq. (4.57), the factor a increases with the average time
between T1s induced by coarsening, 1/(ωr Vr ), and that b decreases with 1/RV . It
is an effect that can be observed experimentally (fig. 4.27).
(a)
10 10
ω r (t a )Vr (t a )
ω r (t 0 )Vr (t 0 )
a(t a , t 0 )
1 1
10 1 10 2 10 3
Age t a (min)
(b)
1 1
b(t a , t 0 )
RV (t 0 )
RV (ta)
0.1 0.1
10 1 10 2 10 3
Age t a (min)
Fig. 4.27: Scaling for frequency a(ta , t0 ) and for modulus b(ta , t0 ) corres-
ponding to the master curve in fig. 4.26. (a) a(ta , t0 ) (•) is compared to
ωr (t0 )Vr (t0 )/(ωr (ta )Vr (ta )) (+), measured by DWS (§1.2.5, chap. 5). (b) (•)
b(ta , t0 ) is compared to RV (t0 )/RV (ta ) (×), measured by optical microscopy. Data
from [30].
4.1.6 Memory and structure Foams, like many other complex fluids, exhibit
a rheological “memory” in their microstructure. So the rheological properties of a
foam and the T1 dynamics depend strongly on the strain history. This memory effect
fades away progressively as the foam coarsens (another example of memory loss with
increasing age!).
208 Rheology
550
10 0
εp
Without
perturbation –1
500 10
Strain
10 –2
450 10 –3
1,000 t p 1,500 4,000
G' (Pa)
Age (s)
400
With
perturbation
350
300
1,000 tp 2,000 3,000 4,000
Age (s)
Fig. 4.28: Effect of rheological memory in shaving foam with φl = 0.08. The evolution
of the shear modulus in time, either during coarsening (◦) or following a perturbation
according to the protocol described in the inset (continuous line). Data from [31].
In order to demonstrate this property, we allow a foam to coarsen over a time tp ,

perturb it briefly with a shear of large amplitude εp (of the order of εy ), then leave
it to coarsen again (fig. 4.28). The shear alters the microstructure without changing
the liquid fraction or bubble size. The elastic modulus decreases in response to the
perturbation, then increases to reach the unperturbed value. This transient softening is
also observed in the glassy dynamics of soft paste, which is interpreted as a rejuvenation
of the system by a temporary increase in the rate of T1s, reducing the capacity of the
system to store elastic energy, and therefore reducing G . In foams, on the other hand,
the T1 rate ωr decreases after the perturbation, then relaxes towards the unperturbed
value. The changes in G and ωr which accompany this “solidification” of the foam
occur more slowly as the waiting time tp becomes longer. The duration of the recovery,
or memory, is determined by the time required to restore the characteristic structure
and dynamics of a coarsening foam through diffusive gas exchange between bubbles.
4.2 Transition from solid to liquid behaviour

4.2.1 Definition of yield stress and yield strain At the macroscopic scale,
the yield stress σy is defined as the maximum shear stress that can be applied to
the sample without causing it to flow; the corresponding strain is the yield strain εy
(p. 171). Shearing a disordered dry 3D foam quasi-statically suggests that at low strain
a foam behaves as a linear elastic solid (σ = G ε), which deforms plastically when ε is
large enough to induce T1 topological changes (cf. §3.2), and finally, when ε > εy , σ
fluctuates around σy and the foam flows. The characteristics of the stress fluctuations
caused by T1s in the flow regime are described in §5. These are visible in the numerical
simulations shown in fig. 4.22a; in that example the yield stress is
γ γ
σy ≈ 0.3 ≈ 0.2 (4.61)
V 1/3 RV
and εy ≈ 0.37. For bubbles of radius RV = 30 μm and surface tension γ = 30 mN.m−1 ,

this gives σy = 200 Pa. Compared to polycrystalline metals, for which εy ≈ 10−3 to
10−2 and σy ≈ 107 to 109 Pa, foams are much more elastic and flow at lower applied
stress.
4.2.2 Experimental methods to measure yielding Various experimental

methods (see §1.2.7, chap. 5) are used in order to characterize the yield stress σy
and yield strain εy of a foam:
• Observe a layer of foam slide down an inclined plane under its own weight (see
experiment 6.2). The layer thins and the flow stops when the thickness h is such
that σy = φl ρl gh sin α, where ρl is the density of the solution and α is the angle
at which the plane is inclined from the horizontal.
• Measure stress as a function of strain during start-up flow at constant shear rate
ε̇ (fig. 4.29a). The maximum stress is σy and the corresponding strain is εy . The
values of σy and εy obtained do not depend on ε̇ if it is small enough.
• Measure the limiting steady-state stress when a decreasing shear rate is applied
(fig. 4.29b). Extrapolating the flow curve σ(ε̇) to ε̇ = 0 gives an estimate of σy ,
known as the dynamic yield stress.
• Measure the complex modulus as a function of the amplitude of an applied os-
cillation ε0 at a fixed frequency (fig. 4.29c). The yield strain εy is the point of
intersection between the plateau in G (for ε0 < εy ) and the power law G ∝ εα 0
(for ε0 > εy ). σy can then be found from σy = εy |G∗ (εy )|. Similarly, we could
deduce σy from the variation in G in response to an oscillating stress, σ0 .
4.2.3 Non-linear viscoelasticity The response to an oscillatory perturbation of

small amplitude σ0 or ε0 (less than the yield strain) is linear and dominated by elastic
behaviour: G G , and G is independent of the amplitude (fig. 4.29c). On the other
hand, above the yield strain the response becomes non-linear. The complex modulus
is thus defined in terms of the fundamental frequency according to eq. (4.11), and the
−3/2
response is viscoplastic: G G and G decreases as ε0 , as shown in fig. 4.29c.
This visco-elasto-plastic behaviour can be described (fig. 4.2) by a spring, a dash-
pot, and a sliding friction element (schematized in fig. 4.29c). The spring represents
the elastic response: its stiffness corresponds to the elastic modulus G. The sliding
friction element describes the plasticity: it is rigid if the force applied to it is lower
than the yield stress σy , but it slides freely beyond it. The dashpot represents the
viscous dissipation, with Newtonian viscosity η given by the loss modulus in the linear
regime, such that G = ηω [44]. This phenomenological model gives a good prediction
of the variations of G∗ with strain amplitude shown in fig. 4.29c.
210 Rheology
.
.
.
. .
.
65 .
.
.
.
(s–1)
.
.
Normalised moduli
σ
/
.
. . .
ε ε
Fig. 4.29: Measuring the yield stress and yield strain of a foam using the methods
described in §1.2.7, chap. 5. (a) Stress as a function of strain during start-up flow at
constant shear rate ε̇ (parallel-plate rheometer). Data from [32]. (b) Flow curves σ(ε̇)
of foams used in oil-well drilling, measured at steady state using a capillary rheometer
under an applied pressure drop of 50 bars. The continuous lines represent fits to a
Herschel–Bulkley relation (eq. (4.6)). Note the absence of a yield stress for φl > 0.40.
Data from [29]. (c) G and G as a function of the strain amplitude ε0 measured
in oscillatory mode (cylindrical Couette rheometer). The continuous curves represent
fits to the elasto-visco-plastic model illustrated, with ηω/G = 0.08. The grey area
corresponds to the region where the strain is localized (see p. 214). Data from [44].
4.2.4 Effect of liquid fraction As for the shear modulus, the values of σy and
εy are greatest in dry foams and decrease when φl increases, reaching zero at φ∗l . In
effect, the strain (and therefore the stress) necessary to reduce to zero the length of
a Plateau border or the area of a film is smaller when the liquid content is greater
(fig. 4.30).
For disordered 3D foams, measurements in oscillatory mode (fig. 4.31) are
summarized in two empirical expressions [54]:
εy = a1 (φ∗l − φl ) (4.62)
(a) (b) (c) (d)
Fig. 4.30: Start of a shear-induced T1 (from a to d) in an ordered wet 2D foam

(φl = 0.06). The process is similar to a T1 in a dry foam until the point at which
the Plateau borders touch (b). The wet vertex is stable at (c), but at (d) it becomes
unstable and drives the completion of the topological change. From [49].
and
γ
σy = a2 (φ∗l − φl )2 , (4.63)
RV
where φ∗l = 0.36 is the volume fraction of close-packed spheres (fig. 4.17), a1 is a
constant between 0.5 and 1, and a2 is a constant beween 0.2 and 0.5. In 2D, the
yielding behaviour is similar [50]. Moreover, the flow curves (fig. 4.29b) indicate a
transition from a viscoelastic liquid (the absence of a yield stress in a dispersion of
bubbles) to a yield stress fluid (σy = 0) in which the yield stress increases as the foam
becomes drier. The transition takes place close to φl = 0.40, which is compatible with
φ∗l = 0.36.
4.3 Foam flow

We now turn to the steady flow of a 3D foam resulting from an applied stress greater
than σy (or an applied strain greater than εy ), but nevertheless sufficiently weak so as
not to induce coalescence or bubble break-up. Current understanding of 3D foam flow
is based on experimental results, complemented by numerical simulations and models,
which constitute the pieces of an as yet unfinished puzzle. We will review some of
them, though not exhaustively.
4.3.1 Effective viscosity At a macroscopic scale, both experiments and bubble

model simulations (§2.2.2, chap. 5) suggest that steady shear of a foam is characterized
by the phenomenological Herschel–Bulkley law (eq. (4.6)). The effective viscosity in
steady flow depends on the shear rate ε̇ according to
σ σy
ηeff ≡ = + ηp ε̇β−1 , (4.64)
ε̇ ε̇
with an exponent β < 1 (fig. 4.32), corresponding to shear-thinning behaviour [17, 26,
29, 32]. Note that the effective viscosity of a foam ηeff is several orders of magnitude
greater than the solution from which it is made (§3.3 and fig. 4.32). The exponent
212 Rheology
100 0.05
0.16
0.21
0.26
10
0.33
G (Pa)
1
0.40
φl
0.1
0.01
ε0
100
φl
0.05
0.16
G (Pa)
10 0.21
0.26
0.33
0.40
1
0.001 0.01 0.1 1 10
ε0
Fig. 4.31: The storage modulus G and loss modulus G as a function of strain
amplitude ε0 in oscillatory mode (ω = 1 rad.s−1 , cone-plate or cylindrical Couette
geometry, see §1.2.7, chap. 5). The sample is a polydisperse foam (R32 ≈ 66 μm and
γ ≈ 20 mN.m−1 ). At the two highest liquid fractions, the measurements are affected
by drainage. Data from [54].
β depends on the rheological properties of the gas/liquid interfaces [17] (§3.3.1): if

the interfaces have a low surface dilatational elasticity (like a solution of SDS above
the cmc), then the dissipation is localized inside the films, and β ≈ 0.5. In the case of
incompressible interfaces (as for a shaving foam or a fatty acid solution), the dissipation
10 8
Effective viscosity (mPa.s)
10 6
Foam
4
10
100
Solution
1
10 –5 10 –4 10 –3 0.01 10 2 10 3
. 0.1
ε (s –1 )
1 10
Fig. 4.32: Effective viscosity (eq. (4.64)) as a function of the shear rate for a shaving
foam (RV = 30 μm and φl = 0.08) and for the (Newtonian) solution from which it
is made (η = 1.8 mPa.s), measured in steady state flow using a cylindrical Couette
geometry (internal radius = 20.0 mm, gap-width = 4.1 mm). The continuous curve
shows a Bingham law (eq. (4.5)) with σy = 15 Pa and ηp = 1.5 Pa.s. Data from [26].
at the interfaces is dominant, and β ≈ 0.2. In practice, the change of β from 0.5 to
0.2 occurs at a surface dilatational modulus (eq. (3.13)) of the order of several tens of
mN.m−1 [17]. Note that the value of β is different from the exponents which arise due
to dissipation at a solid wall (§3.3.3).
The plastic viscosity or consistency ηp also depends on the physico-chemical char-
acteristics of a foam [17]. For a given surface tension, bubble size, and liquid fraction,
ηp is much greater for incompressible interfaces than for perfectly compressible ones.
In other words, the rate of dissipation of mechanical energy increases with the rigidity
of the interfaces.
If the shear rate is not too great, the contribution of the yield stress term in
eq. (4.64) dominates the effective viscosity, which thus varies as γ/(RV ε̇); we call this
the plastic regime. In the limit of large ε̇, the flow is dominated by viscous effects.
Recent experimental and theoretical work suggests that the quantity σ − σy depends
on the capillary number Ca = ηRV ε̇/γ through the relationship [17]
σ − σy
= f (φl ) Ca0.5 , (4.65)
γ/RV
where the function f (φl ) depends not only on the liquid fraction but also on the
disorder of the foam. In steady state flow the bubbles slide past each other and the
contacts between bubbles are thus constantly renewed. The contact dynamics intro-
duce local shear flows which controls the film thickness. This process is governed by
214 Rheology
the competition between the viscous stress η ε̇ and the Laplace pressure γ/RV , which
determines the capillary number.
Macroscopic phenomenological laws like eq. (4.64) or the scalar model represented
in fig. 4.29c do not include all the physical properties that govern foam flow. It is also
necessary to try and capture spatial variations [3, 55].
4.3.2 Is the flow homogeneous? Experiments and simulations show that foam
flow is not always homogeneous. For an experiment in a rheometer, it is therefore
useful to introduce a local shear rate ε̇local , defined over a length-scale smaller than
the gap-width, but larger than the bubble size. At low shear rates, if the stress σ is only
slightly greater than the yield stress σy , “solid” (ε̇local = 0) and “liquid” (ε̇local > 0)
regions may coexist within the sample. When the shear rate varies discontinuously at
the interface between the two, the flowing region is called a shear band.
Other complex fluids consisting of jammed packings, like concentrated emulsions,
soft pastes, gel microbeads, or colloidal dispersions, also exhibit shear banding in
steady flow. Shear bands are not a universal characteristic of complex fluids, insofar as
they depend on physico-chemical parameters—for example the volume fraction of the
dispersed phase or the nature of the interactions between its elements. Moreover, they
are occasionally superimposed upon thixotropic effects16 which further complicates
our understanding of them [28]. We next describe the conditions necessary for shear
bands to appear in a 3D foam (the 2D case is shown in fig. 2.29).
Flow under homogeneous stress Flows in a parallel-plate or cone-plate rheo-
meter, or even a cylindrical Couette rheometer if the gap is narrow (cf. §1.2.7, chap. 5),
are characterized by a constant stress across the gap. We therefore expect that the
shear rate is constant across the gap too, whatever the constitutive law. However,
this is not always found experimentally. Imaging by optical tomography (§1.2.4, chap.
5) shows that in a dry foam, above the yield strain, the velocity profile does not
vary linearly with position, but that shear bands appear, at a variable depth [53].
Heterogeneous flows have also been found in Surface Evolver simulations of ordered
dry 3D foams [52]. On the other hand, bubble model simulations of disordered wet 3D
foams don’t show this [24]. So the physical parameters governing foam flow are still
not well-understood, as the following paragraph confirms.
Flow under heterogeneous stress Flows in a cylindrical Couette rheometer are
characterized by a shear stress which decreases across the gap as 1/r2 , where r is
the distance from the axis of rotation (§1.2.7, chap. 5). Consequently, when a yield
stress fluid is sheared at a rate ε̇ that is so low that viscous stresses are negligible
compared to σy , only a narrow region close to the inner cylinder experiences a stress
greater than σy and flows. The rest of the foam sample then behaves as an elastic
material (fig. 4.33). MRI measurements of the local velocity profiles (§1.2.4, chap. 5)
in the steady shear of foams of various bubble sizes, liquid fractions (φl between 0.05
and 0.12), and interfacial rigidities [48] in a large-gap Couette cell have not detected
any shear banding; instead, the shear rate was found to decrease continuously to
16 A fluid is thixotropic if its effective viscosity changes over time.

Appendix: From the discrete to the continuous 215
σ
σ max Flowing region Ri
Ro
σy
Solid behaviour
σ min r
Ri Ro
Fig. 4.33: Shear of a yield stress fluid in a cylindrical Couette rheometer with internal
radius Ri and external radius Ro (cf. fig. 5.11). A view from above and the profile of
stress aross the gap (σ decreases as 1/r2 ) showing that only the dark grey region close
to the inner cylinder flows.
zero at the interface between the flowing and solid-like regions as σ tended toward
σy (fig. 4.33). These results, combined with torque measurements, suggest that in
steady flow these foams behave like Herschel–Bulkley fluids (eq. (4.6)). It is therefore
necessary to distinguish the heterogeneous flow in a large-gap Couette geometry from
the shear banding described above, a distinction which is also apparent in 2D (see [10]
and §5.4, chap. 2).
5 Appendix: From the discrete to the continuous

The elastic response of a few bubbles, discussed in §3.1, and of a foam with many
bubbles, given by eq. (4.49), are directly comparable. On the other hand, the plastic
response is very different and going from small to large scale requires further consid-
eration. In a dry hexagonal 2D foam, i.e. the periodic repetition of a single bubble
(fig. 4.10), the stress is completely relaxed after the first T1 (fig. 4.11a). If the shear
continues, we observe periods of elastic loading followed by plastic unloading, and the
stress oscillates between 0 and the yield stress. In contrast, fig. 4.29a shows that in
a large disordered foam the stress converges to a constant value in the plastic regime.
An intermediate regime is observed in a sample of about 1,000 bubbles, shown in
fig. 4.22a, with partial relaxations of the stress due to T1s, and small, though not
negligible, stress fluctuations around the average stress. What is significant is that the
relative amplitude of the fluctuations decreases with the number of bubbles and with
disorder.
To be more quantitative, we consider the Stokes experiment in a foam (fig. 4.34a),
in which a small spherical bead of radius Rbead is pulled slowly through a foam at
constant speed. The bead is suspended from a balance by a thin rigid rod, enabling the
force on the bead to be measured. Fig. 4.34b shows an initial period of elastic loading
before the force fluctuates around an average value. In this second regime, linear
elastic loading with constant slope is interrupted by quasi-instantaneous unloading
due to T1s. The T1s are localized around the sphere (fig. 4.34c), and the amplitude
216 Rheology
2.5
2.0
Force (mN)
1.5
2.5
1.0
2.0
0.5
1.5
1,200 1,400
0
0 1,000 2,000 3,000
Time (s)
(a) (b)
(c)
Fig. 4.34: Stokes experiment in a 3D foam. (a) A spherical bead (Rbead = 0.5 cm)
pulled at constant speed (5 μm.s−1 ) through a disordered monodisperse dry foam
(RV = 1.5 mm). Photograph courtesy of I. Cantat and O. Pitois. (b) Force exerted
on the bead as a function of position. (c) Projected map of the T1 density (cf. the
result in the 2D case in fig. 2.28d). Data from [7]. Image courtesy of I. Cantat and
O. Pitois.
of the jump in the force is of the same order of magnitude as if a bubble in direct
contact with the sphere were detached from it: δF ∼ γRV . On dimensional grounds,
the average force is approximated by the product of the yield stress γ/RV and the
2
surface area of the obstacle Rbead . The relative fluctuations are therefore of the order
2 2
of RV /Rbead , and so are smoothed out when the bead becomes very large compared
to the bubbles [7].
Experiments 217
6 Experiments
6.1 Observation of T1s
preparation time: 4 hrs experimental time: 4 hrs
Phenomena demonstrated: T1s and flow.

Reference in the text: §3.2.
Materials:
• equipment for making a 2D foam (see §6.2, chap. 2)
• an obstacle which is the same thickness or thicker than the foam, for example a
slice cut from a plastic cork
• an aquarium air-pump and a flexible tube.
1. Position the obstacle at the centre of the device used to create the foam (centre of
the bowl or CD case). We recommend the assembly between two solid plates (i.e.
the CD case), or assembly between the liquid and the solid plate.
2. In order to create a flow in one direction, insert the end of the tube at one end
of the device and connect it to the pump using the flexible tubing to create air
bubbles continuously at constant gas flow rate. Make the 2D foam according to the
instructions given in §6.2, chap. 2.
3. Observe the T1s, primarily upstream and downstream of the obstacle (figs. 2.28d
and 4.34c).
The presence of the obstacle perturbs what would otherwise be a plug flow, induces
shearing of the bubbles, and causes some edges to disappear, triggering T1s.
6.2 Visualization of the yield stress

preparation time: 1 hr experimental time: 1 hr
Phenomenon demonstrated: yield stress of a foam.

Reference in the text: §4.2.1.
Materials:
• fairly wet shaving foam (φl ∼ 0.20)
• a flat, rough, rigid plate (around 10 cm × 10 cm). To make it rough, sandpaper
can be stuck onto the plate
• something to incline the plate: one end of the plate should rest on a horizontal
plane and the other end should be progressively raised using a wedge or laboratory
jack
• a square or rectangular frame of around 5 cm in length and width and a height
of around 2 cm.
218 Rheology
Fig. 4.35: Flow of a shaving foam down a plane inclined at an angle α to the hori-
zontal. The image on the left is taken at the start of the flow; the image on the right
is taken at the end of the flow. The black lines are drawn on the foam with a liquid
containing a dispersion of carbon particles (printer toner or Indian ink for example),
allowing the flow to be visualized. Photographs courtesy of F. Rouyer.
1. Place the flat plate horizontally.

2. Position the frame in the middle of the plate.
3. Fill the frame with foam and make the upper surface level.
4. Carefully pull the frame away without changing the shape of the foam.
5. Slowly (but not so slowly that the foam coarsens) tilt the plane at an angle α ∼ 10◦ .
6. If the foam doesn’t move, increase the angle slightly.
7. Now leave the foam to spread towards the bottom of the plate and measure its final
thickness hf .
Under the effect of its own weight, the foam is subjected to a shear stress σ =
φl ρl gh sin α. The foam flows when this shear stress exceeds its yield stress σy , and
stops flowing at a height hf for which σy = φl ρl ghf sin α.
7 Exercises
7.1 The Young–Laplace law and the stress in a spherical bubble
(a) Use Batchelor’s expression for the stress tensor (eq. (4.18)) to recover the Young–
Laplace law (eq. (2.3)), giving the pressure inside a spherical gas bubble of radius R
in a liquid at pressure P0 .
(b) Now calculate the contribution of this bubble to the stress tensor and show that
its contribution to the average deviatoric stress is zero. What is the contribution
of the bubble to the average pressure in the foam?
Exercises 219
7.2 Elasticity of a dry 2D foam
(a) Find the yield strain εy and yield stress σy (eq. (4.37)) of a dry hexagonal 2D
foam using the diagram in fig. 4.36. Now deduce the shear modulus Ĝ of the
foam, assuming linear elasticity.
(b) Eq. (4.26) shows that as the yield strain is approached, the relationship between
stress and strain becomes non-linear. Demonstrate the linear elastic behaviour
at small strains by an expansion of eq. (4.26) in ε. Deduce from it the shear
modulus Ĝ.
(c) Compare the approximate (part (a)) and exact (part (b)) expressions for the
modulus.
(d) In a dry hexagonal 2D foam at equilibrium, each vertex is a symmetrical junction of
three films. By placing on each vertex the same small quantity of liquid, we obtain,
according to the decoration theorem, the structure of a wet foam at equilibrium.
The shape of the vertices is illustrated in fig. 4.18. Develop a symmetry argument
to show that the contribution of the decorated vertices to the average shear stress
is zero. Use the fact that the off-diagonal elements of the stress tensor must remain
invariant when a vertex is turned through an angle of 120◦ .
(e) Show that, for the same reason, the linear elastic response of a 2D hexagonal foam
does not change if we decorate its vertices with small, three-sided bubbles.
7.3 Poynting’s law

In a non-linear elastic solid, the stress as a function of strain can be expressed in terms
of the Finger tensor (eq. (4.13)):
σij = f (Bij ). (4.66)
In the following, consider an isotropic undeformed solid.
(a) Show that for the shear strain ε illustrated in fig. 4.1, the Finger tensor is
⎛ ⎞
1 + ε2 ε 0
B=⎝ ε 1 0⎠. (4.67)
0 0 1
A B C
Fig. 4.36: A sheared hexagonal 2D foam, as in fig. 4.10. The basic unit of the network
is the region inside the thin-lined rhombus. (A) Foam at rest before the shear; (B) the
strain reaches the yield strain; (C) T1s occur.
220 Rheology
(b) Show that the tensor B is diagonal in a frame of reference rotated through an
angle θ about the z axis, with
2 cot(2θ) = ε. (4.68)
In a strained isotropic elastic material, the principal axes of B and of the stress
tensor are the same, so in the new frame of reference the stress tensor must be
diagonal:
⎛ ⎞
σ1 0 0
σ = ⎝ 0 σ2 0 ⎠ . (4.69)
0 0 σ3
Find an expression for σ in the (x, y, z) frame of reference, and use it to show that
2 cot(2θ) = N1 /σxy . (4.70)
(c) Deduce Poynting’s law (eq. (4.50)) from eqs. (4.68) and (4.70).
7.4 Stress and strain in a square lattice

Consider a lattice of squares, with side-length a at rest, oriented in the x and y
directions, and assume that each edge of the lattice transmits a force 2γ. This lattice
could represent a 2D foam in which Plateau’s laws are not respected. Assume that all
pressures are zero.
(a) Subject the lattice to an elongational strain εxx = −εyy = ε, εxy = 0. Calculate
the stress tensor, either using Batchelor’s expression (eq. (4.18)) or by determining
directly the force per unit length exerted on a line parallel to the x or y direction.
Note that Hooke’s law in 2D (eq. (4.15)) may also be expressed as σxx =
2Gεxx +Tr(σ)/2, where Tr(σ) = σxx + σyy is the trace of the stress tensor. Hence
deduce that G = γ/a.
(b) Subject the lattice to an affine shear by shifting the point M (x, y) to the point
M (x + εy, y). Using the same methods as in part (a), show that the stresses σxx
and σyy are quadratic functions (to leading order in ε) of the strain, and give their
difference as a function of σxy .
7.5 Elasticity and plasticity

This exercise is a continuation of exercise 7.3, chap. 2.
(a) Elasticity
Consider again the equilibrium shape of the system defined in exercise 7.3, chap.
2, changing a and b whilst preserving ab = S.
(1) Write the energy of the system in terms of the ratio a/b for the stable (and
possibly metastable) configuration, and give the value of a/b for which the
energy is minimized.
(2) Show that for this ratio Fx /b = Fy /a, with Fx the resultant force in the x
direction, exerted on the two right-hand films, and Fy the resultant force in
the y direction exerted on the two upper films.
References 221
(b) Plasticity
When the ratio b/a exceeds the limit of existence of the metastable state, a T1
occurs; determine the energy lost during the relaxation towards a new stable state.
7.6 Compressibility of a foam
(a) The isotherm of compressibility for a dry foam, in equilibrium with a reservoir
of pressure Pext , is χT = −(1/V ) |∂V /∂Pext |T . Assuming that a change in volume
may be made without changing the structure, i.e. by a homothetic transformation,
use one of the methods below to show that [33]
4 γ Sint
χ−1
T = Pext + . (4.71)
9V
Method 1 (pressure argument): differentiate the equation of state (eq. (2.18) or
exercise 7.6, chap. 2), or use Batchelor’s expression (eq. (4.18)).
Method 2 (energy argument): use the thermodynamic relationship ∂ 2 F/∂V 2 T =
1/(V χT ), where F is the free energy of the foam. (See exercise 7.6, chap. 2 for the
terms that must be taken into account in F.)
(b) Make a qualitative comparison of the foam and gas compressibilities, either at
atmospheric pressure Pext , or at the foam pressure Pint . Hence find the bulk
modulus of a dry foam and compare it to the shear modulus.
(c) The isothermal bulk compression modulus of an elastic material is the inverse of
the isotherm of compressibility, defined by KT = −V |∂P/∂V |T , where V is the
volume of the material, P its pressure, and T its temperature. By considering an
infinitesimal variation of pressure δP , show that the bulk compression modulus of
a wet foam with liquid fraction φl is
KT,gas
KT (1 − φl ) ≈ , (4.72)
1 − φl
where KT,gas is the bulk compression modulus of the gas.
References
[1] D.J. Acheson, Elementary Fluid Dynamics, Clarendon Press, Oxford, 1990.
[2] G. Batchelor, J. Fluid Mech., 41, 545, 1970.
[3] S. Bénito, C.-H. Bruneau, T. Colin, C. Gay, and F. Molino, Eur. Phys. J. E,
25, 225, 2008; S. Bénito, F. Molino, C.-H. Bruneau, T. Colin, and C. Gay,
Eur. Phys. J. E, 35, 51, 2012.
[4] S. Besson and G. Debrégeas, Eur. Phys. J. E, 24, 109, 2007.
[5] A.-L. Biance, S. Cohen-Addad, and R. Höhler, Soft Matter, 5, 4672, 2009.
[6] D. Buzza, C.-Y. Lu, and M.E. Cates, J. Phys. II, 5, 37, 1995.
[7] I. Cantat and O. Pitois, Phys. Fluids, 18, 083302, 2006.
[8] I. Cantat, Phys. Fluids, 25, 031303, 2013.
[9] I. Cheddadi, P. Saramito, B. Dollet, C. Raufaste, and F. Graner, Eur.
Phys. J. E, 34, 1, 2011; I. Cheddadi, P. Saramito, F. Graner, J. Rheol., 56,
213, 2012.
222 Rheology

Eur. Phys. J. E 27, 123, 2008.
[11] S. Cohen-Addad, H. Hoballah, and R. Höhler, Phys. Rev. E, 57, 6897, 1998.
[12] S. Cohen-Addad, R. Höhler, and Y. Khidas, Phys. Rev. Lett., 93, 028302,
2004.
[13] S. Cohen-Addad, R. Höhler, and O. Pitois, Ann. Rev. Fluid Mech., 45, 241,
2013.
[14] S.J. Cox, F. Graner, and M.F. Vaz, Soft Matter, 4, 1871, 2008.
Wetting Phenomena: Drops, Bubbles, Pearls, Waves, Springer, New York,
2004.
[16] N.D. Denkov, V. Subramanian, D. Gurovich, and A. Lipsa, Colloids Surf.
A., 263, 129, 2005.
[17] N.D. Denkov, S. Tcholakova, K. Golemanov, K.P. Ananthpadmanabhan,
and A. Lips, Soft Matter, 5, 3389, 2009.
[18] B. Derjaguin, Kolloid Z., 64, 1, 1933.
[19] F. Dias, D. Dutykh, and J.-M. Ghidaglia, Comput. Fluids, 39, 283, 2010.
[20] M. Durand and H.A. Stone, Phys. Rev. Lett., 97, 226101, 2006.
[21] D.J. Durian, Phys. Rev. E, 55, 1739, 1997.
[22] D.A. Edwards, H. Brenner, and D.T. Wasan, Interfacial Transport Processes
and Rheology, Butterworth Heinemann, Stoneham, 1991.
[23] J. Ferry, Viscoelastic Properties of Polymers, Wiley, New York, 1980.
[24] B.S. Gardiner, B.Z. Dlugogorski, and G.J. Jameson, J. Non-Newton. Fluid
Mech., 92, 151, 2000.
[25] B.S. Gardiner, B.Z. Dlugogorski, and G.J. Jameson, J. Rheol., 42, 1437,
1998.
[26] A.D. Gopal and D.J. Durian, J. Colloid Interface Sci., 213, 169, 1999.
[27] A.D. Gopal and D.J. Durian, Phys. Rev. Lett., 91, 188303, 2003.
[28] B. Herzhaft, J. Colloid Interface Sci., 247, 412, 2002.
[29] B. Herzhaft, S. Kakadjian, and M. Moan, Colloids Surf. A, 263, 153, 2005.
[30] R. Höhler and S. Cohen-Addad, J. Phys.: Condens. Matter, 17, R1041, 2005.
[31] R. Höhler, S. Cohen-Addad, and A. Asnacios, Europhys. Lett., 48, 93, 1999.
[32] S.A. Khan, C.A. Schnepper, and R.C. Armstrong, J. Rheol., 32, 69, 1988.
[33] A.M. Kraynik and D.A. Reinelt, J. Colloid Interface Sci., 181, 511, 1996.
[34] A.M. Kraynik and D.A. Reinelt, in Proc. XIVth Int. Congr. Rheology, Seoul,
The Korean Society of Rheology, 2004.
[35] A.M. Kraynik, D.A. Reinelt, and F. van Swol, Phys. Rev. E, 67, 031403,
2003.
2004.
[37] N.P. Kruyt, J. Appl. Mech., 74, 560, 2007.
[38] V. Labiausse, R. Höhler, and S. Cohen-Addad, J. Rheol., 51, 479, 2007.
[39] M.-D. Lacasse, G.S. Grest, and D. Levine, Phys. Rev. E, 54, 5436, 1996.
[40] R.G. Larson, J. Rheol., 41, 365, 1997.
References 223
[41] R.G. Larson, The Structure and Rheology of Complex Fluids, Oxford University
Press, New York, 1999.
[42] A.J. Liu, S. Ramaswamy, T.G. Mason, H. Gang, and D.A. Weitz, Phys. Rev.
Lett., 76, 3017, 1996.
[43] C.W. Macosko, Rheology: Principles, Measurements, and Applications, Wiley-
VCH, Weinheim, 1994.
[44] P. Marmottant and F. Graner, Eur. Phys. J. E, 23, 337, 2007.
[45] S. Marze, A. Saint-Jalmes, and D. Langevin, Colloids Surf. A, 263, 121,
2005.
[46] T.G. Mason, J. Bibette, and D.A. Weitz, Phys. Rev. Lett., 75, 2051, 1995.
[47] N. Mujica and S. Fauve, Phys. Rev. E, 66, 021404, 2002.
[48] G. Ovarlez, K. Krishan, and S. Cohen-Addad, EPL, 91, 68005, 2010.
[49] H.M. Princen, J. Colloid Interface Sci., 91, 160, 1983.
23, 217, 2007.
[51] D.A. Reinelt and A.M. Kraynik, J. Fluid Mech., 311, 327, 1996.
[52] D.A. Reinelt and A.M. Kraynik, J. Rheol., 44, 453, 2000.
[53] F. Rouyer, S. Cohen-Addad, M. Vignes-Adler, and R. Höhler, Phys. Rev.
E, 67, 021405, 2003.
[54] A. Saint-Jalmes and D.J. Durian J. Rheol., 43, 1411, 1999.
[55] P. Saramito, J. Non-Newtonian Fluid Mech., 145, 1, 2007; P. Saramito,
J. Non-Newtonian Fluid Mech., 158, 154, 2009.
[56] L.W. Schwartz and H.M. Princen, J. Colloid Interface Sci., 118, 201, 1987.
[57] D. Stamenović, J. Colloid Interface Sci., 145, 255, 1991.
[58] E. Terriac, J. Etrillard, and I. Cantat, Europhys. Lett., 74, 909, 2006.
[59] S. Vincent-Bonnieu, R. Höhler, and S. Cohen-Addad, Europhys. Lett., 74,
533, 2006.
[60] D. Weaire, Phil. Mag. Lett., 60, 27, 1989.
[61] D. Weaire and S. Hutzler, The Physics of Foams, Clarendon Press, Oxford,
1999.
5
Experimental and numerical methods
The study of foams is interdisciplinary, both in terms of the questions that foams give rise to
and the methods (experimental, simulations, and theoretical) used to study them.
1 Experimental methods
We present some of the experimental techniques used in research laboratories in order
to make and study foams. The appropriate method depends on the length-scale that
should be probed, i.e. liquid/gas interface, single liquid film, bubble, or macroscopic
sample. We will also show how the optical and electrical properties of a foam depend
on its physical parameters.
1.1 Methods used to study interfaces and isolated films

There are numerous techniques used to study gas/liquid interfaces in the presence of
surfactants. We choose to concentrate on those for measuring the surface tension, the
interfacial viscoelasticity, and the properties of thin liquid films. Other techniques such
as ellipsometry, Brewster angle microscopy, and infrared spectroscopy are described
elsewhere [1].
1.1.1 Surface tension There are several different ways to measure surface tension;
we briefly introduce the three that are used most often, and the reader is referred to
[15] for more detailed descriptions.
One possibility is to measure the capillary force exerted on an object immersed in
a solution. In Wilhelmy’s method a thin strip of platinum or filter paper (to enforce
total wetting) is used, while in de Nouy’s method a ring is used. The precision is of
the order of ±0.1 mN.m−1 , and the method is most suited to measuring equilibrium
values.
In the pendant drop method [1, 15], a liquid drop (typically 1 mm in diameter)
is suspended from the end of a needle and its shape determined. Since the pressure
difference at every point of the interface is given by the Young–Laplace law (§2.1.3,
chap. 2), the shape of the drop and its variation in time depend on the surface tension.
The method gives γ(t) for times t from one second to several hours, and the accuracy
of the measurement is again ±0.1 mN.m−1 . The bubble configuration [1, 15], in which
a bubble is held at the end of a needle in a tank filled with solution, and the sessile
drop configuration, a liquid drop placed on a substrate, are similar.
Finally, the maximum-bubble-pressure method gives the surface tension by measur-
ing the pressure at which a bubble detaches from a submerged capillary [43]. A capillary
226 Experimental and numerical methods
of radius Rc (typically of the order of several hundred micrometres) is immersed in

a foaming solution and gas injected to inflate a bubble with known pressure at each
instant. This pressure is 2γ/Rc , by the Young–Laplace law, and is maximal when the
bubble radius is equal to Rc . On further inflation, the bubble detaches from the capil-
lary, and by rapidly repeating the inflation and detachment (usually at frequencies of
a few kHz), it is possible to measure γ(t) on a time-scale of the order of 1 ms to 1 s.
This precision is necessary for surfactants of low molecular mass (see §6.1, chap. 3) for
which the surface tension is significantly lowered immediately after the formation of
an interface; for example, for SDS, above the cmc, the tension drops from 72 mN.m−1
to around 38 mN.m−1 in less than a second.
1.1.2 Surface viscoelasticity

Rheological methods for measuring dilatational viscoelasticity In order to
study the response to an expansion or a compression, it is necessary to vary the area of
a gas/liquid interface. The in-phase and out-of-phase response of the surface tension to
oscillations in interface area over a range of frequencies (experimentally ranging from
a fraction of one Hz to several thousand Hz) gives both the elastic Es and viscous Es
2D moduli (§2.2.3, chap. 3).
The pendant drop or bubble method for measuring surface tension (§1.1.1) can
also be used in oscillatory mode. A sinusoidal oscillation of the volume of the drop
induces a change in the area of the interface and then by tracking the changing shape
of the drop the complex modulus Es∗ can be found, typically for frequencies between
0.01 and 100 Hz.
The detection of capillary waves (whether produced by thermal fluctuations or
induced by an oscillator) allows the dilational viscoelastic properties of an interface to
be measured at high frequency, typically several hundred Hz [52]. In practice, this is
carried out by electro-capillary methods, in which a wave of controlled frequency and
amplitude, for example an alternating voltage of several hundred volts, is applied to
a strip positioned 1 mm above the interface. This raises and lowers the surface of the
liquid without ever touching it. Then the wavelength and the attenuation of the wave
that propagates at the surface are measured optically [52]. The appeal of this method
for measuring interfacial viscoelasticity lies in the fact that the dispersion relation for
a surface wave is different from that in pure liquid, because of the coupling between
the wave in the liquid and the viscoelastic properties of the film in compression. The
propagation of the wave creates regions of expansion and compression of the interface
(fig. 5.1a). The observed wavelength always depends strongly on the surface tension,
whilst the attenuation is linked to the surface dilatational viscoelasticity.
Finally, it is also possible to make measurements in a Langmuir trough (see shaded
section on p. 58). A solution is poured into a shallow container containing a mobile
barrier that allows the area of the interface to be varied (fig. 5.1b). Using a Wilhelmy
plate (§1.1.1) allows variations in surface tension to be measured while the barrier is
translated or slowly oscillated (over a period of seconds or minutes).
Rheological methods for measuring shear viscoelasticity Measuring the 2D
shear moduli is a delicate procedure because it requires simultaneous control of the
Experimental methods 227
(a)
Region of film
compression
Region of film
elongation Gas
Liquid
Direction of wave propagation
(b) Mobile barrier

Interface with surfactants
Liquid
Fig. 5.1: Methods for measuring interfacial viscoelasticity in compression:

(a) Coupling between a capillary wave and expansion/compression of the film.
Counter-rotating circles represent trajectories of fluid particles as the wave passes.
The thick vertical bar represents a metal strip to which an alternating current is
applied to induce a surface wave. (b) Langmuir trough with an oscillating barrier.
shear applied to an interface, measurement of the interfacial stresses and strains, and
accounting for the coupling with flow in the subphase. However, it is possible, both in
oscillatory and continuous modes.
A number of methods are based on generating shear by moving a floating object
such as a magnetized needle (fig. 5.2a) [6].
Other methods include a 2D analogue of a 3D Couette rheometer (§1.2.7), in which
a disc or horizontal bicone is suspended from a vertical thread and placed at the
interface and rotated relative to the outside wall (fig. 5.2b). Measuring (usually by
optical means) its angle of rotation gives the strain, whilst measuring the torque gives
the 2D shear stress [45, 61].
A third group of methods, more suitable for insoluble surfactants, uses a Langmuir
trough (fig. 5.1b). The compression barrier is used to push the free surface of the
liquid into a narrow channel and the interfacial velocity field measured by tracking
markers (e.g. talc) placed on the surface.
Finally, it is possible to observe the flow in a 3D channel such as a Plateau border:
as described in §4, chap. 3, the coupling between the flow at the surface and in the
(a) (b)
Bicone
Gas or
Magnetized liquid 2
floating needle
B
Liquid 1
Interface
View
from
above
Fig. 5.2: Methods for measuring interfacial shear viscoelasticity. (a) A floating mag-
netized needle is oriented by a magnetic field and moved by a magnetic field gradient:
it shears the surface. (b) Shear of the interface using a bicone device: once positioned
precisely at the interface, its circumference shears the interface. The velocity field is
represented by arrows, here in the case where the inner cylinder is turned and the
outer wall is fixed.
bulk is controlled by the surface shear viscosity. By measuring the speed of macroscopic
drainage, we obtain the surface mobility and therefore the shear viscosity [17, 47].
1.1.3 Reflectometry of liquid films The colourful iridescence so easily observed

on the surface of a bubble is the result of interference between the waves of light that
are reflected by each of the two interfaces of a film. This is possible because the films
are so thin: for a film of thickness h illuminated at normal incidence, the intensity of
the reflected light, ignoring multiple reflections, is

4πn0 h
I = RI0 1 − cos , (5.1)
λ
where n0 is the refractive index of the soap solution, λ the wavelength of light used,
I0 the intensity of the light source, and R the reflection coefficient of the interface
(R ≈ 4% for an air/water interface).
For a given wavelength λ, the intensity is minimal when h is a multiple of λ/2n0
and maximal when h is an odd multiple of λ/4n0 . In effect, if one ray is reflected at a
gas/liquid interface and another at a liquid/gas interface, between the two rays there
is a phase shift of π; this is taken into account in eq. (5.1).
If white light is used, each wavelength gives rise to its own particular interference
pattern, and the intensity minimum for each colour occurs for a different film thickness.
All the intensities sum incoherently, which gives a progression of colours (cf. figs. 2.33
and 3.55). A particular colour therefore corresponds to a certain thickness (although
the converse is not true: different thicknesses can show the same colour), and gradients
of thickness give rise to gradients of colour. The colour and the intensity also depend
strongly on the direction from which they are observed.
As the thickness increases, the colour gradations become blurred: green and
pink are obtained first (typically corresponding to thicknesses of several hundred
nanometres) then a mixture of numerous colours called high-order white.
At the other extreme, if the film is very thin, the difference in the optical path length
between the two rays becomes very small compared to λ. Then the only possible phase
shift between the two rays is π and so there can only be nearly destructive interference
and the film becomes almost transparent, known as a black film.1 The thickness of a
black film can be up to 100 nm, so they can be found in a foam (see §2.3.2, chap. 3).
In order to determine the precise thickness of a film as it drains and its thickness
changes, a monochromatic source is used to find the minima and maxima of reflected
intensity (eq. (5.1)). For an absolute measurement, it is necessary to identify the order
of the interference colours (starting with the black film, for example) [7, 54].
Interferometry is also possible with X-rays or neutron beams [3, 20, 50], allowing
structures of the order of nanometres to be probed. X-rays are able to show not only
the thickness but also the structure of the film. Neutron sources have lower flux than
X-rays, but because it is possible to label parts of the material (by replacing hydrogen
atoms with deuterium), they enable precise study of the structure [20].
1.1.4 Film balance technique The film balance technique (also called the porous
plate technique or the thin-film balance) is based on an idea of Mysels [44]. It enables
the study of the properties of a single liquid film held in a frame and the interactions
between interfaces at the scale of thin films [7, 54].
A film is suspended horizontally in a hole (of typically 1 mm in diameter) pierced
in a disc-shaped glass frit (of typical thickness 3 mm; see fig. 5.3a). The frit and the
capillary tube to which it is soldered act as a reservoir of solution and are kept at
constant pressure. The frit is enclosed in a chamber in which the pressure P applied
to the film can be varied (fig. 5.3b). If the repulsive forces between the two interfaces of
the film are large enough, they create a disjoining pressure Πd sufficient to equilibrate
the pressure P . In this case a stable film is obtained, and by measuring the equilibrium
thickness h for each value of P by monochromatic interferometry (§1.1.3), we can
construct the curves Πd (h) (§2.3.2, chap. 3). In the example given for a micellar solution
in fig. 5.4, several branches are found, illustrating the stratification of the micelles in
the film (§2.3, chap. 3). The film doesn’t thin continuously, but its thickness jumps
from one value to another as the pressure is increased. On the other hand, by reducing
the pressure, the curve returns to zero along the same branch each time, indicating
that the organization of the micelles into layers only occurs under confinement.
The film can be visualized directly using a microscope, making it possible to observe
its homogeneity, any instabilities, and even the dynamics associated with a jump in
thickness. Several examples are presented in chap. 3 (figs. 3.17, 3.18, 3.40, and 3.44).
Such direct visualization of the film structure, and thus of possible non-trivial effects
1 Newton called it “black film” because he observed it against a black background.

(a) (b)
Piston
Capillary
Microscope
Glass frit
Film
Pressure P
Film
Fig. 5.3: The film balance technique. (a) A glass frit is soldered to a U-shaped ca-
pillary tube; the film is formed in a hole in the frit. (b) The frit is placed in a sealed
chamber within which the pressure is controlled, and the film is observed through a
microscope.
200
160 SDS (c = 10 cmc)
120
Πd (Pa)
80
40
0
0 10 20 30 40 50 60
Thickness h (nm)
Fig. 5.4: Disjoining pressure Πd (h) as a function of the thickness h, measured with
the film balance technique [7]. The different branches of the curve illustrate the effect
of the confinement described in fig. 3.16.
(thickness heterogenities, aggregation), enables a better understanding of the origin of

the foamability of a solution.
1.2 Methods for studying foams

1.2.1 Making foams The basic principle of producing a foam is to find a way of
incorporating a large volume of gas in a tiny volume of liquid (cf. §5.1, chap. 1).
It is possible to incorporate the gas by beating or whisking a solution (either

manually, or with an electric whisk). The gas/liquid interface is stretched and folded
in such a way as to force air bubbles to enter the liquid. The duration and speed
of the beating controls the homogeneity of the foam obtained. With electric mixers,
large quantities of homogeneous foam may be obtained, with a range of possible liquid
fractions (depending on the amount of solution) between 0.03 and 0.5. In general, the
bubble size distribution is large.
By bubbling gas into a solution it is possible to generate bubbles between typically
0.2 and 10 mm in size. These foams are usually dry (liquid fraction of the order of 10−2
to 10−3 ), because drainage occurs while they are being produced (see §4, chap. 3). The
rate of foam production is limited by the frequency of bubble detachment (typically
1 L.min−1 ). To generate foam more quickly (more than 10 L.min−1 ), it is possible to
simultaneously inject gas and liquid at high pressure through a porous frit, a mesh, or
a constriction, inducing a turbulent flow of the resulting mixture. This process of foam
production is therefore known as turbulent mixing, and, since it requires no moving
parts, it is remarkably robust. Micro-fluidic devices are similar, but do not require
turbulent flow [38, 53]; instead, they produce equal-sized bubbles (or droplets) one at
a time at a rate of the order of several mL.min−1 .
Inside a can of shaving foam, a gas (usually a hydrocarbon or a chlorofluorocarbon)
is liquified under high pressure and emulsified with the surfactant solution. When the
button is pressed, the pressure is released slightly and each hydrocarbon drop becomes
a bubble that is much larger. The excess volume then leaves the can in the form of
a foam. This method produces very homogeneous foams, with bubble diameters of
about 40 μm. The properties of the gas are crucial: in particular it must be liquifiable
under modest pressures (a few atmospheres).
Other techniques exist, for example in situ gas production through a chemical
reaction (as in the case of solid foams made from polyurethane or aluminium [5], or
bread).
1.2.2 Foamability tests In order to test the foamability of a solution, the simplest
method is to shake a flask containing a small amount of the solution and to see if a
foam forms and, if so, for how long it lasts. This is often formalized in a tumbling
tube: a sealed vertical cylinder rotated perpendicular to its axis for a fixed time.
These methods are an easy way to distinguish those solutions which don’t foam from
those that foam very well.
In order to obtain reproducible and quantitative results, a common laboratory test
consists of injecting a constant and controlled flow of gas into the solution (fig. 5.5a).
The quantities measured are the volume of foam produced, and the characteristic time
with which the foam disappears when the bubbling stops. If Vg denotes the volume
of gas injected, poor foamability is characterized by a low value of the ratio Vfoam /Vg
(whatever the temperature T ), indicating that not much gas is trapped in the foam.
When this ratio is close to 1, almost all of the gas has been incorporated into the
foam, indicating quasi-optimal foamability. The size of the bubbles can be changed by
varying the geometry of the gas injector, making it possible to study foamability as a
(a) (b) (c)
H(t)
H(t)
Fig. 5.5: Methods for measuring foamability: (a) bubbling; (b) Ross–Miles method;
(c) commercial apparatus with a mechanical mixer (reproduced by kind permission of
SITA Messtechnik GmbH, Germany).
function of bubble size. It is often observed that, for a given surfactant concentration,
smaller bubbles lead to better foamability.
In the popular Ross–Miles method (fig. 5.5b), foaming solution is allowed to flow
into a beaker from a graduated pipette, fixed at a given height. The foam volume is
measured as a function of time while the solution flows, after it stops, and for different
drop heights.
Difference commercial devices are also now available (fig. 5.5c), which make a
foam with a rotating blade or rotor (often very rapidly, at several hundred revolutions
per minute) in a reservoir containing the solution. The quantity and homogeneity of
the foam produced can be characterized by optical (§1.2.3, §1.2.5) or electrical (§1.2.6)
techniques.
1.2.3 Surface imaging The bubbles at the surface of a foam are easy to observe,
with a magnifying glass or a microscope if necessary. By measuring the diameter of
these surface bubbles, we can characterize the distribution of their sizes and their
average diameter, and by carrying out such measurements at different places, we can
evaluate the homogeneity of the sample.
A photograph of a foam in contact with a transparent wall shows the network
of Plateau borders around the surface bubbles (fig. 5.6a). Changing the lighting
conditions (for example using a diffuse or a point source, either in reflection or in trans-
mission) changes the apparent width of the Plateau borders and the apparent diameter
of the bubbles. Geometric optics and ray tracing can be used to explain this by linking
(a) (b) Observer
W
Wapp
Wall
Wapp θc
γ β
Plateau border
α
Light diffuser
Light source
Fig. 5.6: Surface imaging. (a) Photograph of a monodisperse 3D foam under a glass
plate, illuminated by a uniform source of diffuse transmitted light. The diffuser is the
foam itself. Each Plateau border in contact with the plate appears demarcated by
two dark lines. Photograph courtesy of S. Cohen-Addad. (b) Schematic diagram of a
Plateau border in contact with a transparent wall, illuminated in transmission by a
uniform diffuse light source; α, β, and γ represent several possible optical paths (see
text). The apparent width Wapp of a Plateau border on an image is less than its actual
width W (cf. eq. (5.2)).
the characteristics of the images to the actual dimensions of the Plateau borders and
the bubbles [56]. Then illumination with a uniform source of diffuse transmitted light
shows each Plateau border in contact with the wall as a clear band, demarcated by
two dark lines (fig. 5.6a).
In order to understand a photograph such as this, we construct optical paths to
the observer that obey the laws of geometric optics; see, for example, α, β, and γ in
fig. 5.6b. By traversing these paths in reverse, i.e. from the observer, it is possible
to identify those which originated at the light source. These rays thus establish the
regions that the observer perceives as illuminated. For example, ray α passes far from
the middle of the Plateau border and crosses the interfaces at an angle of incidence
close to 90◦ . It is therefore only weakly deflected and effectively links the observer to
the light source. The paths which cross the wall close to the centre of the Plateau
border are more complex, since they may be reflected or refracted. If, like rays β
and γ, they meet the gas/liquid interface at an angle of incidence equal to or slightly
greater than the critical angle θc for total reflection, they cannot have originated in
the light source. These paths define, on the image viewed by the observer, two dark
lines on either side of the centre of the Plateau border. A geometric calculation gives
the expression for the apparent width of the Plateau border Wapp as a function of its
actual width W [56]:
Wapp = W (1 − sin θc ). (5.2)

At an air/water interface, θc is 48.8◦ , giving Wapp = 0.75 W . In the presence of

surfactants, θc can vary slightly depending on their nature and their concentration.
In the same way, the image of a spherical soap bubble (for example in a very wet
foam), when illuminated by a uniform source of diffuse transmitted light, shows a dark
ring of apparent outer diameter dapp = d cos(θc /2), where d is the actual diameter of
the bubble [56].
1.2.4 Bulk imaging

Optical methods When a foam is sufficiently dry that it appears transparent, it
is possible to observe bubbles deep within a sample using a microscope or magnifying
glass. Remarkably, Matzke [41] counted the number of faces and edges of a large num-
ber of bubbles in this way, and deduced statistical information about the structure of a
foam (cf. §3, chap. 2). Microscopy is sometimes associated with optical tomography [42]
or even stereoscopy [28] (fig. 5.7).
Multiple light scattering, due to reflections and refractions at the liquid/gas in-
terfaces (cf. §1.2.5), prohibits observation of the internal structure of a foam sample.
The different types of 3D microscopy (confocal, two-photon, or structured illumination
microscopy, for example), although common in biology, are still not widely used for
foams.
X-ray tomography Foams scatter X-rays much less than visible light, and this is
exploited to image the bulk of a foam by tomography. Synchrotrons provide a source
of X-rays that is even bright enough to image wet foams [36]. A small volume of foam
(a) (b)
Time
Fig. 5.7: 3D optical imaging. (a) Structure deep within the foam, obtained by op-
tical tomography. The foam is illuminated by a laser sheet scanner and imaged in
successive slices, and then its structure can be reconstructed in 3D. The image shows
one of these slices, with circles indicating the vertices in this slice and squares/crosses
indicating the vertices in the slice above/below, respectively. Photograph courtesy of
C. Monnereau, reproduced by kind permission of [42], c 2001 The American Physical
Society. (b) Coarsening tracked by optical stereoscopy, showing the disappearance of
a bubble with four faces over time. Photographs taken from two different angles of
observation (only one of which is shown here) allow the 3D coordinates of the bubble’s
four vertices to be located precisely. Photograph courtesy of S. Hilgenfeldt, reproduced
from [28], c 2001 The American Physical Society.
(of the order of a cubic centimetre) is placed in a parallel horizontal beam of X-rays in
front of a detector with typically 1,000 × 1,000 pixels or more. The resulting 2D image
is similar to a medical X-ray: the X-rays are slightly absorbed by the liquid phase but
the gas phase has no effect on them. The grey level of a pixel is therefore proportional
to the total amount of liquid crossed by the beam along its straight path through
the sample. The sample is turned about a vertical axis and roughly one thousand 2D
images are recorded at different angles. From these images a mathematical algorithm
is applied to reconstruct the 3D structure of the foam. The time required to create a
high-resolution 3D image (fig. 5.8) is currently a fraction of a second.
The 3D images consist of elementary cubes, called voxels (in analogy with pixels in
a digital 2D picture), with sides of the order of 10 μm in length, which is sufficient to
clearly resolve the Plateau borders. The thin films must be reconstructed a posteriori
after segmentation (see §3.1.1). It is then possible to record for each bubble its faces,
its volume, and its evolution in time, for example to follow foam coarsening (§3.2,
chap. 3).
Imaging by nuclear magnetic resonance Nuclear magnetic resonance (NMR)

probes a material by analysing the electromagnetic signal emitted by the spins of
atomic nuclei when they are placed in an intense stationary magnetic field and excited
by an appropriate radio-frequency wave. The signal emitted depends both on the
nature of the material probed and the excitation characteristics. With magnetic field
gradients, it is possible to excite each point in space differently, and deduce the nature
and the position of each element of material from the signal: this is the principle
of medical magnetic resonance imaging (MRI), which has been used to study foam
coarsening [24].
If the material moves, it “experiences” different values of the magnetic field, and
the signal received depends on the field experienced by each material element along
Fig. 5.8: X-ray tomography: 3D image of a foam with liquid fraction φl = 0.15. Each
bubble is identified (see fig. 5.18) and coloured in a different shade of grey. Image
courtesy of J. Lambert; data from [36]. See Plate 12.
the whole of its trajectory, and therefore, inter alia, its velocity. The exploitation of
this information allows us to generate a 3D map of velocities within the material, and
thus to probe the flow of foams at the local scale [49] (cf. §4, chap. 4). To reduce the
acquisition time, only a single component of the velocity is measured, or the region
studied is limited to a single plane.
1.2.5 Multiple light-scattering We describe here how to take advantage of the

fact that, due to reflection and refraction at the many liquid/gas interfaces, foams scat-
ter light strongly. By characterizing the propagation of light through a foam (eq. (5.3)
below), we can determine the average bubble size if the liquid fraction is known, and
vice versa (eq. (5.6)). Furthermore, a coherent light source creates interference patterns
which provide details about the dynamics of topological changes (eq. (5.7)).
Transport mean free path Geometric optics describe the propagation of light
within a foam consisting of bubbles that are larger than the wavelength of light. When
a light ray meets a gas/liquid interface, it is partly refracted and partly reflected.
By following each of the two possible trajectories until they meet another interface,
and repeating, we can construct a complex tree of rays. The “memory” of the initial
direction of propagation is lost after a few refractions and/or reflections, over a char-
acteristic distance known as the transport mean free path, denoted ∗ . We model the
propagation of a light ray through a foam as an effective random walk, consisting of
paths ∗ in statistically independent directions.
In a very wet foam, the distance ∗ is comparable to the size of a bubble: in fact,
the liquid between two bubbles acts as a lens which strongly deflects a ray. On the
other hand, the thin films between neighbouring bubbles in a dry foam do not deflect
the transmitted rays, and ∗ greatly exceeds the bubble size. In this description, the
only characteristic length-scale is the bubble size and so we expect that, at given liquid
fraction φl , ∗ varies like the average bubble radius RV . In fact, experimentally [57],
RV
∗ = 2 √ . (5.3)
φl
The idea of a random walk occurs in other physical phenomena, such as the Brownian
motion of a molecule in solution. At the macroscopic scale, this motion is described in
terms of a concentration which varies in space and time according to Fick’s law [1]. By
analogy, we will show that light transport across an opaque medium is described by
a light energy density obeying the same laws, which enables the propagation of light
across a sample of multiply scattering foam to be modelled.
Diffuse light transmission Diffuse transmission spectroscopy (DTS) measures the
diffuse light intensity transmitted through a foam by illuminating a strip of foam with
a collimated light beam of intensity I0 , and measuring the total light intensity I which
crosses the sample. The transmission factor T r = I/I0 is linked to the thickness of the
strip L and to the transport mean free path ∗ . In the absence of optical absorption
in the sample, and in the multiple scattering regime (L ∗ ), we have [57]
1 + ze
Tr = . (5.4)
L/∗ + 2ze
The extrapolation parameter ze , of order one, characterizes the conditions at the op-
tical limits between the sample and the transparent plates which delimit it. For a foam
confined between two glass plates, we have [57]
ze = 0.74 + 3.44 φl . (5.5)
In the limit L ∗ , eqs. (5.3) to (5.5) imply that, omitting prefactors,
∗ RV
Tr ∼ ∼ √ . (5.6)
L L φl
A foam becomes more opaque as it becomes wetter, as the bubbles become smaller,
or as the sample becomes thicker. A measurement of optical transmission enables the
average bubble size to be determined for a known liquid fraction, and vice versa. As
this measurement is instantaneous and non-destructive, it is appropriate for measur-
ing the temporal evolution of the average bubble size during coarsening (fig. 3.27) or
of the liquid fraction during drainage (fig. 3.29). Relative measurements, compared
to a given bubble size or a given initial liquid fraction, are easy to perform; for ab-
solute measurements, the technique must be calibrated using a sample with known
diffusivity.
Dynamic light scattering Very often, we seek statistical information about
the dynamics of topological changes within a foam, whether induced by coarsen-
ing (§3.2.1, chap. 3) or by an imposed mechanical perturbation (see §4.2.1, chap.
4 or §4.2.3, chap. 4). This information can be obtained by a dynamic light-
scattering technique known as diffusing-wave spectroscopy (DWS), which probes the
local dynamics in turbid media such as foams, concentrated emulsions, and pastes
(fig. 5.9) [60].
In DWS, the sample is illuminated with a coherent light beam and either the
transmitted or back-scattered light is recorded on a screen placed close to the sample.
A pattern of “speckles” (regions of different light intensity) is observed (fig. 5.9a). The
speckles are the result of interference between all the light rays arriving at the screen.
(a) (b) g2(Δt)

I(t)
1
Laser
PM Correlator
Time t
0
Turbid Delay time
medium
Fig. 5.9: Diffusing-wave spectroscopy (DWS). (a) Speckle pattern. The scale bar cor-
responds to an angle of 0.02 mrad. Photograph courtesy of S. Cohen-Addad. (b) DWS
(in transmission) is used to probe the dynamics in a turbid medium: the speckle in-
tensity is measured using a photomultiplier, and the autocorrelation function g2 (Δt)
is calculated using a signal correlator.
A change of structure in the sample, for example a T1, modifies the phase of those
rays which cross the rearranged zone, and changes the intensity of the speckles, which
appear to “twinkle”. To quantify this, we measure the intensity of the speckle pattern
I(t) as a function of time, and calculate its (normalized) autocorrelation,
I(t)I(t + Δt) − I(t)2

g2 (Δt) = 2 . (5.7)
I(t)2 − I(t)
By taking an average, shown with angle brackets, we avoid having to choose a

particular speckle or a particular bubble structure with fixed disorder. There are
two equivalent ways of taking the average: if the internal dynamics of the sample
(due to coarsening or flow, for example) modifies the structure over time, we use
a photomultiplier to measure the intensity of a single speckle, and we average its
autocorrelation over time t (fig. 5.9b); if, on the other hand, the dynamics are slow
or non-ergodic, we use a “multispeckle” technique instead, in which a large number of
speckles is recorded simultaneously with a CCD camera [4, 12, 13], and the correlation
between their respective intensities is averaged at instants t and t + Δt.
In the limit Δt → 0, g2 (Δt) tends to 1, indicating that over short time intervals
the structure of the foam is unchanged. As Δt increases, the internal dynamics in-
creasingly modify the structure, due for example to T1s, and the correlation between
the structures at times t and t + Δt is lost.
Statistical models of light propagation and topological changes permit a more
quantitative analysis. For example, the characteristic time over which g2 (Δt) decreases
in a coarsening foam is proportional to the average time interval τ between T1s in a
given volume [19], which increases as coarsening proceeds (fig. 3.25). It is also pos-
sible to infer the fraction of the sample that undergoes T1s per unit time, for example
during shear [27], to measure the duration of a T1 [4], or even to characterize the
intermittency of the dynamics [12].
1.2.6 Conductivity The electrical properties of a foam are strongly dependent

on its liquid content. Given a relationship between the electrical conductivity and φl ,
measuring the former would allow us to infer the latter.
In the very wet limit, the bubbles separate into isolated spheres. The relative
conductivity Λ of this bubbly liquid, defined as the conductivity of the sample divided
by that of the foaming solution, is given by the Maxwell relationship [21]:
2φl
Λ= . (5.8)
3 − φl
In the dry limit, the foam consists of a network of randomly oriented, conduct-
ing Plateau borders, in which the resistance varies as the inverse of φl according
to eq. (2.38). The relative conductivity of the foam is now given by Lemlich’s
relationship [21]:
φl
Λ= . (5.9)
3
0.8
Relative conductivity
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
Liquid fraction
Fig. 5.10: Relative electrical conductivity as a function of liquid fraction [21]. The
experimental measurements (symbols) obtained for foams and suspensions are well-
described by eqs. (5.10) and (5.11) (thick line) over the whole range of liquid fraction.
Eq. (5.9) (thin dotted line) is a linear approximation valid for liquid fractions of up to
a few percent. See [21] for further details.
Measurements of the conductivity of foams and suspensions for liquid fractions

between 0.02 and 0.9 suggest that it is continuous over the whole range of φl , as
shown in fig. 5.10. The empirical expression (a one-parameter best-fit to the data in
fig. 5.10)
2φl (1 + 12φl )
Λ= (5.10)
6 + 29φl − 9φ2l
describes all the data well and agrees with the predictions of Lemlich and Maxwell in
the two limits [21].
Conversely, the liquid fraction can be expressed in terms of the relative conductiv-
ity as
3Λ(1 + 11Λ)
φl = . (5.11)
1 + 25Λ + 10Λ2
Neither average bubble size, polydispersity, or interfacial properties affect these re-
lationships. Thus, no matter what the system (foam, emulsion, suspension of a
non-conducting phase) or its physico-chemistry, the relative conductivity depends only
on the degree of dispersion. Consequently, given the conductivity of the solution, which
depends on the electrolyte concentration and the electrode geometry, a conductivity
measurement of a sample of foam gives direct access to its liquid fraction (see fig. 3.29
for example).
This conductivity measurement is simple to use: from the electrical point of view,
the foam behaves as a circuit consisting of a resistor and a capacitor in parallel.
The measurements are carried out with an alternating electrical voltage, typical 1 V
oscillating at 1 kHz (see experiment §7.4, chap. 3).
1.2.7 Rheometry To study the rheological behaviour of a 3D foam it should be

probed with the simplest and most well-defined perturbations possible. Once a con-
stitutive law relating stress, strain, strain-rate, and structural parameters (cf. chap. 4)
is established, it should be possible to predict the flow in any given geometry.
Methods To impose a controlled deformation on a foam, we inject the foam into

the cell of a rheometer, which measures the strain under imposed stress or the stress
under imposed strain. The choice of cell geometry depends on the information required;
the characteristics of the strains and stresses which the sample undergoes in each
possible geometry are described in fig. 5.11. A detailed presentation of the methods
of rheometry can be found in [39], for example.
The four geometries “sliding plate”, “parallel plate”, “cone-plate”, or “cylindrical
Couette” are all a priori suitable for measuring the linear response, under the condition
that the gap size is much greater than the bubble size (typically at least 20 bubbles).
To study the non-linear regime, a geometry for which the stress is the same everywhere
within the gap is preferable, as in the sliding plate, cone-plate (if the cone angle is
small), or even the cylindrical geometry (if the gap is small compared to the radius of
the cylinder). By measuring the force exerted by the sample perpendicular to the plate,
we obtain the first normal stress difference which accompanies the shear (cf. §4.1.2,
chap. 4). Finally, to study the flow of a foam under the effect of a pressure gradient,
we use a capillary rheometer.
We have seen in §4.3.2, chap. 4 and fig. 2.29 that the flow of a foam is sometimes
heterogeneous, or localized, whether or not the stress is homogeneous. Under these
conditions, it is not possible to deduce the local strain from the motion of the walls,
and it is necessary to observe the sample itself, for example optically or with MRI
(cf. §1.2.4).
Wall slip A frequent problem in foam rheometry is the slip of the sample at the
walls of the rheometer [40], which are covered by a film of the foaming solution (§3.3.3,
chap. 4). The degree of slip depends on the material from which the wall is made:
plastic is often more slippery than steel. Wall slip is distinct from shear banding,
which can occur in the bulk, far from the walls. It is often straightforward to detect
wall slip experimentally, since the parameters in the constitutive law appear to depend
on the thickness of the sample or the size of the rheometer. It distorts the analysis
of the rheological measurements since the motion of the walls or the pressure drop is
greater than what is actually experienced by the foam. To eliminate it, it is necessary
to roughen the walls with (water-resistant) sandpaper or by etching grooves in the
surface, perpendicular to the direction of flow; the grain of the paper or the depth of
the grooves must be greater than the bubble diameter.
Sliding plate
Area S
Parallel plate
Cone-plate
Cylindrical Couette
Capillary rheometer
Fig. 5.11: Schematic diagrams of the different rheometer geometries used in foam
rheology. The variables are strain ε, strain rate ε̇, shear stress σ, force F and torque M,
velocity V and angular velocity Ω, tangential fluid velocity vθ , pressure drop ΔP , and
volumetric flow rate Q.
2 Numerical simulations
In this section, we describe the main numerical tools used to simulate the static struc-
ture and rheology of foams. Some have high levels of precision while others generate
statistics for large numbers of bubbles.
Numerical simulations are an important tool for both the engineer and the re-
searcher. Whilst an analytical approach is often useful for ordered foams, simulations
are usually necessary to understand disordered foams, given the larger number of
parameters involved: vertex positions, edge lengths, and curvatures, and, in 3D, the
curvature of the faces. They enable foam behaviour to be predicted without actual ex-
perimentation, and allow the different physical parameters to be varied independently,
something which is often impossible experimentally, for example to ignore gravity or
gas diffusion, and even control polydispersity. Moreover, some quantities, such as the
fields of pressure or local velocity, and film curvatures, which are difficult or impossible
to measure experimentally, are accessible and it becomes possible to explore parameter
ranges beyond the reach of experiments.
The simulation techniques presented here are adapted either to predicting the
foam’s static (equilibrium) state, or its flow. Even then, the choice of simulation tool
can depend on the significance of dissipation in the problem studied, the range of liquid
fraction, the dimension (2D or 3D), and the compromise sought between obtaining the
fine detail of the structure, the number of bubbles to be considered, and the calculation
time.
2.1 Predicting static structure

To obtain the structure of a foam at equilibrium, i.e. the shape and pressure of
the bubbles, it is necessary to satisfy Plateau’s laws and the Young–Laplace law
everywhere [8], or, as we describe here, find a local minimum of energy.
Starting from an arbitrary initial configuration in which the bubbles are connected
and tessellate space, for example a Voronoi tesselation [58] of randomly scattered
points, the interfaces are deformed so as to minimize the energy, respecting the volume
of each bubble and any further constraints at the boundaries. During the search for a
minimum, the topology of the initial tessellation may be modified by one or more T1s.
We present below the two most popular methods for predicting the equilibrium
shape of bubbles and their pressures. Although designed with dry foams in mind, they
can be adapted to the wet case. They are in principle equivalent and both are freely
available on the Internet, with each having different advantages.
2.1.1 Potts model The Potts model [30] searches randomly for an energy min-
imum. This enables a minimum, which may be very different from the initial
configuration, to be found rapidly, even for a large number of bubbles. Developed
by R.B. Potts in 1952 for statistical physics, it has been used to simulate the grains
(microcrystals) in crystalline materials [51] as well as foams [29]. It forms the basis of
the free software CompuCell [30], for which extensive training is available.
It is a lattice model: as in a digital photograph, a foam is represented by small
surface elements (pixels in 2D) or volume elements (voxels in 3D) with fixed positions.
Numerical simulations 243
Fig. 5.12: Potts model simulation [26], here in 2D. (a) Bubble i consists of all pixels
which are labelled with the number i, with i running here from 1 to 7. (b) Polydisperse
foam at equilibrium. The shading indicates the number of sides of a bubble. 589 bubbles
on a lattice of 256 × 256 pixels, with periodic boundary conditions to left and right
and hard walls at the top and bottom.
Bubble i is defined as all those pixels labelled with the same number (or index) i.
In a dry foam, each pixel belongs to only one bubble, so the bubbles tessellate space
(fig. 5.12). The interface between bubbles i and j is exactly those sites where a pixel
labelled i neighbours a pixel labelled j. The energy of the foam is equal to the contact
area (i.e. the area of the interface between sites with different indices) multiplied by
2γ. Additional terms in the energy functional (see exercise 4.2) ensure conservation of
the bubble volumes.
The energy minimization is sequential : a pixel located at the edge of bubble i is
chosen at random and associated with the neighbouring bubble j, i.e. it takes the index
of one of its neighbours. If this operation lowers the energy, it is retained (otherwise
the pixel is returned to bubble i). In this case, it reduces the size of bubble i and
increases the size of bubble j and moves the interface between them (a small variation
in bubble volume is therefore permitted). It can happen that this puts bubble j into
contact with a new bubble k: in other words, it may cause a T1. This operation
is repeated thousands of times, and the total energy gradually decreases to a local
minimum.
The model also allows the effect of an external force to be incorporated. It also
permits, with a certain probability (which decreases when the energy cost increases),
that a pixel can change bubble even if this increases the energy. This allows the
exploration of different energy minima and means that the model is appropriate
for non-deterministic systems with fluctuations, for example physical systems at a
temperature which is significantly greater than their energy variations (divided by
Boltzmann’s constant) or biological cells [30]. Finally, it allows the simulation of a
million 2D or 3D bubbles [55] (see fig. 3.24), with a degree of precision of the order
of one percent of the value of the interface area (and thus of the energy) [29].
Fig. 5.13: Surface Evolver simulation [31]. (a) Starting from an initial configuration
(on the left), each face of each bubble is divided into small triangles that move in
order to decrease the total energy, until a local minimum is found (on the right).
(b) Example of an ordered monodisperse foam in 3D, the Weaire–Phelan structure.
This is the structure with the least surface area per unit volume known (see shaded
section on p. 37) [59]. Six barrel-shaped bubbles touch each other on their hexagonal
faces. They are therefore aligned in columns along three perpendicular axes. The gaps
are filled with pentagonal dodecahedra (the two bubbles at the far right of the figure).
2.1.2 Surface Evolver software The Surface Evolver (fig. 5.13) seeks an energy
minimum in a deterministic way [9]. Developed with grain growth and minimal surfaces
in mind, it has also been used successfully in other fields in which surface energy is
important. Its creator, K. Brakke, has been providing updates and documentation for
free on his Internet site [31] since 1992.
The surfaces of the bubbles are discretized into a mesh of small triangles which
determines the precision; it can be refined as the calculation proceeds, allowing the
shape of each bubble to be described with a level of precision as high as desired
(see for example fig. 2.17). In the case of 2D foams, the surface elements are the
edges themselves, which can be represented as arcs of circles and do not need to be
subdivided.
The variables of the problem are the attributes of these surface elements (vertex
position, edge curvature, etc.), generating a parameter space of very large dimension.
The software calculates the energy gradient in this vector space and determines the
direction of greatest energy decrease. All the elements are then moved in this direction
together: the entire structure is affected simultaneously. If during this process the
interface between two bubbles becomes smaller than some given critical size, a T1 can
be triggered, the new surface tesselated, and the minimization continued.
The calculation converges when the energy varies less than some specified amount.
It is necessary to strike a compromise between the required precision (mesh size,
convergence criterion, critical interface size before triggering a T1) and the calculation
time. The software is used for simulations of up to a thousand bubbles in 3D [35].
2.2 Predicting dynamics

When the dynamics are slow, a foam passes through a succession of equilibrium states
(§2.2.1); we can study this using the static foam simulations described above. If the
flows are more rapid, then the foam is not at equilibrium. Its structure is determined
by the competition between three forces: surface tension, pressure forces, and viscous
forces, which are functions of the local velocity gradient. For very rapid flows or wet
foams, inertia must also be taken into account. We present below a model that works
in both 2D and 3D (§2.2.2), and two models limited to 2D (§2.2.3, 2.2.4).
2.2.1 Quasi-static models Potts model and Surface Evolver can both be used
in 2D or in 3D to simulate slow dynamics, either in coarsening (§3.2, chap. 3), where
the bubble volumes evolve slowly, or quasi-static flows (fig. 5.14).
The foam structure is first driven to an equilibrium state determined by the initial
conditions, which may include, for example, the position of the walls which contain
the foam. To simulate the flow induced by the movement of these walls, for example,
the walls are moved step by step and a new minimum of energy is sought at each step
(fig. 4.22b). The evolution of the structure towards equilibrium during each relaxation
is of no physical significance: only the successive equilibrium states are relevant.
In Surface Evolver, a T1 is triggered when a face separating two bubbles, or an
edge, becomes smaller than a critical value, chosen in advance. With a higher critical
value, the T1s happen more easily, providing an indirect way of taking into account
the liquid fraction [48] (cf. fig. 4.30 and §3.3.2). In the Potts model, it is the pixel
size which determines the critical value [48].
2.2.2 Bubble model The bubble model, first described by Durian in 2D [18], and
later extended to 3D [23], enables viscous dynamics to be probed. This approach is
more phenomenological than the preceding ones, giving a fast method of calculation
which does not respect the exact shape of the bubbles. The structure is determined
by the position of the centres of spherical bubbles of given, constant, diameter. The
bubbles can overlap, and interact via a repulsive contact force, proportional to their
overlap, and through frictional forces between neighbouring bubbles, proportional to
(a) (b)
Fig. 5.14: A dry monodisperse 2D foam flowing from left to right around an obstacle.
Good agreement is obtained between these two quasi-static simulations based on dif-
ferent algorithms [48]. (a) Surface Evolver, with an initially disordered foam; the grey
bubbles are those which initially form a straight line at the entrance to the channel.
(b) Potts model, with an initially ordered foam; the grey bubbles are those which have
5 or 7 sides.
Fig. 5.15: Bubble model simulation of a 2D foam under shear. The bubbles are rep-
resented by slightly (in this wet foam) overlapping circles: greater overlap represents
greater bubble deformation for given liquid fraction, and lower liquid fraction in gen-
eral. The bubbles cut by horizontal lines are fixed and move in the direction of the
arrows to impose a shear on the rest of the foam. The dotted and continuous circles
correspond to two successive instants in time, and the line segment at the centre of
each bubble represents its motion. In this example, a T1 takes place towards the top
right corner, where significant motion can be seen, even though the movement of the
walls is imperceptible (reproduced from [18],
c 1997 The American Physical Society).
their relative velocity (fig. 5.15). These frictional forces do not take into account the
flow in the liquid phase (see §3.3, chap. 4).
From a given initial state, the forces exerted on each bubble, which depend uniquely
on the (known) positions of the centres and the (unknown) relative velocities, are
calculated. Since inertia is neglected, the sum of these forces must be zero, which
enables the velocity of each bubble to be determined. The bubbles are then moved,
and the process repeated. It is a similar process to molecular dynamics, which is used
to predict the motion of a group of atoms [2].
2.2.3 Viscous froth Model The flow of a 2D foam between two plates is special
because of the additional contribution to the dissipation from the plates (see §5.4,
chap. 2, fig. 2.29, and also [11]): the friction forces are functions of the velocity
(and not the velocity gradient as in 3D) and are relatively well understood (see §3.3,
chap. 4).
There are many different simulations of the dynamic behaviour of a 2D foam. The
most precise ones exploit a full discretization of the structure (viscous froth model [34]):
each edge is represented by a certain number of elementary line segments, acted on by
the pressure in adjacent bubbles, by the surface tension, and by a friction force at the
plates, proportional to velocity. As for the bubble model, the sum of forces on each
segment is zero, which gives the velocity of each segment. If this motion leads to the
disappearance of an edge, a T1 is triggered (see §2.2.1), allowing the prediction of the
effect of frictional forces on T1s (fig. 5.16).
Fig. 5.16: Movement of a 2D foam along a U-shaped channel simulated with the
viscous froth model, which includes the friction with the plates bounding the foam.
Lower arrow: at low velocity the foam moves as a plug and the bubbles keep the same
neighbours. Upper arrow: above a critical driving velocity, the bubbles on the inside
of the bend move more quickly and a T1 occurs [34].
2.2.4 Vertex model For 2D flows with a large number of bubbles, further sim-
plification is necessary. One possibility is to describe the foam by the vertex positions
and their connectivity (vertex model [46]). The edges are approximated by straight
lines (fig. 5.17) and the dynamics are reproduced with an algorithm similar to the
one above.
Fig. 5.17: Simulation of the flow of a foam between two plates using the vertex model.
The foam is monodisperse and disordered with one defect: one bubble larger than the
others. There are periodic boundary conditions and the foam moves to the right at a
given rate. This simulation validated the theoretical prediction of the velocity of the
large bubble, which is higher than the average (see §5.4, chap. 2) [10].
3 Methods of image analysis

An image of a foam, particularly in 2D, is remarkably informative as it contains almost
all the physical information required (except for the viscous dissipation). In general, it
is necessary to start by treating the image in some way (§3.1), identifying the bubbles
and their boundaries for example, although images from numerical simulations can be
used directly, without treatment. The procedures for gathering data from the treated
images (§3.2) are often the same for experiments and simulations. The 2D case is
simpler (§3.3).
3.1 Image treatment

The image obtained from a camera is a measure of the grey level in each pixel. If a foam
is illuminated and observed from the same side (reflection), the liquid appears brighter
than the gas (fig. 5.18a); if illuminated from behind (transmission), the reverse is true.
To make the most of the images, it is necessary to first segment them, that is, to
identify individual bubbles. Various methods, adapted to either 2D or 3D, allow the
recognition of bubbles in spite of low contrast, noise, poor resolution, or variations in
intensity. Here we consider only the case of high-quality 2D images, the easiest to treat
and analyse (§3.3.1).
3.1.1 Segmentation For an image with good contrast and many pixels per
bubble, segmentation can be automated with the following steps:
• Pre-treatment: compensate for slight variations in the lighting and in the
background, for example by filtering the image to reduce this noise.
• Thresholding: grey-levels greater than a given value (dependent on the image
being treated), called the “threshold”, are turned white (the liquid phase in
fig. 5.18b); the others become black (the gas phase in fig. 5.18b).
• Dilation (predominantly for dry foams): sometimes a bubble is not completely
surrounded by “white” and therefore appears to be connected to a neighbour.
Fig. 5.18: Main steps in image treatment, here in 2D: (a) raw image (courtesy of I.
Cantat); (b) image after thresholding and dilation; (c) image after skeletonization
and segmentation. The liquid appears brighter than the gas.
Methods of image analysis 249
In this case, we assign the “white” value to the n nearest neighbours of each of
the white pixels; this disconnects the bubbles, and those bubbles smaller than n
pixels are lost.
• Skeletonization of the continuous (white) phase: the pixels at the edge of the white
region are changed to black as long as this doesn’t alter the topology. The final
image then consists of black bubbles surrounded by a white border exactly one
pixel thick (fig. 5.18c).
• Segmentation: identification of each individual bubble. The same number—the
number of the bubble—is attributed to all touching black pixels (fig. 5.18c); this
allows each bubble to be coloured and easily identified (see fig. 5.8).
3.1.2 Bubble tracking To characterize the kinematics of the foam (cf. §3.2.5),
a succession of images is required (a video), at time intervals Δt sufficiently short to
capture the evolution of interest. For example, an image every 30 min is sufficient to
follow slow coarsening, an image every second is enough for a quasi-static flow, and a
high-speed camera (up to 1,000 images per second) allows the study of rapid flows.
Even if the image is not segmented (i.e. the bubbles have not been individually
identified), subtracting one image from the next in the sequence gives a qualitative
visualization which is often sufficient. In this case, it is possible to shift the image so
as to find the optimal correlation with the next one, which allows the average velocity
field to be determined.
If we segment the image, we can also determine the individual speed of each bubble.
This requires that we follow each bubble from one image to the next, thus matching
a bubble k in image m with a bubble k in image m + 1. If each bubble advances less
than one diameter between two successive images, one method is to determine if the
centre of mass (xk,m , yk,m ) of the bubble k in the image m lies in bubble k in image
m + 1. If this is the case, then k and k are the two successive indices of this particular
bubble.
If the average velocity field v(x, y) is known, we can even follow the bubbles which
advance more than one diameter between two successive images by seeking the bubble
in image m + 1 that contains the point (xk,m + vx Δt, yk,m + vy Δt). Conversely, if we
follow each bubble, we can measure the velocity of its centre of mass, and thus average
these speeds, weighted by the bubble volumes (or by the mass of the liquid films) in
order to obtain v(x, y).
3.1.3 Basic data Ideally, for each bubble, a 3D image gives access to raw
data on:
(i) all its interior pixels, its centre of mass, its volume, and the local liquid fraction;
(ii) its number of faces, the face areas, and curvatures;
(iii) the number of edges per face, edge length, curvature, and cross-section;
(iv) the number of vertices and their position.
In all of this we have assumed that we know all the voxels of each bubble. In practice,
this is the case for a segmented image from X-ray or optical tomography or NMR (see
§3.1.1), and for Surface Evolver or Potts simulations. However, there are occasions
when this does not hold, for example tomography or NMR images of foams that
are too dry for the liquid films separating the bubbles to be seen, or vertex model
simulations.
3.2 Image analysis

How can we exploit all of the information available in an image? From the basic data
(§3.1.3) we can measure many quantities. We give here a non-exhaustive list.
Each measurement is carried out in response to a certain question. Depending
on the problem posed, we might measure the complete distribution of a quantity, or
content ourselves with its average or standard deviation.
3.2.1 Topology, disorder A bubble is characterized by its number of neighbours,

i.e. faces, f . This is used to determine the topological disorder of the foam, which is
a measure of how closely the different bubbles resemble each other. It is characterized
by the variance (dispersion of the values around the average) of the distribution of the
number of faces in the foam, μf2 = (f − f )2 , or equivalently μf2 = f 2 − 2f f +
f 2 = f 2 − f 2 . The resulting
! number is positive, and we often take its square root,
the standard deviation δf = μf2 , or its dimensionless form
"
δf f 2
= − 1. (5.12)
f f
2
Around its perimeter each face has a number of edges n, whose distribution also has a
standard deviation. It is another measure of topological disorder, this time for a single
bubble: the more irregular the bubble is, the greater δn/n becomes.
This sort of topological analysis is evident in the empirical Aboav–Weaire relation-
ship, which links the number of faces of a bubble f to the average number of faces of
its neighbours, f neigh (§3.1.4, chap. 2). It is observed that if f f neigh is plotted as a
function of f , the result is approximately a straight line (fig. 2.11):
f f neigh = (f − af ) f + af f + μf2 . (5.13)
Note that if the number of faces f of a bubble is large, then the average number of
faces of its neighbours is small. The parameter af is a characteristic of the foam, which
can be measured by multiplying eq. (5.13) by f and averaging over all bubbles in the
foam [33]:
1
af = f − f f 2 f neigh − f . (5.14)
μ2
3.2.2 Bubble size, polydispersity A further characteristic of a bubble is its size.
We have access to the volume V of each bubble from MRI, optical or X-ray tomography,
3D microscopy, simulations, and in 2D foams (in which case we measure the area A, see
§3.3). The bubble radius is available from other experiments, including light scattering
and 2D images of 3D foams (projection, radiography, slices, microscopy).
The radius has several possible definitions (as opposed to the volume V , which is
uniquely defined) and different experiments measure different radii. We can in principle
define several characteristic sizes of a bubble, for example the value RS = (S/4π)1/2
associated with the surface (the radius of a sphere with the same surface area S) or
RV = (3V /4π)1/3 associated with the volume (the radius of a sphere with the same
volume V ). For a bubble obeying Plateau’s laws, eq. (2.19) gives RS ≈ 1.05 RV .
Given a definition of the radius R, there are several ways to define a characteristic
length based on an average over the whole foam, using the moments of the distribution
of R. Thus, Ravg = RV is the average radius and Rrms = RV2 1/2 is the mean square
radius. More generally, we construct a family of characteristic lengths:
RVn
Rn,n−1 = (5.15)
RVn−1
where n is a whole number (see fig. 5.19). For example, R10 = Ravg and R21 =
2
Rrms /Ravg ; R32 , the Sauter mean radius, is used in rheological constitutive models
(§4.1.1, chap. 4) as well as models of osmotic pressure (§4.2, chap. 2). Their values
are close, and for a monodisperse foam (i.e. in which all the bubbles have the same
volume) they are equal.
The correspondence between measurements of radius and volume is not direct. In
particular, in a 2D slice of a 3D foam the apparent size of the bubble depends on the
position of the cut, so that bubbles of the same volume can appear to have different
radii. A monodisperse foam may thus show a polydisperse cross-section. Likewise, if we
see only the free surface of the foam or observe those bubbles that touch a (transparent)
Fig. 5.19: Different values of the average bubble size. The radius of each circle rep-
resents the value of the quantity indicated in the circle, defined according to eq. (5.15)
and measured for the polydisperse foam shown in the background. Photograph cour-
tesy of K. Feitosa, D. Durian [22], reproduced by kind permission of Springer Science
+ Business Media.
wall (fig. 5.6), the apparent distribution of bubble sizes doesn’t directly reflect the
actual volumes.
The geometric disorder is the (dimensionless) standard deviation of the distribution
of the bubble volumes, analogous to eq. (5.12):
"
δV V 2
μV2 = (V − V ) 2
and = − 1. (5.16)
V V
2
It is almost equivalent to considering the polydispersity in the bubble radii [35]:
#
$2/3
R32 RV3 V 2/3
p= −1= − 1 = # $ − 1. (5.17)
RV3
1/3 RV2 V 2/3
The size disorder is given by δRV /RV = (RV2 /RV 2 − 1)1/2 . If δV /V and
δRV /RV are small compared to 1, they are approximately related by δV /V =
3 δRV /RV . These three measures, δV /V , δRV /RV , and p, are zero for a
monodisperse foam and positive otherwise.
With regard to correlations, a large bubble often has smaller neighbours. This
negative correlation is characterized in a similar way to the topology (eq. (5.14)) by
the parameter
1 # 2 $
aV = V − V
V (Vneigh − V ) . (5.18)
μ2
3.2.3 Surface area, deformation It is important to know the total surface area
of a foam, especially because this gives its elastic energy (eq. (2.15)). It is difficult to
exploit however, as the differences in energy between the different possible arrange-
ments of a given group of bubbles are often too small to be detectable experimentally.
In order to resolve this, Surface Evolver simulations are often used to calculate the
ratio S/V 2/3 for each bubble. Then the regularity of the bubble can be assessed by
comparing it with a sphere or a regular bubble (eqs. (2.16) and (2.19)).
The deformation of a bubble can be characterized by the “inertia matrix”:
⎛# $ ⎞
x2 xy xz
⎜ ⎟
⎜ # 2$ ⎟
⎜ yx y yz ⎟ (5.19)
⎜ ⎟
⎝ ⎠
# 2$
zx zy z
where the average is over the coordinates (x, y, z) of all the pixels of the bubble (with
the origin taken at its barycentre). This matrix, with units of m2 , describes the prin-
cipal directions of deformation of the bubble and its anisotropy; the eigenvalues give
the directions in which a bubble is elongated.
The deformation of a foam (or a subregion) is characterized by the “texture” [25]:

⎛ # 2$ ⎞
X XY XZ
⎜ ⎟
⎜ # 2$ ⎟
⎜ Y X Y Y Z ⎟ (5.20)
⎜ ⎟
⎝ ⎠
# 2$
ZX ZY Z
where the average is over all vectors, with components (X, Y, Z), which connect the
barycentres of two neighbouring bubbles. This matrix, also expressed in m2 , describes
the principal directions of deformation and the anisotropy of a foam; there is an
equivalent definition in 2D (fig. 5.20).
When a dry foam is deformed, each bubble undergoes a similar deformation, so
the (intra-cellular) inertia and the (inter-cellular) texture are correlated. In a wetter
foam, each bubble becomes more rounded, so that its inertia is more isotropic while
the texture continues to reflect the deformation of the whole foam.
3.2.4 Forces, pressure, curvature Bubbles exert forces on each other which
originate from three sources: gas pressure, film tension, and viscous friction. Dynamic
quantities, like viscous stresses, cannot be measured directly on the image.
On the other hand, the image does allow access to geometric quantities, like the
contribution of pressures and films to the stress (eq. (4.18)), for which it is necessary
to know the surface tension γ. In 2D, it is sometimes possible to make direct meas-
urements (see §3.3.3), although this is usually difficult, because of sensitivity to the
pixellization of the images.
To determine the average pressure in a region, we measure the average of the bubble
pressures, weighted by their volume. The contributions of the films is determined by
integrating the last term of eq. (4.18) over each face, which requires knowledge of the
exact shape of the face (and therefore its curvature).
Fig. 5.20: Deformation of a foam, here in 2D. In (a), the foam is almost isotropic:
the texture of a bubble, calculated from its links with neighbouring bubbles, can be
represented by a circle (the two eigenvalues of the matrix in eq. (5.20) are equal,
denoted λ0 ). In (b), the foam is deformed: the texture is represented by an elongated
ellipse in which the axes are the two distinct eigenvalues, λ1 = λ2 . Images courtesy of
C. Raufaste, P. Marmottant, reproduced by kind permission of The European Physical
Journal [25].
To measure the curvature (which is also useful for determining the geometric
charge, eq. (2.23)), it is necessary to fit to the image a surface of constant mean
curvature. The fit may be improved by noting that the curvature arises from the pres-
sure difference between bubbles (eq. (2.4)), which implies relationships between the
curvature of adjacent faces (eqs. (2.5) and (2.24)). In other words, by determining
simultaneously the pressures and the curvatures best approximating the faces, we can
improve the accuracy. For N bubbles, and therefore N unknown pressures, there are
about N f /2 faces, and the same number of measured curvatures. Then there are
f /2 times more givens than unknowns, which improves the approximation to the
pressures and curvatures.
The structure of a foam observed in an experiment can be reconstructed in the
Surface Evolver. Using the fact that each face has constant mean curvature, the
pressures and curvatures can then be calculated.
3.2.5 Evolution in time Following individual bubbles over time allows, for ex-
ample, measurement of their velocity (cf. §3.1.2) and their growth rate (see §3.1,
chap. 3) defined at a fixed number of faces, because we can only track the bubble
in the absence of T1s.
To identify the topological changes, we generate the list of bubbles that share a face
(neighbours). If this list changes between two successive images, it is because there
has been a topological change, allowing us to count the number of T1s, T2s, or film
ruptures per unit time and relate them to the number of faces.
We can also measure how the texture changes (eq. (5.20)) between two images,
allowing us to treat all the topological changes simultaneously, and to study their
direction as well as their characteristic size [16, 25].
3.3 Image analysis, liquid fraction, and stress in 2D

The quantities defined in §3.1 and §3.2 each have an equivalent in a 2D foam. For the
topology and the geometry, it is sufficient to replace f with n, and V with A. There
are some features, however, which are specific to 2D foams.
3.3.1 2D image analysis In 2D, images are easy to obtain experimentally (cer-
tainly more so than in 3D), and especially to segment. On an image which shows
good contrast and homogeneous lighting (for example with a diffuse light source),
the bubbles can be individually identified (fig. 5.18). Even if the foam is very dry
(between two glass plates) the edges of the bubbles, and thus the faces, are visible be-
cause of the surface Plateau borders (fig. 5.6a). In simulations, the data is again often
more precise. At equilibrium the edges are circular arcs, which facilitates numerical or
analytical treatment.
3.3.2 2D liquid fraction Defining the liquid fraction in 2D poses considerable

problems (cf. §5.3, chap. 2). In a Hele–Shaw cell (fig. 2.25a) it is possible to define
the liquid fraction as in 3D, i.e. as the fraction of the volume between the plates that
the liquid occupies. But this depends on whether the walls are wetted or not by the
soap solution, and this definition is only weakly related to the physical properties of
Exercises 255
the foam. In the bubble raft (fig. 2.25c), or the liquid–glass arrangement (fig. 2.25b
and 2.25d), the total volume of the foam is not precisely defined, and therefore neither
is the liquid fraction.
In fact, in all realizations of a 2D foam, it is necessary to define and measure
an effective liquid fraction, appropriate to the problem under investigation. Thus for
measurements of flow, it is possible to establish a link between the gas and the liquid
flows. Or for the study of mechanical properties, we measure the distance between two
vertices at the moment when a T1 is initiated (§2.2.1) [48]. Finally, if we are interested
in what happens in the vicinity of a glass wall (friction, for example), there are optical
methods [56] for estimating the shape and size of surface Plateau borders (fig. 5.6).
3.3.3 2D stress The 3D surface tension depends only on the composition of the
soap solution. In 2D it is replaced by the line tension, which also depends on the bubble
geometry (cf. §5.2, chap. 2): it must be measured directly on each occasion, which is
not always easy [14, 48].
Foams in a liquid–glass experiment probably constitute the unique (experimental)
situation where the pressure in each bubble can be measured directly. If the pressure
increases in a bubble, it descends slightly into the liquid phase without noticably
changing its volume. The change in surface area, viewed from above, is small but it
is nonetheless sufficient to deduce by image analysis the change in pressure from its
relaxed state (fig. 2.28c) [16].
In a very dry 2D foam in a Hele–Shaw cell it is possible to measure the stresses due
to the bubble edges. Their curvature can be measured by fitting them to circular arcs,
but if the foam is at, or close to, equilibrium (quasi-static case), the stress depends
only weakly on the curvatures, and it is sufficient to determine the position of the
vertices and treat the edges as if they were straight [32].
4 Exercises
4.1 Measurement of the average liquid fraction of a foam

A U-tube can give an idea of the average liquid fraction of a foam, using Archimedes’
principle. If the bottom of the tube is filled with the foaming solution and foam is
placed on one side of the tube, the liquid on the other side rises.
For a foam with volume Vfoam = 1 mL in a tube of diameter 3 mm, what is the
liquid fraction φl if the liquid on the other side rises to a height of 8.5 mm?
4.2 Pressure in the Potts model

The free enthalpy expression G on which the Potts model is based (called for simplicity
the energy in §2.1.1) for the whole foam is

(Vk − Vko )2
G = γd2p [1 − δ(ni , nj )] + K ,
Vko
pixel i neighbours j bubbles k
where dp is the edge length of a pixel, ni is the index of pixel i, δ(ni , nj ) = 1 if ni = nj

and 0 otherwise, Vko = nk RT /Pext is the reference volume of bubble k, and K is a
physical constant to be determined.
(a) Show that this expression does correspond, in the limit of small variations in
volume, to the free enthalpy of the foam, and give the value of K. You should
write the free enthalpy of the foam as the sum of the gas and liquid enthalpy and
the interfacial energy. Then consider a transformation in which the volume of the
bubbles changes from Vko to Vk at constant external temperature and pressure,
assume that the gas in the foam is a perfect gas, and calculate the change in free
enthalpy during the transformation.
In practice, the foam structure that is obtained doesn’t depend on the value of
K chosen, although its value must be sufficiently large that the differences (Vk −
Vko )/Vko are much less than 1 (barely compressible foam). For numerical reasons,
we are limited to a value of K much smaller than the value found previously and
we are able to confirm that (Vk − Vko )/Vko 1.
(b) Give the pressure Pk in bubble k as a function of Vk , Vko , and K, and show that
Pk = −∂G/∂Vk in the limit of small variations in volume.
References
[1] A.W. Adamson and A.P. Gast, Physical Chemistry of Surfaces, Wiley, New
York, 6th ed., 1997.
[2] M.P. Allen and D.J. Tildesley, Computer Simulation of Liquids, Oxford
University Press, Oxford, 2002.
[3] M. Axelos and F. Boué, Langmuir, 19, 6598, 2003.
[4] R. Bandyopadhyay, A.S. Gittings, S.S. Suh, P.K. Dixon, and D.J. Durian,
Rev. Sci. Instr., 76, 093110, 2005.
[5] J. Banhart, J. Metals, 52, 22, 2000.
[6] G.B. Bantchev and D.K. Schwartz, Langmuir, 19, 2673, 2003.
[7] V. Bergeron, J. Phys: Condens. Matter, 11, R215, 1999.
[8] F. Bolton and D. Weaire, Phil. Mag. B, 65, 473, 1992.
[9] K. Brakke, Exp. Math., 1, 141, 1992.
[10] I. Cantat and R. Delannay, Eur. Phys. J. E, 18, 55, 2005.
Eur. Phys. J. E, 27, 123, 2008.
[12] L. Cipelletti, H. Bissig, V. Trappe, P. Ballesta, and S. Mazoyer, J. Phys:
Condens. Matter, 152, S257, 2003.
[13] S. Cohen-Addad and R. Höhler, Phys. Rev. Lett., 86, 4700, 2001.
[14] S. Courty, B. Dollet, F. Elias, P. Heinig, and F. Graner, Europhys. Lett.,
64, 709, 2003; B. Dollet, F. Elias, F. Graner, Europhys. Lett., 88, 69901,
2009.
References 257

[16] B. Dollet and F. Graner, J. Fluid Mech., 585, 181, 2007.
[17] W. Drenckhan, H. Ritacco, A. Saint-Jalmes, D. Langevin, P. Mac-
Guiness, A. Saugey, and D. Weaire, Phys. Fluids, 19, 102101, 2007.
[18] D.J. Durian, Phys. Rev. E, 55, 1739, 1997.
[19] D.J. Durian, D.A. Weitz, and D.J. Pine, Science, 252, 686, 1991.
[20] J. Etrillard, M. Axelos, I. Cantat, F. Artzner, A. Renault, and F. Boué,
Langmuir, 21, 2229, 2005.
[21] K. Feitosa, S. Marze, A. Saint-Jalmes, and D.J. Durian, J. Phys.: Condens.
Matter, 17, 6301, 2005.
[22] K. Feitosa and D.J. Durian, Eur. Phys. J. E, 26, 309, 2008.
[23] B.S. Gardiner, B.Z. Dlugogorski, and G.J. Jameson, J. Non-Newton. Fluid
Mech., 92, 151, 2000.
[24] C.P. Gonatas, J.S. Leigh, A.G. Yodh, J.A. Glazier, and B. Prause, Phys.
Rev. Lett., 75, 573, 1995.
[25] F. Graner, B. Dollet, C. Raufaste, and P. Marmottant, Eur. Phys. J. E,
25, 349, 2008.
[26] F. Graner, Y. Jiang, E. Janiaud, and C. Flament, Phys. Rev. E, 63, 011402,
2001.
[27] P. Hébraud, F. Lequeux, J.P. Munch, and D.J. Pine, Phys. Rev. Lett., 78,
4657, 1997.
[28] S. Hilgenfeldt and A. van Doornum, APS Meeting Abstracts, 13014, 2003.
[29] E.A. Holm, J.A. Glazier, D.J. Srolovitz, and G.S. Grest, Phys. Rev. A, 43,
2662, 1991.
[30] http://www.compucell3d.org/
[31] http://www.susqu.edu/brakke/evolver/
[32] E. Janiaud and F. Graner, J. Fluid Mech., 532, 243, 2005; S. Ishihara and K.
Sugimura, J. Theor. Biol., 313, 201, 2012; S. Ishihara, K. Sugimura, S.J. Cox,
I. Bonnet, Y. Bellaı̈che, and F. Graner, Euro. Phys. J. E, in press, 2013.
[33] S. Jurine, S. Cox, and F. Graner, Colloids Surf. A, 263, 18, 2005.
[34] N. Kern, D. Weaire, A. Martin, S. Hutzler, and S.J. Cox, Phys. Rev. E,
70, 041411, 2004.
2004.
[36] J. Lambert, I. Cantat, A. Renault, F. Graner, J.A. Glazier, I. Vere-
tennikov, and P. Cloetens, Colloids Surf. A, 263, 295, 2005.
[37] J. Lambert, R. Mosko, I. Cantat, P. Cloetens, J.A. Glazier, F. Graner,
and R. Delannay, Phys. Rev. Lett., 104, 248304, 2010.
[38] E. Lorenceau, Y. Yip Cheung Sang, R. Höhler, and S. Cohen-Addad,
Phys. Fluids, 18, 097103, 2006.
[39] C.W. Macosko, Rheology: Principles, Measurements, and Applications, Wiley-
VCH, Weinheim, 1994.
[40] S. Marze, D. Langevin, and A. Saint-Jalmes, J. Rheol., 52, 1091, 2008.
[41] E.B. Matzke, Am. J. Botany, 33, 58, 1946.
[42] C. Monnereau and M. Vignes-Adler, Phys. Rev. Lett., 80, 5228, 1998.
[43] K.J. Mysels, Langmuir, 2, 428, 1986.
[44] K.J. Mysels and M.N. Jones, Discuss. Faraday Soc., 42, 42, 1966.
[45] R. Nagarajan, S.I. Chung, and D.T. Wasan, J. Colloid Interface Sci., 204,
53, 1998.
[46] T. Okuzono and K. Kawasaki, Phys. Rev. E, 51, 1246, 1995.
[47] O. Pitois, C. Fritz, and M. Vignes-Adler, J. Colloid Interface Sci., 282, 458,
2005.
23, 217, 2007.
[49] S. Rodts, J.C. Baudez, and P. Coussot, Europhys. Lett., 69, 636, 2005.
[50] D. Sentenac and J.-J. Benattar, Phys. Rev. Lett., 81, 160, 1998.
[51] D.J. Srolovitz, M.P. Anderson, G.S. Grest, and P.S. Sahni, Scripta metall.,
17, 241, 1983.
[52] C. Stenvot and D. Langevin, Langmuir, 4, 1179, 1988.
[53] H.A. Stone, A.D. Stroock, and A. Ajdari, Ann. Rev. Fluid Mech., 36, 381,
2004.
[54] C. Stubenrauch and R. von Klitzing, J. Phys.: Condens. Matter, 15, R1197,
2003.
[55] G.L. Thomas, R.M.C. de Almeida, and F. Graner, Phys. Rev. E, 74, 021407,
2006.
[56] A. van der Net, L. Blondel, A. Saugey, and W. Drenckhan, Colloids Surf.
A, 309, 159, 2007.
[57] M.U. Vera, A. Saint-Jalmes, and D.J. Durian, Appl. Opt., 40, 4210, 2001.
[58] G.F. Voronoı̈, J. Reine Angew. Math., 133, 97, 1908; 134, 198, 1908; 136, 67,
1909.
[59] D. Weaire and R. Phelan, Phil. Mag. Lett., 69, 107, 1994.
[60] D.A. Weitz and D.J. Pine, “Diffusing-wave spectroscopy”, in Dynamic Light
Scattering, Chap. 16, Clarendon, Oxford, 1993.
[61] C. Zakri, A. Renault, and B. Berge, Physica B, 248, 208, 1998.
Notation
Note that page numbers are not provided for symbols that are used throughout the text
c concentration
cmc critical micelle concentration 146
cs bulk concentration below the interface 85
d average bubble diameter
dm typical size of a surfactant molecule 146
e thickness of a 2D foam (distance between
the plates) 57
ex , ey , ez reference unit vectors in the laboratory frame
f number of faces of a bubble 32
g = 9.8 m.s−2 gravity, in the direction −ez
h thickness of a film 20
kB = 1.38 × 10−23 J.K−1 Boltzmann’s constant
length of a Plateau border 46
ij length of the edge separating bubbles i and j in 2D 35
∗ transport mean free path of light 236
n number of edges of a face (in 3D);
number of sides of a bubble (in 2D) 31
n unit vector normal to a surface
n̂ unit vector normal to a line, in the plane
of the interface
p pressure in the liquid phase 48
q geometric charge 41
qt topological charge 32
r radius of curvature of the cross-section
of a Plateau border 47
r1 , r2 radii of curvature of a Plateau border 47
s surface area of the cross-section of a Plateau
border 48
u local velocity in the liquid phase
v velocity of a bubble
x, y horizontal directions
z vertical direction, oriented upwards
A surface area of a 2D bubble 38

Bij Finger tensor 174
Bo = ηs /(ηr) Boussinesq number 120
260 Notation
Ca = ηu/γ capillary number 190

C r = (dV /dt)/V 1/3 relative growth rate of a bubble 102
Dv bulk diffusion coefficient of a surfactant 85
Df bulk diffusion coefficient of a gas through a liquid 109
Deff effective diffusion coefficient of gas across a film 109
E energy of a foam 35
EGM Gibbs–Marangoni elastic modulus 87
Es∗ = Es + iEs complex surface dilatational elastic modulus 88
Fij deformation gradient tensor 174
G static shear modulus 169
Ĝ static shear modulus in 2D (in N.m−1 ) 180
G∗ = G + iG complex shear modulus 173
G∗s = Gs + iGs complex surface shear modulus 88
H = 1/R1 + 1/R2 mean curvature of a film 26
He Henry’s coefficient 109
J compliance 172
Kp reduced permeability of a Plateau border 117
Lint total length of interface in a 2D foam
(E = λLint ) 59
N number of bubbles in a foam
N1 , N2 normal stress differences 175
P gas pressure in a bubble
Patm = 105 Pa atmospheric pressure
Pc capillary pressure 48
R radius of curvature of a film 25
Rij radius of curvature of the edge separating
bubbles i and j in 2D 40
R1 and R2 principal radii of curvature of a film 25
R32 Sauter mean radius 251
RS equivalent spherical radius of a bubble with
surface area S 38
RV equivalent spherical radius of a bubble with
volume V 38
Si surface area of bubble i 35
Sij area ofthe face separating bubbles i and j
(E = i<j 2γSij ) 27
Sint total area of air/water interface in a foam
(E = γSint ) 35
T temperature
T1 type 1 topological transformation 78
T2 type 2 topological transformation 81
V bubble volume
Vm molar volume of gas 109
Vr typical volume of a topological change 205
X displacement field 174
Notation 261
α √ foam permeability 118

δa = 3 − π/2 geometric coefficient 124
δb = 1.74 geometric coefficient 124
γ surface tension of a single interface 23
γ0 = 72 × 10−3 N.m−1 surface tension of pure water in air 23
ε shear strain 168, 174
ε̇ = dε/dt strain rate 169
εij strain tensor 174
εy yield strain 171
η bulk viscosity, η = 10−3 Pa.s of water at 20 ◦ C 169
ηd surface dilatational viscosity 87
ηp plastic viscosity 171
ηs surface shear viscosity 88
κij = 1/Rij curvature of the edge separating bubbles
i and j in 2D 40
λ ∼ γe line tension of an interface in 2D 59
λc = γ/ρl g capillary length 54
λD Debye length 93
μ chemical potential 83
μf2 = f 2 − f 2 variance in the number of faces 250
μV2 = V 2 − V 2 variance in the bubble volumes 252
ρ density of a foam 21
ρl density of a liquid 21
σ shear stress 168
σ̂ shear stress in 2D (in N.m−1 ) 180
σij stress tensor 174
s
σij surface stress tensor 87
σy yield stress 171
φl liquid fraction (volume fraction of the liquid
phase of a foam) 21
φ∗l critical liquid fraction 195
Γ surface concentration of surfactant 82

Γ∞ saturated surface concentration of surfactant 85
Πo osmotic pressure of a foam 51
Πd disjoining pressure 48
Πl = γ0 − γ Langmuir surface pressure 83
Index
2D (foam) 55, 65, 254 of gas 100, 102, 109

of surfactants 85
A dilation (see expansion)
adsorption 18, 82, 84, 90, 98, 148, 161 dimple 92, 136
antifoam/antifoaming agent 140 dissipation 126, 151, 186–190, 209,
avalanche 140 211, 246
drainage
B coupled with coarsening 133
film 135, 152
Batchelor 175, 218 equation 127
biological tissue 12 forced 115, 129, 157
black film 137, 155, 229 free 114, 132, 156, 158, 162
common 98 characteristic time 132
Newton 98 DTS (see multiple light scattering)
Boussinesq number 120–123 DWS (see multiple light scattering)
bubbly liquid 21, 76
E
C
edge (see Plateau border)
capillary elasticity
length 54 foam 178, 194, 198, 202
number 190–193, 213 surface/interfacial 86, 88, 91, 136
pressure 48, 97, 98, 135 Gibbs (see Gibbs)
suction 20, 48, 97 Langmuir (see Langmuir)
charge elastic limit 183
electrostatic 94 elongation 62, 178, 220
geometric 41–44, 100, 103 emulsion 11, 30, 51, 76, 77, 147, 237
topological 32, 43, 161 energy
close packing 21, 46 dissipated 151, 187
cmc (see micelle) elastic 169, 187, 208
coalescence (see rupture) local minimum 22, 36, 180, 184
coarsening 99 surface 17, 23, 35, 72, 197
Mullins 44, 102, 104 Euler’s formula 31, 32, 43, 71, 81
Lifshitz–Slyozov–Wagner 76 expansion/compression 87, 211, 226
Ostwald ripening 76
von Neumann 42, 100, 151, 161 F
characteristic time 104, 108
coupling with drainage 133 ferrofluid 149
colour in films (see iridescence) film balance technique 136, 229
conductivity 115, 158, 238 fluctuations 26, 91, 137, 215
convective instability 128 foamability 82, 98, 231
creep 172, 202, 205 G
curvature
mean 27, 30, 102 Gibbs
radius of 27, 48 elasticity 87
equation 83
D
H
Darcy’s law 117
Debye length 93 Hamaker (constant) 92
diffusion hexagonal structure 12, 61, 179, 184,
coefficient 85, 102, 109, 152 196, 219
264 Index
hole (opening in film) 138 Poiseuille flow 117

honeycomb (see hexagonal structure) polymer 10, 147
hydrophilic/hydrophobic 18, 147 pressure
hysteresis 184 capillary (see capillary)
I disjoining 19, 20, 49, 92, 94, 97, 112, 230
osmotic 21, 51
idealized bubble (with zero mean curvature) surface 24, 83, 162
37, 44, 103
ideal foam 26, 30, 31, 59, 100 Q
interface
mobile/immobile 90, 120, 122, 130, 136 quasi-static 62, 170, 178, 183, 186,
compressible/incompressible 188–193, 198, 245
213
iridescence 20, 67, 129, 228 R
K rheometer 210, 214

Reynolds
Kelvin cell 33, 50, 68, 181, 183 number 116
L velocity 135
rupture 78, 134, 140
Langmuir
monolayer 58 S
trough 227
equation 83 Sauter mean radius 194, 251
foam 56, 58 SDS 19, 69, 110, 146, 211, 225
Laplace Law (see Young–Laplace Law) self-similar (growth) 105, 161
Lewis relation 38 shear 169, 172, 240
liquid fraction experiment 172, 210, 228, 241
2D 60 modulus 89, 88, 169, 173, 180–183
3D 21, 110–115, 238 surface 86, 88, 120, 228
localization of flow 214, 215 rate 88, 169, 172, 211
length solid foam 8
Debye (see Debye)
stability (see foamability)
capillary (see capillary)
strain 168, 170, 173, 181, 240
local minimum (see energy)
surface 87, 88, 121
M rate 171
yield 171, 183, 210, 217
magnetic resonance 215, 235
stratification 95
Marangoni effect 91
marginal regeneration 136, 155 stress 168, 174, 177, 215, 240
micelle 23, 83, 146, 229 deviatoric 174
MRI (see magnetic resonance) normal 174, 198, 198
Mullins (see coarsening) surface 87, 121
multiple light scattering 236 yield 171, 208–211, 218
suction (see capillary)
N surface
NMR (see magnetic resonance) elasticity (see elasticity)
node (see vertex) energy (see energy)
pressure (see pressure)
O strain (see strain)
Ostwald ripening (see coarsening) stress (see stress)
tension (see tension)
P viscosity (see viscosity)
permeability 118, 119 surface area
plasticity 62, 170, 183, 209, 215, 220 effective 112
Plateau minimal 19, 244
border 21, 47, 188, 233 specific 6, 119, 127
laws 20, 26, 30, 46, 59, 70, surfactant/surface-active molecule 18, 24, 63,
180, 183 82, 131, 145, 153, 190, 225
Index 265
T V
T1 36, 78, 183, 186, 205, 208, 211, 215, 217, van der Waals 19, 92, 94
245, 246, 254 vertex 21, 30, 46, 125, 126, 129, 132, 194, 219,
T2 81, 104, 151, 254 234, 247
tension viscosity
film 26, 69 foam 169, 171, 211
line 42, 59, 178 surface 86, 87, 120, 152, 190, 204
surface 17, 23, 27, 64, 82, 91, 225 von Neumann’s law (see coarsening)
(characteristic) time
of drainage (see drainage) W
of coarsening (see coarsening)
Weaire–Phelan 37, 244
tissue (see biological)
tomography 2, 21, 234, 235 Y
topological
change (see T1) yield
charge (see charge) stress (see stress)
rearrangement (see T1) strain (see strain)
topology 31, 38, 100, 250 Young–Laplace law 17, 25, 26, 69, 218

Foams - Structure and Dynamics-Oxf PDF

Uploaded by

Copyright:

Available Formats

Foams - Structure and Dynamics-Oxf PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Foams - Structure and Dynamics-Oxf PDF

Uploaded by

Copyright:

Available Formats

This page intentionally left blank

I. Cantat, S. Cohen-Addad, F. Elias, F. Graner, R. Höhler,

The bubbly world of foams

: Materials generally found in a kitchen or bathroom

: cheap, less than about 5 dollars, euros, or pounds

5 2D and quasi-2D foams 55

5 Experimental and numerical methods 225

1 The foams around us

1.2 On your plate and in your glass

Fig. 1.2: Three-dimensional imaging experiment of the air bubbles in a chocolate

1.3 Detergents and cosmetics

1.4 Spontaneous or undesirable foams

2.1 Physico-chemical constituents

2.2 Geometrical and physical properties

2.3 Mechanical properties

3 What are foams used for?

3.1 Desirable functions

3. Isolate, conﬁne, smother

6. Give a ﬂuid the behaviour of a solid

3.2 Mineral ﬂotation

Circulation of ground Recirculation

4 Solid foams and other cellular systems

polyethylene, polyvinylchloride), which is already polymerized and hardens on cooling.

4.2 Other cellular structures

A biological tissue (epithelium) is a cellular arrangement in two dimensions. In a

Diamant Fullerène Nanotube

Phenomena demonstrated: making a foam by whisking, foamability, liquid fraction,

Phenomena demonstrated: making a foam by shaking, foam disorder, loss of transpar-

Phenomena demonstrated: making a foam by beating and blowing, foam disorder,

5.2 Chocolate mousse

diﬃculty level: cost of materials:

Phenomena demonstrated: making a foam by whisking, foamability, liquid fraction,

Comments: The variations are countless. You could:

1 Description at all length-scales

1.1 At the scale of a gas/liquid interface

The shape of a liquid/liquid or gas/liquid interface is determined by the existence of

From interface to ﬁlm: the role of surfactants Whether it is a foam in a river

1.2 At the scale of a ﬁlm

Film thickness A newly created ﬁlm is not generally at equilibrium. It is a few

1.3 At the scale of a bubble

1.4 At the scale of a foam

2 Local equilibrium laws

2.1 Equilibrium of ﬂuid interfaces

dEsurf = γdS, (2.1)

where γ is deﬁned as the surface (or superﬁcial) tension [22].

Fig. 2.5: Experimental measurement of surface tension. A soap ﬁlm is formed by

2.2 Plateau’s laws

Hij + Hjk + Hki = 0. (2.5)

This is a strong constraint on the equilibrium structure of a foam.

a – Force balance on half a bubble

b – Free energy balance over the whole bubble

Summary of Plateau’s laws

2.2.3 Condition at solid walls It is necessary to add to these laws a condition

3.1 Number of neighbours: topology

In two dimensions, Euler’s formula is

3.1.2 Average number of neighbours In the shaded section below we show a

Consequences of Euler’s formula and Plateau’s laws in 2D [30, 62].

Then Euler’s formula in 2D (eq. (2.9)) gives

n n χEuler

The overpressure in a bubble is of the order of 4γ/RV Patm . Consequently, P ≈ Patm